Advanced Datasheet: GR740 LEON 4FT Processor (2016)

GR740
Quad Core LEON4 SPARC V8 Processor
2016 Preliminary Data Sheet and User’s Manual
The most important thing we build is trust
Features
• Fault-tolerant quad-processor SPARC V8 integer unit
with 7-stage pipeline, 8 register windows, 4x4 KiB
instruction and 4x4 KiB data caches.
• Double-precision IEEE-754 floating point units
• 2 MiB Level-2 cache
• 64-bit PC100 SDRAM memory interface with ReedSolomon EDAC
• 8/16-bit PROM/IO interface with EDAC
• SpaceWire router with eight SpaceWire links
• 2x 10/100/1000 Mbit Ethernet interfaces
• PCI Initiator/Target interface
• MIL-STD-1553B interface
• 2x CAN 2.0 controller interface
• 2x UART, SPI, Timers and watchdog, 16+22 pin GPIO
• CPU and I/O memory management units
• Multi-processor interrupt controller with support for
asymmetric and symmetric multiprocessing
• SpaceWire Time Distribution Protocol controller and
support for time synchronisation
Description
The GR740 device is a radiation-hard system-onchip featuring a quad-core fault-tolerant LEON4
SPARC V8 processor, eight port SpaceWire router,
PCI initiator/target interface, MIL-STD-1553B
interface, CAN 2.0 interfaces and 10/100/1000
Mbit Ethernet interfaces.
Specification
• System frequency: 250 MHz (estimated worst
case pending characterisation)
• Main memory interface: PC100 SDRAM
• SpaceWire router with SpaceWire links: 200
Mbit/s minimum
• 33/66 MHz (TBC) PCI 2.3 initiator/target 
interface
• Ethernet 10/100/1000 Mbit MACs
• CLGA625 package
32-bit APB
S
IRQ(A)MP
96-bit
PC100
SDRAM
PROM
IO
8/16-bit
SDRAM
CTRL w.
EDAC
PROM
& IO
CTRL w.
EDAC
S
32-bit APB
S
S Statistics
LEON4
STAT.UNIT
Memory
Scrubber
S M
Memory bus
128-bit AHB
Caches
L2
Cache
MX
S
S S
AHB/APB
Bridges
S
M
S
Slave IO bus
32-bit AHB
S
AHB
Status
S
S
Clock gating S
unit
Temperature S
sensor
M
FPU
X
MMU
MMU
LEON4
Caches
MX
Processor bus MX
S 128-bit AHB
AHB/AHB
Bridge
M
M
AHB Bridge
M
IOMMU
S
S
M
PCI
DMA
S
S
M
PCI
Target
S
GPIO port
0-1
S
UART
S
Bootstrap
GP register
S
TDP
controller
S M
SpW router
S
S
S
Timer unit 0
watchdog
MMU
MX
PCI
Master
Timer units
1-4
FPU
LEON4
Caches
X
Pad / PLL
controller
FPU
AHB/AHB S
Bridge
M
M
AHB/APB
Bridge
S
S
Debug bus
S
SpW RMAP
DCL
M
DSU4
32-bit AHB
S
X
AHB
Status
S
M
JTAG
DCL
AHBTRACE
X
Master IO bus
32-bit AHB
M
M
CAN
MIL-STD
Controller
1553B
S S
S
M
M
Ethernet
S S
S
SPI
controller
Applications
The GR740 device is targeted at high-performance general purpose 
processing. The architecture is suitable for both symmetric and 
asymmetric multiprocessing. Shared resources can be monitored to 
support mixed-criticality applications.
GR740-UM-DS, Jun 2016, Version 1.4
www.cobham.com/gaisler
GR740
Table of contents
1
Introduction.............................................................................................................................. 8
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10
2
Architecture............................................................................................................................ 14
2.1
2.2
2.3
2.4
2.5
3
Clock inputs........................................................................................................................................... 34
Clock loop for SDRAM ........................................................................................................................ 34
Reset scheme ......................................................................................................................................... 35
Clock multiplexing for main system clock, SDRAM and SpaceWire .................................................. 36
PLL control and configuration .............................................................................................................. 37
PLL watchdog ....................................................................................................................................... 38
PCI clock ............................................................................................................................................... 38
MIL-STD-1553B clock ......................................................................................................................... 38
Clock gating unit ................................................................................................................................... 38
Debug AHB bus clocking...................................................................................................................... 39
Notes on Ethernet interface clock and mode switch ............................................................................. 39
Technical notes....................................................................................................................... 40
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
6
Bootstrap signals ................................................................................................................................... 25
Pin multiplexing .................................................................................................................................... 26
Complete signal list ............................................................................................................................... 29
Clocking and reset.................................................................................................................. 34
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
5
Overview ............................................................................................................................................... 14
Cores...................................................................................................................................................... 16
Memory map ......................................................................................................................................... 17
Interrupts ............................................................................................................................................... 20
Plug & play and bus index information................................................................................................. 20
Signals.................................................................................................................................... 25
3.1
3.2
3.3
4
Scope ....................................................................................................................................................... 8
Preliminary data sheet limitations ........................................................................................................... 8
Updates and feedback.............................................................................................................................. 8
Software support...................................................................................................................................... 8
Development board ................................................................................................................................. 8
Reference documents .............................................................................................................................. 9
Document revision history .................................................................................................................... 10
Acronyms .............................................................................................................................................. 11
Definitions ............................................................................................................................................. 12
Register descriptions ............................................................................................................................. 13
GRLIB AMBA plug&play scanning ..................................................................................................... 40
Processor register file initialisation and data scrubbing ........................................................................ 40
PROM-less systems and SpaceWire RMAP ......................................................................................... 40
System integrity and debug communication links ................................................................................ 41
ASMP configurations ............................................................................................................................ 41
Clock gating .......................................................................................................................................... 41
Software portability............................................................................................................................... 42
Level-2 cache ........................................................................................................................................ 42
Time synchronisation ............................................................................................................................ 43
LEON4 - Fault-tolerant High-performance SPARC V8 32-bit Processor ............................. 45
6.1
6.2
Overview ............................................................................................................................................... 45
LEON4 integer unit ............................................................................................................................... 47
GR740-UM-DS, Jun 2016, Version 1.4
2
www.cobham.com/gaisler
GR740
6.3
6.4
6.5
6.6
6.7
6.8
6.9
6.10
6.11
7
Floating-point Control Unit ................................................................................................... 76
7.1
7.2
7.3
8
Overview ............................................................................................................................................. 116
Operation ............................................................................................................................................. 116
Registers .............................................................................................................................................. 119
LEON4 Hardware Debug Support Unit............................................................................... 120
13.1
13.2
13.3
13.4
13.5
13.6
14
Overview ............................................................................................................................................. 107
Operation ............................................................................................................................................. 107
Registers .............................................................................................................................................. 110
Bridge connecting Debug AHB bus to Processor AHB bus ................................................ 116
12.1
12.2
12.3
13
Overview ............................................................................................................................................... 96
Operation ............................................................................................................................................... 96
Limitations............................................................................................................................................. 96
SDRAM back-end operation ................................................................................................................. 96
Fault-tolerant operation ....................................................................................................................... 100
Registers .............................................................................................................................................. 103
Memory Scrubber and AHB Status Register ....................................................................... 107
11.1
11.2
11.3
12
Overview ............................................................................................................................................... 82
Operation ............................................................................................................................................... 82
Operation ............................................................................................................................................... 85
Registers ................................................................................................................................................ 88
SDRAM Memory Controller with Reed-Solomon EDAC .................................................... 96
10.1
10.2
10.3
10.4
10.5
10.6
11
Overview ............................................................................................................................................... 78
Functional description ........................................................................................................................... 78
Level 2 Cache controller ........................................................................................................ 82
9.1
9.2
9.3
9.4
10
Floating-Point register file..................................................................................................................... 76
Floating-Point State Register (FSR)...................................................................................................... 76
Floating-Point Exceptions and Floating-Point Deferred-Queue ........................................................... 76
High-performance IEEE-754 Floating-point Unit ................................................................. 78
8.1
8.2
9
Cache system ......................................................................................................................................... 53
Memory management unit..................................................................................................................... 56
Floating-point unit................................................................................................................................. 57
Co-processor interface........................................................................................................................... 58
AMBA interface .................................................................................................................................... 58
Multi-processor system support ............................................................................................................ 60
ASI assignments .................................................................................................................................... 61
Configuration registers .......................................................................................................................... 66
Software considerations ........................................................................................................................ 74
Overview ............................................................................................................................................. 120
Operation ............................................................................................................................................. 120
AHB Trace Buffer ............................................................................................................................... 121
Instruction trace buffer ........................................................................................................................ 123
DSU memory map............................................................................................................................... 124
DSU registers ...................................................................................................................................... 126
JTAG Debug Link with AHB Master Interface ................................................................... 136
14.1
14.2
14.3
Overview ............................................................................................................................................. 136
Operation ............................................................................................................................................. 136
Registers .............................................................................................................................................. 137
GR740-UM-DS, Jun 2016, Version 1.4
3
www.cobham.com/gaisler
GR740
15
SpaceWire codec with AHB host Interface and RMAP target............................................. 138
15.1
15.2
15.3
15.4
15.5
15.6
15.7
15.8
15.9
16
AHB Trace buffer tracing Master I/O AHB bus .................................................................. 161
16.1
16.2
16.3
17
Overview ............................................................................................................................................. 227
Bridge operation .................................................................................................................................. 227
General access protection and address translation .............................................................................. 230
Access Protection Vector..................................................................................................................... 231
IO Memory Management Unit (IOMMU) functionality..................................................................... 233
Fault-tolerance..................................................................................................................................... 236
Statistics............................................................................................................................................... 237
ASMP support ..................................................................................................................................... 237
Registers .............................................................................................................................................. 238
Gigabit Ethernet Media Access Controller (MAC) ............................................................. 248
19.1
19.2
19.3
19.4
19.5
19.6
19.7
19.8
20
Overview ............................................................................................................................................. 166
Operation ............................................................................................................................................. 166
SpaceWire ports................................................................................................................................... 176
AMBA ports ........................................................................................................................................ 178
Configuration port ............................................................................................................................... 201
IOMMU - Bridge connecting Master I/O AHB bus ............................................................ 227
18.1
18.2
18.3
18.4
18.5
18.6
18.7
18.8
18.9
19
Overview ............................................................................................................................................. 161
Operation ............................................................................................................................................. 161
Registers .............................................................................................................................................. 162
SpaceWire router.................................................................................................................. 166
17.1
17.2
17.3
17.4
17.5
18
Overview ............................................................................................................................................. 138
Operation ............................................................................................................................................. 138
Link interface ...................................................................................................................................... 139
Time-Code distribution ....................................................................................................................... 141
Receiver DMA channels...................................................................................................................... 141
Transmitter DMA channels ................................................................................................................. 146
RMAP.................................................................................................................................................. 150
AMBA interface .................................................................................................................................. 154
Registers .............................................................................................................................................. 155
Overview ............................................................................................................................................. 248
Operation ............................................................................................................................................. 248
Tx DMA interface ............................................................................................................................... 249
Rx DMA interface ............................................................................................................................... 251
MDIO Interface ................................................................................................................................... 254
Ethernet Debug Communication Link (EDCL) .................................................................................. 254
Media Independent Interfaces ............................................................................................................. 256
Registers .............................................................................................................................................. 257
32-bit PCI/AHB bridge ........................................................................................................ 262
20.1
20.2
20.3
20.4
20.5
20.6
20.7
20.8
Overview ............................................................................................................................................. 262
Configuration....................................................................................................................................... 262
Operation ............................................................................................................................................. 268
PCI Initiator interface.......................................................................................................................... 271
PCI Target interface............................................................................................................................. 272
DMA Controller .................................................................................................................................. 273
PCI trace buffer ................................................................................................................................... 275
Interrupts ............................................................................................................................................. 276
GR740-UM-DS, Jun 2016, Version 1.4
4
www.cobham.com/gaisler
GR740
20.9
20.10
21
MIL-STD-1553B / AS15531 Interface ................................................................................ 285
21.1
21.2
21.3
21.4
21.5
21.6
21.7
21.8
22
Overview ............................................................................................................................................. 341
Operation ............................................................................................................................................. 341
Registers .............................................................................................................................................. 342
Multiprocessor Interrupt Controller with extended ASMP support..................................... 346
26.1
26.2
26.3
27
Overview ............................................................................................................................................. 331
PROM access ...................................................................................................................................... 331
Memory mapped IO ............................................................................................................................ 333
8-bit and 16-bit PROM access............................................................................................................. 334
8- and 16-bit I/O access....................................................................................................................... 335
Burst cycles ......................................................................................................................................... 335
Memory EDAC ................................................................................................................................... 336
Bus Ready signalling........................................................................................................................... 336
Registers .............................................................................................................................................. 338
General Purpose Timer Units............................................................................................... 341
25.1
25.2
25.3
26
Overview ............................................................................................................................................. 327
Operation ............................................................................................................................................. 327
Registers .............................................................................................................................................. 330
Fault-tolerant 8/16-bit PROM/IO Memory Interface .......................................................... 331
24.1
24.2
24.3
24.4
24.5
24.6
24.7
24.8
24.9
25
Overview ............................................................................................................................................. 307
Interface............................................................................................................................................... 308
Protocol ............................................................................................................................................... 308
Status and monitoring.......................................................................................................................... 308
Transmission........................................................................................................................................ 308
Reception............................................................................................................................................. 311
Global reset and enable ....................................................................................................................... 314
Interrupt ............................................................................................................................................... 314
Registers .............................................................................................................................................. 315
Memory mapping ................................................................................................................................ 325
Bridge connecting Slave I/O AHB bus to Processor AHB bus............................................ 327
23.1
23.2
23.3
24
Overview ............................................................................................................................................. 285
Electrical interface............................................................................................................................... 285
Operation ............................................................................................................................................. 286
Bus Controller Operation .................................................................................................................... 287
Remote Terminal Operation ................................................................................................................ 292
Bus Monitor Operation........................................................................................................................ 296
Clocking and reset ............................................................................................................................... 296
Registers .............................................................................................................................................. 297
CAN 2.0 Controllers with DMA.......................................................................................... 307
22.1
22.2
22.3
22.4
22.5
22.6
22.7
22.8
22.9
22.10
23
Reset .................................................................................................................................................... 276
Registers .............................................................................................................................................. 277
Overview ............................................................................................................................................. 346
Operation ............................................................................................................................................. 346
Registers .............................................................................................................................................. 350
General Purpose I/O Ports ................................................................................................... 360
27.1
27.2
Overview ............................................................................................................................................. 360
Operation ............................................................................................................................................. 360
GR740-UM-DS, Jun 2016, Version 1.4
5
www.cobham.com/gaisler
GR740
27.3
28
UART Serial Interfaces ........................................................................................................ 367
28.1
28.2
28.3
28.4
28.5
28.6
28.7
29
Overview ............................................................................................................................................. 403
Protocol ............................................................................................................................................... 403
Functionality........................................................................................................................................ 403
Registers .............................................................................................................................................. 409
AMBA AHB controller with plug&play support................................................................. 423
37.1
37.2
38
Overview ............................................................................................................................................. 397
Operation ............................................................................................................................................. 397
Registers .............................................................................................................................................. 399
SpaceWire - Time Distribution Protocol Controller ............................................................ 403
36.1
36.2
36.3
36.4
37
Overview ............................................................................................................................................. 396
Register Bank For I/O and PLL configuration registers ...................................................... 397
35.1
35.2
35.3
36
Overview ............................................................................................................................................. 395
Operation ............................................................................................................................................. 395
Registers .............................................................................................................................................. 395
Temperature sensor controller.............................................................................................. 396
34.1
35
Overview ............................................................................................................................................. 393
Operation ............................................................................................................................................. 393
Registers .............................................................................................................................................. 393
Register for bootstrap signals............................................................................................... 395
33.1
33.2
33.3
34
Overview ............................................................................................................................................. 386
Multiple APB interfaces ...................................................................................................................... 388
Registers .............................................................................................................................................. 389
AHB Status Registers .......................................................................................................... 393
32.1
32.2
32.3
33
Overview ............................................................................................................................................. 382
Operation ............................................................................................................................................. 382
Registers .............................................................................................................................................. 383
LEON4 Statistics Unit (Performance Counters).................................................................. 386
31.1
31.2
31.3
32
Overview ............................................................................................................................................. 372
Operation ............................................................................................................................................. 372
Registers .............................................................................................................................................. 375
Clock gating unit.................................................................................................................. 382
30.1
30.2
30.3
31
Overview ............................................................................................................................................. 367
Operation ............................................................................................................................................. 367
Baud-rate generation ........................................................................................................................... 368
Loop back mode .................................................................................................................................. 369
FIFO debug mode................................................................................................................................ 369
Interrupt generation ............................................................................................................................. 369
Registers .............................................................................................................................................. 370
SPI Controller supporting master and slave operation ........................................................ 372
29.1
29.2
29.3
30
Registers .............................................................................................................................................. 361
Overview ............................................................................................................................................. 423
Operation ............................................................................................................................................. 423
AMBA AHB/APB bridge with plug&play support ............................................................. 425
38.1
38.2
Overview ............................................................................................................................................. 425
Operation ............................................................................................................................................. 425
GR740-UM-DS, Jun 2016, Version 1.4
6
www.cobham.com/gaisler
GR740
39
Electrical description ........................................................................................................... 426
39.1
39.2
39.3
39.4
39.5
40
Absolute maximum ratings ................................................................................................................. 426
Operating conditions ........................................................................................................................... 426
Input voltages, leakage currents and capacitances .............................................................................. 426
Power supplies..................................................................................................................................... 427
AC characteristics................................................................................................................................ 427
Mechanical description ........................................................................................................ 442
40.1
40.2
40.3
40.4
Component and package ..................................................................................................................... 442
Package pinout diagram ...................................................................................................................... 442
Pin assignment..................................................................................................................................... 443
Package drawing.................................................................................................................................. 458
41
Temperature and thermal resistance..................................................................................... 459
42
Ordering information ........................................................................................................... 460
43
Errata.................................................................................................................................... 461
GR740-UM-DS, Jun 2016, Version 1.4
7
www.cobham.com/gaisler
GR740
1
Introduction
1.1
Scope
This document is the preliminary data sheet for the GR740 device. The GR740 was developed in an
activity funded by the European Space Agency and is part of the agency’s roadmap for standard
microprocessor components.
The work has been performed by Cobham Gaisler AB, located in Göteborg, Sweden. The architecture
has been reviewed by representatives from the European Space Agency and Airbus Defence and
Space (formerly EADS Astrium).
1.2
Preliminary data sheet limitations
Note that this document is a preliminary data sheet:
1.3
•
Advanced data sheet - Product in development
•
Preliminary data sheet - Shipping prototype
•
Data sheet - Shipping space-grade product
Updates and feedback
Updates are available at http://www.gaisler.com/gr740
Feedback can be sent to Cobham Gaisler AB support: support@gaisler.com
For commercial questions please contact sales@gaisler.com
1.4
Software support
The GR740 design is supported by standard toolchains provided by Cobham Gaisler. Toolchains can
be downloaded from http://www.gaisler.com.
1.5
Development board
A development board with the GR740 device is available. Please see http://www.gaisler.com/gr-cpcigr740.
GR740-UM-DS, Jun 2016, Version 1.4
8
www.cobham.com/gaisler
GR740
1.6
Reference documents
[AMBA]
AMBA Specification, Rev 2.0, ARM Limited
[SPARC]
The SPARC Architecture Manual, Version 8, SPARC International Inc.
[V8E]
SPARC-V8 Supplement, SPARC-V8 Embedded (V8E) Architecture
Specification, SPARC-V8E, Version 1.0, SPARC International Inc.
[CCSDS]
Time Code Formats, CCSDS 301.0-B-4, Blue Book, Issue 4, November
2010 www.CCSDS.org
[SPW]
Space engineering: SpaceWire - Links, nodes, routers and networks,
ECSS-E-ST-50-12C, July 2008
[RMAP]
Space engineering: SpaceWire - Remote memory access protocol, ECSSE-ST-50-52C, February 2010
[SPWCUC]
High Accuracy Time Synchronization over SpaceWire Networks, SPWCUC-REP-0003, Version 1.1, September 2012
[SPWID]
Space engineering: SpaceWire Protocol Identification, ECSS-E-ST-5051C, 5 February 2010
[SPWPNP]
Space Engineering: SpaceWire Plug-and-Play protocol, ECSS-E-ST-5054C, Draft, March 2013
[SPWD]
SpaceWire-D - Deterministic Control and Data Delivery over SpaceWire
Networks, Draft B, April 2010, ESA Contract Number 220774-07-NL/
LvH
[SPWINT]
Yuriy Sheynin, Distributed Interrupts in SpaceWire Interconnections,
International SpaceWire Conference, Nara, November 2008 (outdated)
GR740-UM-DS, Jun 2016, Version 1.4
9
www.cobham.com/gaisler
GR740
1.7
Document revision history
Change record information is provided in table 1.
Table 1. Change record
Version
Date
Note
1.0
2015 April
First public release of GR740 document.
1.1
2015 November Fix typo of CE/NE bit in AHBSTAT section.
Clarify that Level-2 cache is unified.
Correct L4STAT section 31.1 to state that the unit has sixteen counters.
Correct GRSPWROUTER documentation: Error in the description of the ICODEGEN register. The UA bit is independent of the setting of the AH bit. It is not required for AH to be set
in order for UA to have effect.
Corrected AHBTRACE TIMETAG register APB address offset in table caption.
Corrected MEMSCRUB APB address offsets in table captions for two last range registers.
Corrected SPICTRL MASK register access attributes.
Added missing reset values for L2C Scrub delay register and Access control register.
Document TCTRL register WS and WN fields in timer unit section.
Correct reset value for LEON4 %asr17.DBP, CCTRL.DS and %tbr.
Update pinlist in section 40.3
Updated front page and back page.
Converted to new headers and footers.
Corrected description for EDCL 1 bootstrap signals (GPIO[5:4])
Corrected register table headings and add value for trace buffer FDEPTH field in GRPCI2
section 20.
1.2
2016 January
Correct name of TOV field in DSU Instruction trace buffer control register 1
Add note about pulsed interrupts in interrupt controller section
Update footer
1.3
2016 February
Correct information on LEON4 AMBA access size in section 6.7.4.
Correct typos in %ASR22-23 description in section 6.10.3
Correct typo on Memory scrubber Error Threshold registers, BECTE field.
Add package drawing in section 40.4.
1.4
2016 June
Change status from advanced to preliminary data sheet
Correct to PCIMODE_ENABLE=HIGH in table 27, row 1, column 3.
Correct Level-2 cache tag and checkbit register layout in section 9.4
Correct SDCFG2 register reference in section 10.4.6.
Correct reference to description of tick-out connection SpaceWire router register descriptions under section 17.4.8.
Added description in section 4.11 of how to handle Ethernet TXCLK and mode switch to
Gigabit operation.
Clarify in section 34.1 that temperature sensor is disabled on current prototype and engineering model devices.
Update description of GRGPIO IFLAG register in section 27.3.10.
Add note about development board in new section 1.5.
Clarify trace point usage in sections 6.9.1 and 13.4.
Clarify in section 17.5.3 that SpaceWire router RTR.RTCOMB register is only accessible via
RMAP.
Rephrase unified cache description in Level-2 cache section 9.1.
Update Level-2 cache error injection description in sections 9.3.6 and 9.4.5.
Minor updates to supplies in section 39.
GR740-UM-DS, Jun 2016, Version 1.4
10
www.cobham.com/gaisler
GR740
1.8
Acronyms
Table 2. Acronyms
Acronym
Comment
AHB
Advanced High-performance bus, part of [AMBA]
AMBA
Advanced Microcontroller Bus Architecture
AMP
See ASMP
APB
Advanced Peripheral Bus, part of [AMBA]
ASMP
Asymmetric Multi-Processing (in the context of this document: different OS instances running on own processor cores)
BCH
Bose-Hocquenghem-Chaudhuri, class of error-correcting codes
CAN
Controller Area Network, bus standard
CPU
Central Processing Unit, used to refer to one LEON4 processor core.
DCL
Debug Communication Link. Provides a bridge between an external interface and on-chip
AHB bus.
DDR
Double Data Rate
DMA
Direct Memory Access
DSU
Debug Support Unit
EDAC
Error Detection and Correction
EDCL
Ethernet Debug Communication Link
FIFO
First-In-First-Out, refers to buffer type
FPU
Floating Point Unit
Gb
Gigabit, 109 bits
GB
Gigabyte, 109 bytes
GiB
Gibibyte, gigabinary byte, 230 bytes, unit defined in IEEE 1541-200
I/O
Input/Output
IP, IPv4
Internet Protocol (version 4)
ISR
Interrupt Service Routine
JTAG
Joint Test Action Group (developer of IEEE Standard 1149.1-1990)
kB
Kilobyte, 103 bytes
KiB
Kibibyte, 210 bytes, unit defined in IEEE 1541-2002
L2
Level-2, used in L2 cache abbreviation
MAC
Media Access Controller
Mb, Mbit
Megabit, 106 bits
MB, Mbyte
Megabyte, 106 bytes
MiB
Mebibyte, 220 bytes, unit defined in IEEE 1541-2002
OS
Operating System
PCI
Peripheral Component Interconnect
PROM
Programmable Read Only Memory. In this document used to signify boot-PROM.
RAM
Random Access Memory
RMAP
Remote Memory Access Protocol
SEE
Single Event Effects
SEL/SEU/SET
Single Event Latchup/Upset/Transient
GR740-UM-DS, Jun 2016, Version 1.4
11
www.cobham.com/gaisler
GR740
Table 2. Acronyms
1.9
Acronym
Comment
SMP
Symmetric Multi-Processing
SPARC
Scalable Processor ARChitecture
TCP
Transmission Control Protocol
UART
Universal Asynchronous Receiver/Transmitter
UDP
User Datagram Protocol
Definitions
This section and the following subsections define the typographic and naming conventions used
throughout this document.
1.9.1
Bit numbering
The following conventions are used for bit numbering:
•
The most significant bit (MSb) of a data type has the leftmost position
•
The least significant bit of a data type has the rightmost position
•
Unless otherwise indicated, the MSb of a data type has the highest bit number and the LSb the
lowest bit number
1.9.2
Radix
The following conventions is used for writing numbers:
•
Binary numbers are indicated by the prefix "0b", e.g. 0b1010.
•
Hexadecimal numbers are indicated by the prefix "0x", e.g. 0xF00F
•
Unless a radix is explicitly declared, the number should be considered a decimal.
1.9.3
Data types
Byte (BYTE)
8 bits of data
Halfword (HWORD)
16 bits of data
Word (WORD)
32 bits of data
Double word (DWORD)
64 bits of data
Quad word (4WORD)
128-bits of data
GR740-UM-DS, Jun 2016, Version 1.4
12
www.cobham.com/gaisler
GR740
1.10
Register descriptions
An example register, showing the register layout used throughout this document, can be seen
in table 3. The values used for the reset value fields are described in table 4, and the values
used for the field type fields are described in table 5. Fields that are named RESERVED,
RES, or R are read-only fields. These fields can be written with zero or with the value read
from the same register field.
Table 3. <Address> - <Register acronym> - <Register name>
31
24 23
16 15
8
7
0
EF3
EF2
EF1
EF0
<Reset value for EF3>
<Reset value for EF2>
<Reset value for EF1>
<Reset value for EF0>
<Field type for EF3>
<Field type for EF2>
<Field type for EF1>
<Field type for EF0>
Table 4.
31: 24
Example field 3 (EF3) - <Field description>
23: 16
Example field 2 (EF2) - <Field description>
15: 8
Example field 1 (EF1) - <Field description>
7: 0
Example field 0 (EF0) - <Field description>
Reset value definitions
Value
Description
0
Reset value 0.
1
Reset value 1. Used for single-bit fields.
0xNN
Hexadecimal representation of reset value. Used for multi-bit fields.
0bNN
Binary representation of reset value. Used for multi-bit fields.
NR
Field not reset
*
Special reset condition, described in textual description of the field. Used for example when reset
value is taken from a pin.
-
Don’t care / Not applicable
Table 5. Field type definitions
Value
Description
r
Read-only. Writes have no effect.
w
Write-only. Used for a writable field in a register where the field’s read-value has no meaning.
rw
Readable and writable.
rw*
Readable and writable. Special condition for write, described in textual description of field.
wc
Write-clear. Readable, and cleared when written with a 1
cas
Readable, and writable through compare-and-swap. Only applies to SpaceWire Plug-and-Play registers.
GR740-UM-DS, Jun 2016, Version 1.4
13
www.cobham.com/gaisler
GR740
2
Architecture
2.1
Overview
The system is built around five AMBA AHB buses; one 128-bit Processor AHB bus, one 128-bit
Memory AHB bus, two 32-bit I/O AHB buses and one 32-bit Debug AHB bus. The Processor AHB
bus houses four LEON4FT processor cores connected to a shared L2 cache. The Memory AHB bus is
located between the L2 cache and the main external memory interface (SDRAM) and attaches a memory scrubber.
The two separate I/O AHB buses connect peripheral cores. Slave interfaces of the PCI master/target
and PROM/IO memory controller are placed on one bus (Slave I/O AHB bus). All master/DMA interfaces are placed on the other bus (Master I/O AHB bus). The Master I/O AHB bus connects to the
Processor AHB bus via an AHB/AHB bridge that provides access restriction and address translation
(IOMMU) functionality. The IOMMU also has an AHB master interface connected to the Memory
AHB bus. The AHB master interface to use when propagating traffic from a core on the Master I/O
AHB bus is dynamically configurable.
Peripheral unit register interfaces such as timers, interrupt controllers, UARTs, general purpose I/O
port, SPI controller, MIL-STD-1553B interface, Ethernet MACs, CAN controllers, and SpaceWire
router AMBA interfaces are connected via two AHB/APB bridges that are attached to the Processor
AHB bus.
The fifth bus, a dedicated 32-bit Debug AHB bus, connects a debug support unit (DSU), one AHB
trace buffer monitoring the Master I/O AHB bus and several debug communication links. The Debug
AHB bus allows for non-intrusive debugging through the DSU and direct access to the complete system, as the Debug AHB bus is not placed behind an AHB bridge with access restriction functionality.
The chapters in this document have been grouped after the bus topology. The first chapters describe
cores connected to the Processor AHB bus, followed by the Memory AHB bus, Debug AHB bus,
Master I/O AHB bus and finally Slave I/O AHB bus and APB buses.
GR740-UM-DS, Jun 2016, Version 1.4
14
www.cobham.com/gaisler
GR740
The GR740 has the following on-chip functions:
•
•
•
•
•
•
•
•
•
•
4x LEON4 SPARC V8 processor cores with MMU and GRFPU floating-point unit
Level-2 cache, 4-ways, BCH protection, supports locking of 1-4 ways
Debug Support Unit (DSU) with instruction and AHB trace buffers
Ethernet, JTAG and SpaceWire debug communication links
96-bit PC100 SDRAM memory controller with Reed-Solomon EDAC
Hardware memory scrubber
8/16-bit PROM/IO controller with BCH EDAC
I/O Memory Management Unit (IOMMU) with support for eight groups of DMA units
8-port SpaceWire router/switch with four on-chip AMBA ports with RMAP
SpaceWire TDP controller
•
•
•
•
2x 10/100/1000 Mbit Ethernet MAC
32-bit 33/66 (TBC) MHz PCI master/target interface with DMA engine
MIL-STD-1553B interface controller
2x CAN 2.0B controllers
•
•
•
•
•
•
•
•
•
•
2x UART
SPI master/slave controller
Interrupt controller with extended support for asymmetric multiprocessing
1x Timer core with five timers, time latch/set functionality and watchdog functionality
4x Timer core with four timers and time latch/set functionality
Separate AHB and PCI trace buffers
Temperature sensor
Clock gating unit
LEON4 statistics unit (performance counters)
Pad and PLL control unit
•
AHB status registers
GR740-UM-DS, Jun 2016, Version 1.4
15
www.cobham.com/gaisler
GR740
2.2
Cores
The design is based on the following IP cores from the GRLIB IP Library:
Table 6. Used IP cores
Core
Function
Documented in
section
Vendor
Device
AHB2AHB
Uni-directional AHB/AHB bridge
12, 23
0x01
0x020
AHBJTAG
JTAG/AHB Debug interface
14
0x01
0x01C
AHBSTAT
AHB Status Register
32
0x01
0x052
AHBTRACE
AHB trace buffer
16
0x01
0x017
APBCTRL
AHB/APB bridge
38
0x01
0x006
IRQ(A)MP
Multiprocessor interrupt controller
26
0x01
0x00D
APBUART
8-bit UART with FIFO
28
0x01
0x00C
DSU4
LEON4 Debug Support Unit
13
0x01
0x049
MMCTRL
Memory controller
10
0x01
0x05D
GPTIMER
Modular timer unit with watchdog
25
0x01
0x011
GR1553B
MIL-STD-1553B / AS15531 interface
21
0x01
0x04D
GRCAN
CAN 2.0 controller with DMA
22
0x01
0x03D
GRCLKGATE
Clock gating unit
30
0x01
0x02C
GRETH_GBIT
10/100/1000 Ethernet MAC with DCL
19
0x01
0x01D
GRGPIO
General Purpose I/O Port
27
0x01
0x01A
GRGPRBANK
General Purpose Register Bank
35
0x01
0x08F
GRGPREG
General Purpose Register
33
0x01
0x087
GRIOMMU
AHB/AHB bridge with protection (IOMMU)
18
0x01
0x04F
GRPCI2
Fast 32-bit PCI bridge
20
0x01
0x07C
GRSPW2
SpaceWire codec with RMAP
15
0x01
0x029
GRSPWROUTER
SpaceWire router switch
17
0x01
0x08B
GRSPWTDP
SpaceWire - Time Distribution Protocol
36
0x01
0x097
FTMCTRL
8/16/32-bit memory controller with EDAC
24
0x01
0x054
L2CACHE
Level 2 cache
9
0x01
0x04B
L4STAT
LEON4 statistical unit
31
0x01
0x047
LEON4
LEON4 SPARC V8 32-bit processor
6
0x01
0x048
MEMSCRUB
Memory scrubber
11
0x01
0x057
SPICTRL
SPI controller
29
0x01
0x02D
GR740THSENS
GR740 Temperature sensor controller
34
0x01
0x099
The information in the last two columns is available via plug’n’play information in the system and is
used by software to detect units and to initialize software drivers.
GR740-UM-DS, Jun 2016, Version 1.4
16
www.cobham.com/gaisler
GR740
2.3
Memory map
The memory map of the internal AHB and APB buses as seen from the processor cores can be seen
below. Software does not need to be aware that a bridge is positioned between the processor and a
peripheral core since the address mapping between buses is one-to-one.
Table 7. AMBA memory map, as seen from processors
Core
Address range
Area
Bus
L2CACHE
0x00000000 - 0x7FFFFFFF
L2 cache memory area. Covers
SDRAM memory area.
Processor
GRPCI2
0x80000000 - 0xBFFFFFFF
PCI memory area
Slave I/O
FTMCTRL
0xC0000000 - 0xCFFFFFFF
PROM area
Slave I/O
0xD0000000 - 0xDFFFFFFF
Memory mapped I/O area
0xE0000000 - 0xEFFFFFFF
Unused. This memory range is occupied on the Debug AHB bus and is
not visible from the processors. A
separate table below shows the mapping.
Processor
0xF0000000 - 0xF03FFFFF
L2 cache configuration registers
Processor
0xF0400000 - 0xFF7FFFFF
Unused
Processor
GRPCI2
0xFF800000 - 0xFF83FFFF
PCI I/O area
Slave I/O
GRIOMMU
0xFF840000 - 0xFF847FFF
IOMMU configuration registers
Slave I/O
0xFF848000 - 0xFF87FFFF
Unused
Slave I/O
0xFF880000 - 0xFF880FFF
SpaceWire router configuration port
Slave I/O
0xFF881000 - 0xFF8FEFFF
Unused
Slave I/O
0xFF8FF000 - 0xFF8FFFFF
Slave I/O bus plug&play area
Slave I/O
0xFF900000 - 0xFF9FFFFF
APB bridge 0
Processor
APBUART0
0xFF900000 - 0xFF9000FF
UART 0 registers
Processor
APBUART1
0xFF901000 - 0xFF9010FF
UART 1 registers
Processor
GRGPIO0
0xFF902000 - 0xFF9020FF
General purpose I/O port registers
Processor
FTMCTRL
0xFF903000 - 0xFF9030FF
PROM/IO controller registers
Processor
A
IRQ(A)MP
0xFF904000 - 0xFF907FFF
Interrupt controller registers
Processor
L2CACHE
GRSPWROUTER
APBBRIDGE0
P
GPTIMER0
0xFF908000 - 0xFF9080FF
Timer unit 0 registers
Processor
B
GPTIMER1
0xFF909000 - 0xFF9090FF
Timer unit 1 registers
Processor
B
GPTIMER2
0xFF90A000 - 0xFF90A0FF
Timer unit 2 registers
Processor
R
GPTIMER3
0xFF90B000 - 0xFF90B0FF
Timer unit 3 registers
Processor
I
GPTIMER4
0xFF90C000 - 0xFF90C0FF
Timer unit 4 registers
Processor
D
GRSPWROUTER
0xFF90D000 - 0xFF90DFFF
SpaceWire router AMBA interface 0
Processor
G
GRSPWROUTER
0xFF90E000 - 0xFF90EFFF
SpaceWire router AMBA interface 1
Processor
E
GRSPWROUTER
0xFF90F000 - 0xFF90FFFF
SpaceWire router AMBA interface 2
Processor
0
GRSPWROUTER
0xFF910000 - 0xFF910FFF
SpaceWire router AMBA interface 3
Processor
GRETH_GBIT0
0xFF940000 - 0xFF9400FF
Gigabit Ethernet MAC 0 registers
Processor
GRETH_GBIT1
0xFF980000 - 0xFF9800FF
Gigabit Ethernet MAC 1 registers
Processor
APBBRIDGE0
0xFF990000 - 0xFF9FFEFF
Unused
Processor
APBBRIDGE0
APBBRIDGE1
0xFF9FF000 - 0xFF9FFFFF
APB bus 0 plug&play area
Processor
0xFFA00000 - 0xFFAFFFFF
APB bridge 1
Processor
GR740-UM-DS, Jun 2016, Version 1.4
17
www.cobham.com/gaisler
GR740
Table 7. AMBA memory map, as seen from processors
Core
Address range
Area
Bus
A
GRPCI2
0xFFA00000 - 0xFFA000FF
PCI core registers
Processor
P
GRCAN0
0xFFA01000 - 0xFFA013FF
CAN 2.0 controller 0
Processor
B
GRCAN1
0xFFA02000 - 0xFFA023FF
CAN 2.0 controller 1
Processor
B
SPICTRL
0xFFA03000 - 0xFFA030FF
SPI controller
Processor
R
GRCLKGATE
0xFFA04000 - 0xFFA040FF
Clock gating unit
Processor
I
GR1553B
0xFFA05000 - 0xFFA050FF
MIL-STD-1553B controller
Processor
D
AHBSTAT0
0xFFA06000 - 0xFFA060FF
AHB status register monitoring Processor AHB bus
Processor
AHBSTAT1
0xFFA07000 - 0xFFA070FF
AHB status register monitoring
Slave I/O AHB bus
Processor
GRGPIO1
0xFFA08000 - 0xFFA080FF
General purpose I/O register for mul- Processor
tiplexed pins.
GRGPREG
0xFFA09000 - 0xFFA090FF
Register for bootstrap signals
Processor
GR740THSENS
0xFFA0A000 - 0xFFA0A0FF
Temperature sensor
Processor
GRGPRBANK
0xFFA0B000 - 0xFFA0B0FF
General purpose register bank
Processor
GRSPWTDP
0xFFA0C000 - 0xFFA0C1FF
CCSDS TDP controller
Processor
L4STAT
0xFFA0D000 - 0xFFA0D1FF
LEON4 Statistics Unit
Processor
APBBRIDGE1
0xFFA0D200 - 0xFFAFFEFF
Unused
Processor
APBBRIDGE1
0xFFAFF000 - 0xFFAFFFFF
APB bus 1 plug&play area
Processor
0xFFB00000 - 0xFFDFFFFF
Unused
Processor
0xFFE00000 - 0xFFE000FF
SDRAM controller registers
Memory
0xFFE00100 - 0xFFE00FFF
Unused
Memory
0xFFE01000 - 0xFFE010FF
Memory scrubber registers
Memory
0xFFE01100 - 0xFFEFEFFF
Unused
Memory
0xFFEFF000 - 0xFFEFFFFF
Memory bus plug&play area
Memory
0xFFF00000 - 0xFFFFEFFF
Unused
Processor
0xFFFFF000 - 0xFFFFFFFF
Processor bus plug&play area
Processor
G
E
1
MMCTRL
MEMSCRUB
When connecting to the system via one of the debug communication links (JTAG, Ethernet, USB, or
SpaceWire) connected to the Debug AHB bus, several debug support cores will be visible. Table 8
below lists the address map of these cores. Note that cores in the address range 0xE0000000 0xEFFFFFFF are not accessible from the processors or from any cores on the Master I/O AHB bus.
Accesses to this range from any core not located on the Debug AHB bus will result in an AMBA
ERROR response. Apart from the area 0xE0000000 - 0xEFFFFFFF, the AMBA memory space seen
via the debug communication links is identical to the address space seen from other cores in the system.
Accesses to unused AMBA AHB address space will result in an AMBA ERROR response, this
applies to the memory areas that are marked as "Unused" in the table above. Accesses to unused areas
located on one of the AHB/APB bridges will not have any effect, note that these unoccupied address
ranges are not marked as "Unused" in the table above. No AMBA ERROR response will be given for
memory allocated to one of the APB bridges.
GR740-UM-DS, Jun 2016, Version 1.4
18
www.cobham.com/gaisler
GR740
Table 8. AMBA address range 0xE0000000 - 0xEFFFFFFF on Debug AHB bus
Core
Address range
Comment
DSU4
0xE0000000 - 0xE07FFFFF
Debug Support Unit area for processor 0
0xE1000000 - 0xE17FFFFF
Debug Support Unit area for processor 1
0xE2000000 - 0xE27FFFFF
Debug Support Unit area for processor 2
0xE3000000 - 0xE37FFFFF
Debug Support Unit area for processor 3
APBBRIDGED
0xE4000400 - 0xE40FFFFF
APB bridge on Debug AHB bus
A GRSPW2
0xE4000000 - 0xE40000FF
SpaceWire RMAP target with AMBA interface
P
0xE4000200 - 0xE40003FF
LEON4 Statistics unit, secondary port
B APBBRIDGED
0xE4000200 - 0xE403FFFF
Unused
D GRPCI2
L4STAT
0xE4040000 - 0xE407FFFF
GRPCI2 secondary PCI trace buffer interface
APBBRIDGED
0xE4080000 - 0xE40FFEFF
Unused
APBBRIDGED
0xE40FFF00 - 0xE40FFFFF
Debug APB bus plug&play area
0xE4100000 - 0xEEFFFFFF
Unused
0xEFF00000 - 0xEFF1FFFF
AHB trace buffer, tracing master I/O AHB bus
0xEFF20000 - 0xEFFFEFFF
Unused
0xEFFFF000 - 0xEFFFFFFF
Debug AHB bus plug&play area
AHBTRACE
GR740-UM-DS, Jun 2016, Version 1.4
19
www.cobham.com/gaisler
GR740
2.4
Interrupts
The table below indicates the interrupt assignments. Note that the table below describes interrupt bus
lines, these can be remapped in the interrupt controller.
Table 9. Interrupt assignments
Interrupt
Core
Comment
1
GPTIMER0
GPTIMER unit 0, timer 1
2
GPTIMER0
GPTIMER unit 0, timer 2
3
GPTIMER0
GPTIMER unit 0, timer 3
4
GPTIMER0
GPTIMER unit 0, timer 4
5
GPTIMER0
GPTIMER unit 0, timer 5
6
GPTIMER1
Shared interrupt for all timers on GPTIMER unit 1
7
GPTIMER2
Shared interrupt for all timers on GPTIMER unit 2
8
GPTIMER3
Shared interrupt for all timers on GPTIMER unit 3
9
GPTIMER4
Shared interrupt for all timers on GPTIMER unit 4
10
IRQ(A)MP
Extended interrupt line.
11
GRPCI/PCIDMA
PCI master/target and PCI DMA
12
Unassigned
13
Unassigned
Suitable for use by software for inter-processor and
inter-process synchronization.
14
Unassigned
15
Unassigned
Note: Not maskable by processor
The GPIO port has configuration registers that determine the mapping between general purpose I/O lines
and the four interrupt lines allocated to the GPIO port.
16
GRGPIO0 /1 / CAN
17
GRGPIO0 /1 / CAN
18
GRGPIO0 /1 / CAN
19
GRGPIO0/1/ SPICTRL
Interrupt lines 16 -18 are shared between the GPIO port
and CAN controllers.
Interrupt line 19 is shared between the GPIO port and
the SPI controller.
20
SPWROUTER AMBA I/F 0
SpaceWire router AMBA interface 0
21
SPWROUTER AMBA I/F 1
SpaceWire router AMBA interface 1
22
SPWROUTER AMBA I/F 2
SpaceWire router AMBA interface 2
23
SPWROUTER AMBA I/F 3
SpaceWire router AMBA interface 3
24
GRETH_GBIT0
Gigabit Ethernet MAC 0
25
GRETH_GBIT1
Gigabit Ethernet MAC 1
26
GR1553B
MIL-STD-1553B interface controller
27
AHBSTAT/ST65THSENS
Shared by all AHB Status registers in design and by
temperature sensor.
28
MEMSCRUB/L2CACHE
Memory scrubber and L2 cache
29
APBUART0
UART 0
30
APBUART1
UART 1
31
GRIOMMU / GRSPWTDP
IOMMU register interface interrupt.
CCSDS TDP controller interrupt
2.5
Plug & play and bus index information
The format of GRLIB AMBA Plug&play information is given in sections 37 and 38. The address
ranges of the plug&play configuration areas are given in the preceding section and is also replicated
for each core in the tables below. The plug&play areas are used by software to detect the system-on-
GR740-UM-DS, Jun 2016, Version 1.4
20
www.cobham.com/gaisler
GR740
chip architecture. The values in the tables below are fixed. The tables also include the bus indexes for
all masters and slaves on the system’s AHB and APB buses.
The plug & play memory map and bus indexes for AMBA AHB masters on the Processor AHB bus
are shown in table 10.
Table 10. Plug & play information for masters on Processor AHB bus
Core
Index
Function
Address range
LEON4
0
LEON4 SPARC V8 Processor
0xFFFFF000 - 0xFFFFF01F
LEON4
1
LEON4 SPARC V8 Processor
0xFFFFF020 - 0xFFFFF03F
LEON4
2
LEON4 SPARC V8 Processor
0xFFFFF040 - 0xFFFFF05F
LEON4
3
LEON4 SPARC V8 Processor
0xFFFFF060 - 0xFFFFF07F
GRIOMMU
4
AHB/AHB bridge with protection functionality
0xFFFFF080 - 0xFFFFF09F
AHB2AHB
5
Uni-directional AHB/AHB bridge connecting Debug
AHB bus to Processor AHB bus
0xFFFFF0B0 - 0xFFFFF0BF
The plug & play memory map and bus indexes for AMBA AHB slaves on the Processor AHB bus are
shown in table 11.
Table 11. Plug & play information for slaves on Processor AHB bus
Core
Index
Function
Address range
L2CACHE
0
Level 2 cache
0xFFFFF800 - 0xFFFFF81F
AHB2AHB
1
Uni-directional AHB/AHB bridge connecting Processor AHB bus to Slave I/O bus
0xFFFFF820 - 0xFFFFF83F
APBCTRL
2
AHB/APB bridge 0
0xFFFFF840 - 0xFFFFF85F
APBCTRL
3
AHB/APB bridge 1
0xFFFFF860 - 0xFFFFF87F
The plug & play memory map and bus indexes for AMBA AHB masters on the Memory AHB bus are
shown in table 12.
Table 12. Plug & play information for masters on Memory AHB bus
Core
Index
Function
Address range
L2CACHE
0
Level 2 cache
0xFFEFF000 - 0xFFEFF01F
MEMSCRUB
1
Memory scrubber
0xFFEFF020 - 0xFFEFF03F
GRIOMMU
2
IOMMU secondary AHB master interface
0xFFEFF040 - 0xFFEFF05F
The plug & play memory map and bus indexes for AMBA AHB slaves on the Processor AHB bus are
shown in table 13.
Table 13. Plug & play information for slaves on Memory AHB bus
Core
Index
Function
Address range
MMCTRL
0
SDRAM controller
0xFFEFF800 - 0xFFEFF81F
MEMSCRUB
1
Memory scrubber
0xFFEFF820 - 0xFFEFF83F
GR740-UM-DS, Jun 2016, Version 1.4
21
www.cobham.com/gaisler
GR740
The plug & play memory map and bus indexes for AMBA AHB masters on the Debug AHB bus are
shown in table 14.
Table 14. Plug & play information for masters on Debug AHB bus
Core
Index
Function
Address range
AHBJTAG
0
JTAG Debug Communication Link
0xEFFFF000 - 0xEFFFF01F
GRSPW2
1
SpaceWire codes with AMBA interface and RMAP
target
0xEFFFF020 - 0xEFFFF03F
GRETH_GBIT
EDCL 0
2
10/100/1000 Mbit Ethernet Debug Communication
Link
0xEFFFF040 - 0xEFFFF05F
GRETH_GBIT
EDCL 1
3
10/100/1000 Mbit Ethernet Debug Communication
Link
0xEFFFF060 - 0xEFFFF07F
The plug & play memory map and bus indexes for AMBA AHB slaves on the Processor AHB bus are
shown in table 15.
Table 15. Plug & play information for slaves on Debug AHB bus
Core
Index
Function
Address range
DSU4
0
LEON4 Debug Support Unit
0xEFFFF800 - 0xEFFFF81F
AHB2AHB
1
Uni-directional AHB/AHB bridge connecting Debug
AHB bus to Processor AHB bus
0xEFFFF820 - 0xEFFFF83F
APBCTRL
2
AHB/APB bridge
0xEFFFF840 - 0xEFFFF85F
AHBTRACE
3
AHB trace buffer
0xEFFFF860 - 0xEFFFF87F
The plug & play memory map and bus indexes for AMBA AHB masters on the Slave I/O AHB bus
are shown in table 16.
Table 16. Plug & play information for masters on Slave I/O AHB bus
Core
Index
Function
Address range
AHB2AHB
0
Uni-directional AHB/AHB bridge connecting Processor AHB bus to Slave I/O bus
0xFF8FF000 - 0xFF8FF01F
The plug & play memory map and bus indexes for AMBA AHB slaves on the Slave I/O AHB bus are
shown in table 17.
Table 17. Plug & play information for slaves on Slave I/O AHB bus
Core
Index
Function
Address range
FTMCTRL
0
PROM/IO controller
0xFF8FF800 - 0xFF8FF81F
GRPCI2
1
PCI master interface
0xFF8FF820 - 0xFF8FF83F
GRIOMMU
2
IOMMU register interface
0xFF8FF840 - 0xFF8FF85F
GRSPWROUTER
3
SpaceWire router AMBA configuration interface
0xFF8FF860 - 0xFF8FF87F
GR740-UM-DS, Jun 2016, Version 1.4
22
www.cobham.com/gaisler
GR740
The bus indexes for AMBA AHB masters on the Master I/O AHB bus are shown in table 18. The
Master I/O AHB bus does not have an AMBA plug&play area.
Table 18. Bus index information for masters on Master I/O AHB bus
Core
Index
Function
Address range
GRPCI2
0
PCI target
Not applicable
GRPCI2
1
PCI DMA
Not applicable
GRETH_GBIT 0
2
10/100/1000 Ethernet MAC 0
Not applicable
GRETH_GBIT 1
3
10/100/1000 Ethernet MAC 1
Not applicable
SPWROUTER
4
SpaceWire router AMBA interface 0
Not applicable
SPWROUTER
5
SpaceWire router AMBA interface 1
Not applicable
SPWROUTER
6
SpaceWire router AMBA interface 2
Not applicable
SPWROUTER
7
SpaceWire router AMBA interface 3
Not applicable
GR1553B
8
MIL-STD-1553B interface
Not applicable
GRCAN
9
CAN 2.0 controller
Not applicable
The bus index for the AMBA AHB slave on the Master I/O AHB bus is shown in table 19.
Table 19. Bus index information for slaves on Master I/O AHB bus
Core
Index
Function
Address range
GRIOMMU
0
IOMMU slave interface
Not applicable
The plug & play memory map and bus indexes for AMBA APB slaves connected via the AHB/APB
bridges on the Slave I/O AHB bus are shown in tables 20 and 21.
Table 20. Plug & play information for APB slaves connected via the first APB bridge on Slave I/O AHB bus
Core
Index
Function
Address range
APBUART
0
UART 0
0xFF9FF000 - 0xFF9FF007
APBUART
1
UART 1
0xFF9FF008 - 0xFF9FF00F
GRGPIO
2
General Purpose I/O Port
0xFF9FF010 - 0xFF9FF017
FTMCTRL
3
PROM/IO memory controller
0xFF9FF018 - 0xFF9FF01F
IRQ(A)MP
4
Multiprocessor interrupt controller with AMP extension
0xFF9FF020 - 0xFF9FF027
GPTIMER
5
General Purpose Timer Unit 0
0xFF9FF028 - 0xFF9FF02F
GPTIMER
6
General Purpose Timer Unit 1
0xFF9FF030 - 0xFF9FF037
GPTIMER
7
General Purpose Timer Unit 2
0xFF9FF038 - 0xFF9FF03F
GPTIMER
8
General Purpose Timer Unit 3
0xFF9FF040 - 0xFF9FF047
GPTIMER
9
General Purpose Timer Unit 4
0xFF9FF048 - 0xFF9FF04F
GRSPWROUTER
10
SpaceWire router AMBA interface 0
0xFF9FF050 - 0xFF9FF057
GRSPWROUTER
11
SpaceWire router AMBA interface 1
0xFF9FF058 - 0xFF9FF05F
GRSPWROUTER
12
SpaceWire router AMBA interface 2
0xFF9FF060 - 0xFF9FF067
GRSPWROUTER
13
SpaceWire router AMBA interface 3
0xFF9FF068 - 0xFF9FF06F
GRETH_GBIT
14
10/100/1000 Mbit Ethernet MAC
0xFF9FF070 - 0xFF9FF077
GRETH_GBIT
15
10/100/1000 Mbit Ethernet MAC
0xFF9FF078 - 0xFF9FF07F
GR740-UM-DS, Jun 2016, Version 1.4
23
www.cobham.com/gaisler
GR740
Table 21. Plug & play information for APB slaves connected via the second APB bridge on Slave I/O AHB bus
Core
Index
Function
Address range
GRPCI2
0
PCI configuration register interface
0xFFAFF000 - 0xFFAFF007
GRCAN
1
CAN 2.0 controller
0xFFAFF008 - 0xFFAFF00F
GRCAN
2
CAN 2.0 controller
0xFFAFF010 - 0xFFAFF017
SPICTRL
3
SPI controller
0xFFAFF018 - 0xFFAFF01F
GRCLKGATE
4
Clock gating unit register interface
0xFFAFF020 - 0xFFAFF027
GR1553B
5
MIL-STD-1553B interface
0xFFAFF028 - 0xFFAFF02F
AHBSTAT
6
AHB Status register interface
0xFFAFF030 - 0xFFAFF037
AHBSTAT
7
AHB Status register interface
0xFFAFF038 - 0xFFAFF03F
GRGPIO
8
General purpose I/O port
0xFFAFF040 - 0xFFAFF047
GRGPREG
9
General purpose register for bootstrap control
0xFFAFF048 - 0xFFAFF04F
ST65THSENS
10
Temperature sensor
0xFFAFF050 - 0xFFAFF057
GRGPRBANK
11
General purpose register bank
0xFFAFF058 - 0xFFAFF05F
GRSPWTDP
12
SpaceWire - Time Distribution Protocol
0xFFAFF060 - 0xFFAFF067
L4STAT
13
LEON4 Statistics Unit register interface
0xFFAFF068 - 0xFFAFF06F
The plug & play memory map and bus indexes for AMBA APB slaves connected via the AHB/APB
bridge on the Debug AHB bus are shown in table 22.
Table 22. Plug & play information for APB slaves connected via APB bridge on Debug AHB bus
Core
Index
Function
Address range
GRSPW2
0
SpaceWire codec AMBA interface with RMAP target
0xE40FF000 - 0xE40FF007
L4STAT
1
LEON4 Statistics Unit
0xE40FF008 - 0xE40FF00F
GRPCI2
2
GRPCI2 trace buffer secondary interface
0xE40FF010 - 0xE40FF017
GR740-UM-DS, Jun 2016, Version 1.4
24
www.cobham.com/gaisler
GR740
3
Signals
3.1
Bootstrap signals
The power-up and initialisation state is affected by several external signals as shown in table 23. The
bootstrap signals taken via GPIO are saved when the on-chip system reset is released. This occurs
after deassertion of the SYS_RESETN input and lock of all active PLLs (see also reset description in
section 4). This means that if a core, such as the Ethernet controller, is clock gated off and then reset
and enabled at a later time, the bootstrap signal value will be taken from the saved value present in a
general purpose register described in section 33.
Table 23. Bootstrap signals
Bootstrap signal
Description
DSU_EN
Enables the Debug Support Unit (DSU) and other cores on the Debug AHB bus. If DSU_EN is
HIGH the DSU and the cores on the Debug AHB bus will be clocked. If DSU_EN is LOW the
DSU and all cores on the Debug AHB bus will be clock gated off.
The value of the DSU_EN signal also controls if the Ethernet Debug Communication Links
(EDCLs) should be enabled. If DSU_EN is LOW the EDCLs will be disabled and clock gated
after reset, otherwise they will be enabled.
BREAK
Puts all processors in debug mode when asserted while DSU_EN is HIGH. When DSU_EN is
LOW, BREAK is assigned to the timer enable bit of the watchdog timer and also controls if the
first processor starts executing after reset.
PCIMODE_ENABLE
Enables PCI mode. If the bootstrap signal MEM_IFWIDTH is HIGH then PCIMODE_ENABLE selects if the top-half of the SDRAM interface should be used for the PCI controller
(HIGH) or Ethernet port 1 (LOW).
MEM_IFWIDTH
Selects the width of SDRAM interface. If this signal is LOW then the external memory interface
uses 64 data bits with up to 32 check bits. If this signal is HIGH then the external memory interface uses 32 data bits with up to 16 check bits and the top half of the SDRAM interface is used
for PCI or Ethernet port 1, as determined by the PCIMODE_ENABLE bootstrap signal.
MEM_CLKSEL
The value of this signal determines the clock source for the SDRAM memory. If this signal is
low then the memory clock and the system clock has the same source, otherwise the source for
the memory clock is the MEM_EXTCLOCK clock input.
GPIO[5:0]
Sets the least significant address nibble of the IP and MAC address for Ethernet Debug Communication Link (EDCL) 0 and 1 . Setting the full address nibble high disables the corresponding
EDCL at reset (the EDCL can be enabled via software at a later stage).
GPIO[1:0] sets the least significant bits of the nibble for EDCL 0 and EDCL1
GPIO[3:2] sets the top nibble bits for EDCL 0 and GPIO[5:4] set the top nibble bits for EDCL1.
GPIO[7:6]
Selects SpaceWire router Distributed Interrupt configuration
"00" - Interrupts with acknowledgement mode (32 interrupts with acknowledgements);
"01" - Extended interrupt mode (64 interrupts, no acknowledgements);
"10" - Distributed interrupts disabled, all Dist. Interrupt codes treated as Time-Codes;
"11" - Dist. interrupt disabled, Control code treated as Time-Code if CTRL flags are zero.
GPIO[9:8]
Selects if Ethernet Debug Communication Link 0 (GPIO[8]) and Link 1(GPIO[9]) traffic should
be routed over the Debug AHB bus (HIGH) or the Master I/O AHB bus (LOW).
GPIO[10]
Selects the PROM width. 0: 8-bit PROM, 1: 16-bit PROM
GPIO[11]
Controls the clock gate settings for the SpaceWire router.
GPIO[13:12]
Sets the two least significant bits of the SpaceWire router’s instance ID.
GPIO[14]
Controls reset value of PROM/IO controller’s PROM EDAC enable (PE) bit. When this input is
’1’ at reset, EDAC checking of the PROM area will be enabled.
GPIO[15]
Selects if the PROM/IO interface should be enabled after reset. If this signal is HIGH then the
PROM/IO interface is enabled. Otherwise the PROM/IO interface pins are routed to their alternative functions.
GR740-UM-DS, Jun 2016, Version 1.4
25
www.cobham.com/gaisler
GR740
Table 23. Bootstrap signals
3.2
Bootstrap signal
Description
PLL_BYPASS[2:0]
Bypass PLL and use clock input directly. 2: SpW clock, 1: SDRAM clock, 0: System clock PLL
bypass.
PLL_IGNLOCK
The PLL outputs of the device are gated until the PLL lock outputs have been asserted. Setting
this signal HIGH disables this clock gating for all PLLs, and also removes the lock signals from
the reset generation.
Pin multiplexing
The device shares pin between the following groups of interfaces:
•
Part of the PROM/IO interface shares pins with UART 0, UART 1, CAN 0, CAN 1, SpaceWire
debug and MIL-STD-1553B. The pins can also be controlled as general-purpose I/O.
•
The top half of the SDRAM interface shares pins with PCI and Ethernet port 1.
The sections below describes multiplexing for the affected interfaces. Section 35 describes the peripheral through which software controls the multiplexing.
3.2.1
PROM/IO interface multiplexing
The selection between the PROM/IO interface and the other low-speed interfaces on the same pins is
done at boot time via the bootstrap signal GPIO[15]. When GPIO[15] is HIGH during reset, then the
full PROM/IO interface will be available. When GPIO[15] is LOW after reset, the alternative function is routed to the shared pins.
The multiplexing has been designed so that even if starting with all the multiplexed pins set to their
alternative (peripheral) mode, enough dedicated PROM/IO pins are still available to access an 8-bit,
64 KiB boot PROM for bootstrapping the system.
After reset, the setting can be reconfigured on a pin by pin basis by software using a register interface
(see the General Purpose Register Bank section). The register interface can also reconfigure the multiplexed I/O:s to function as general-purpose I/Os.
If only a subset of the alternative functions are desired and a larger PROM or IO interface is desired,
then then GPIO[15] should be kept HIGH during reset and software can then during boot assign a
subset of the signals to alternative functions. In this case, the effect of address lines tied to peripherals
on the board toggling during the first PROM accesses before they have been re-configured to their
correct function will need to be considered at the system design level.
A few inputs belonging to the SpaceWire debug and UART CTS signals are shared with GPIO bus
pins without any explicit multiplexing, these inputs are simply connected to both functions at the
same time. Note that the UART CTS signals are ignored by default and will therefore not affect
UART operation unless flow control is enabled in the UART’s control register.
GR740-UM-DS, Jun 2016, Version 1.4
26
www.cobham.com/gaisler
GR740
Table 24. Multiplexed PROM/IO interface pins with alternative functions and control register bit position
Pin name*
Primary function
Alternative function
GPIO2 function
Register
bank
FTMEN /
ALTEN bit
position**
Functi
on
Dir Function
Dir
UART0
TCD
O
GPIO2[21]
IO
21
ADDR[26] O
UART1
TXD
O
GPIO2[20]
IO
20
ADDR[25] O
1553B
TXAP
O
GPIO2[19]
IO
19
Core
Function
Dir Core
PROMIO_ADDR[27]
FTMCTRL
ADDR[27] O
PROMIO_ADDR[26]
FTMCTRL
PROMIO_ADDR[25]
FTMCTRL
PROMIO_ADDR[24]
FTMCTRL
ADDR[24] O
1553B
TXAN
O
GPIO2[18]
IO
18
PROMIO_ADDR[23]
FTMCTRL
ADDR[23] O
1553B
RXEN
A
O
GPIO2[17]
IO
17
PROMIO_ADDR[22]
FTMCTRL
ADDR[22] O
1553B
TXBP
O
GPIO2[16]
IO
16
PROMIO_ADDR[21]
FTMCTRL
ADDR[21] O
1553B
TXBN
O
GPIO2[15]
IO
15
PROMIO_ADDR[20]
FTMCTRL
ADDR[20] O
1553B
RXEN
B
O
GPIO2[14]
IO
14
PROMIO_ADDR[19]
FTMCTRL
ADDR[19] O
SPWD
TXD
O
GPIO2[13]
IO
13
PROMIO_ADDR[18]
FTMCTRL
ADDR[18] O
SPWD
TXS
O
GPIO2[12]
IO
12
PROMIO_ADDR[17]
FTMCTRL
ADDR[17] O
UART0
RTS
O
GPIO2[11]
IO
11
PROMIO_ADDR[16]
FTMCTRL
ADDR[16] O
UART1
RTS
O
GPIO2[10]
IO
10
PROMIO_DATA[7]
FTMCTRL
DATA[7]
IO
UART0
RXD
I
GPIO2[9]
IO
9
PROMIO_DATA[6]
FTMCTRL
DATA[6]
IO
UART1
RXD
I
GPIO2[8]
IO
8
PROMIO_DATA[5]
FTMCTRL
DATA[5]
IO
CAN
RX0
I
GPIO2[7]
IO
7
PROMIO_DATA[4]
FTMCTRL
DATA[4]
IO
CAN
RX1
I
GPIO2[6]
IO
6
PROMIO_DATA[3]
FTMCTRL
DATA[3]
IO
1553B
RXAP
I
GPIO2[5]
IO
5
PROMIO_DATA[2]
FTMCTRL
DATA[2]
IO
1553B
RXAN
I
GPIO2[4]
IO
4
PROMIO_DATA[1]
FTMCTRL
DATA[1]
IO
1553B
RXBP
I
GPIO2[3]
IO
3
PROMIO_DATA[0]
FTMCTRL
DATA[0]
IO
1553B
RXBN
I
GPIO2[2]
IO
2
PROMIO_CEN[1]
FTMCTRL
CEN[1]
O
CAN
TX0
O
GPIO2[1]
IO
1
IO_SN
FTMCTRL
IO_SN
O
CAN
TX1
O
GPIO2[0]
IO
0
* See section 40.3 for pin assignments
** See section 35
Table 25. Shared GPIO interface pins with slow interfaces
Pin name* Primary function
Second function
Core
Function
Dir
Core
Function
Dir
GPIO[7]
GPIO1
GPIO[27]
IO
SPWD
RXD
I
GPIO[6]
GPIO1
GPIO[26]
IO
SPWD
RXS
I
GPIO[5]
GPIO1
GPIO[25]
IO
UART0
CTSN
I
GPIO[4]
GPIO1
GPIO[24]
IO
UART1
CTSN
I
* See section 40.3 for pin assignments
GR740-UM-DS, Jun 2016, Version 1.4
27
www.cobham.com/gaisler
GR740
3.2.2
SDRAM interface multiplexing
The top half of the SDRAM interface shares pins with PCI and Ethernet port 1. The selection between
full SDRAM, PCI and Ethernet is made with the bootstrap signals MEM_IFWIDTH and PCIMODE_ENABLE. This configuration is static and should be kept constant during the runtime
of the device (a change will require a full reset of the device). Some of the data mask (DQM)
bits are used as clock inputs in the alternative modes, and their direction will therefore depend on configuration.
Table 26. Selection between SDRAM, PCI and Ethernet 1
MEM_IFWIDTH
PCIMODE_ENABLE
SDRAM interface
Ethernet port 1
PCI
0
0
64 data bits, 32 check bits
Unavailable
Unavailable
32 data bits, 16 check bits
Available
Unavailable
Unavailable
Available
1
1
0
1
Table 27. Multiplexed SDRAM interface pins with PCI or Ethernet interfaces
ETHERNET1 function
SDRAM function
(MEM_IFWIDTH=HIGH,
(MEM_IFWIDTH=LOW) PCIMODE_ENABLE=HIGH)
ETH1_
PCI function
(MEM_IFWIDTH=HIGH,
PCIMODE_ENABLE=LOW)
PCI_
Function
Dir
Function
Dir
Function
Dir
MEM_DQ[95]
DQ[95]
IO
TXD[7]
O
AD[31]
IO
MEM_DQ[94]
DQ[94]
IO
TXD[6]
O
AD[30]
IO
MEM_DQ[93]
DQ[93]
IO
TXD[5]
O
AD[29]
IO
MEM_DQ[92]
DQ[92]
IO
TXD[4]
O
AD[28]
IO
MEM_DQ[91]
DQ[91]
IO
TXD[3]
O
AD[27]
IO
MEM_DQ[90]
DQ[90]
IO
TXD[2]
O
AD[26]
IO
MEM_DQ[89]
DQ[89]
IO
TXD[1]
O
AD[25]
IO
MEM_DQ[88]
DQ[88]
IO
TXD[0]
O
AD[24]
IO
MEM_DQ[87]
DQ[87]
IO
TXEN
O
AD[23]
IO
MEM_DQ[86]
DQ[86]
IO
TXER
O
AD[22]
IO
MEM_DQ[85]
DQ[85]
IO
(none)
I
AD[21]
IO
MEM_DQ[84]
DQ[84]
IO
(none)
I
AD[20]
IO
MEM_DQ[83]
DQ[83]
IO
(none)
I
AD[19]
IO
MEM_DQ[82]
DQ[82]
IO
(none)
I
AD[18]
IO
MEM_DQ[81]
DQ[81]
IO
(none)
I
AD[17]
IO
MEM_DQ[80]
DQ[80]
IO
(none)
I
AD[16]
IO
MEM_DQ[63]
DQ[63]
IO
RXD[7]
I
AD[15]
IO
MEM_DQ[62]
DQ[62]
IO
RXD[6]
I
AD[14]
IO
MEM_DQ[61]
DQ[61]
IO
RXD[5]
I
AD[13]
IO
MEM_DQ[60]
DQ[60]
IO
RXD[4]
I
AD[12]
IO
MEM_DQ[59]
DQ[59]
IO
RXD[3]
I
AD[11]
IO
MEM_DQ[58]
DQ[58]
IO
RXD[2]
I
AD[10]
IO
MEM_DQ[57]
DQ[57]
IO
RXD[1]
I
AD[9]
IO
MEM_DQ[56]
DQ[56]
IO
RXD[0]
I
AD[8]
IO
MEM_DQ[55]
DQ[55]
IO
RXDV
I
AD[7]
IO
Pin name*
GR740-UM-DS, Jun 2016, Version 1.4
28
www.cobham.com/gaisler
GR740
Table 27. Multiplexed SDRAM interface pins with PCI or Ethernet interfaces
ETHERNET1 function
SDRAM function
(MEM_IFWIDTH=HIGH,
(MEM_IFWIDTH=LOW) PCIMODE_ENABLE=HIGH)
ETH1_
PCI function
(MEM_IFWIDTH=HIGH,
PCIMODE_ENABLE=LOW)
PCI_
Function
Dir
Function
Dir
Function
Dir
MEM_DQ[54]
DQ[54]
IO
RXER
I
AD[6]
IO
MEM_DQ[53]
DQ[53]
IO
COL
I
AD[5]
IO
MEM_DQ[52]
DQ[52]
IO
CRS
AD[4]
IO
MEM_DQ[51]
DQ[51]
IO
MDINT
AD[3]
IO
MEM_DQ[50]
DQ[50]
IO
(none)
I
AD[2]
IO
MEM_DQ[49]
DQ[49]
IO
(none)
I
AD[1]
IO
MEM_DQ[48]
DQ[48]
IO
(none)
I
AD[0]
IO
MEM_DQ[47]
DQ[47]
IO
(none)
I
CBE[3]
IO
MEM_DQ[46]
DQ[46]
IO
(none)
I
CBE[2]
IO
MEM_DQ[45]
DQ[45]
IO
(none)
I
CBE[1]
IO
MEM_DQ[44]
DQ[44]
IO
(none)
I
CBE[0]
IO
MEM_DQ[43]
DQ[43]
IO
(none)
I
FRAME
IO
Pin name*
MEM_DQ[42]
DQ[42]
IO
(none)
I
REQ
O
MEM_DQ[41]
DQ[41]
IO
(none)
I
GNT
I
MEM_DQ[40]
DQ[40]
IO
(none)
I
IRDY
IO
MEM_DQ[39]
DQ[39]
IO
(none)
I
TRDY
IO
MEM_DQ[38]
DQ[38]
IO
(none)
I
PAR
IO
MEM_DQ[37]
DQ[37]
IO
(none)
I
PERR
IO
MEM_DQ[36]
DQ[36]
IO
(none)
I
SERR
IO
MEM_DQ[35]
DQ[35]
IO
(none)
I
DEVSEL
IO
MEM_DQ[34]
DQ[34]
IO
(none)
I
STOP
IO
MEM_DQ[33]
DQ[33]
IO
(none)
I
INTA
IO
MEM_DQ[32]
DQ[32]
IO
(none)
I
INTB
I
MEM_DQM[11]
DQM[11]
O
GTXCLK
I
M66EN
I
MEM_DQM[10] DQM[10]
O
TXCLK
I
HISTN
I
MEM_DQM[7]
DQM[7]
O
RXCLK
I
IDSEL
I
MEM_DQM[6]
DQM[6]
O
(none)
I
PCI_CLK
I
MEM_DQM[5]
DQM[5]
O
(none)
I
INTC
I
MEM_DQM[4]
DQM[4]
O
(none)
I
INTD
I
* See section 40.3 for pin assignments
3.3
Complete signal list
The listing below shows all interface signals, sorted by interface. Some of these signals are located on
shared pins as indicated in the table, therefore some physical pins will map to more than one entry in
this table. Section 3.2 and the device pin assignments in section 40.3 detail the pin sharing.
Table 28. External signals
Name
Usage
Pin sharing
Direction
Polarity
SYS_RESETN
System reset
No
In
Low
SYS_EXTLOCK
External clocks locked (for reset generation), tie high if unused
No
In
High
GR740-UM-DS, Jun 2016, Version 1.4
29
www.cobham.com/gaisler
GR740
Table 28. External signals
Name
Usage
Pin sharing
Direction
Polarity
SYS_CLK
System clock
No
In
-
MEM_EXTCLOCK
Alternate clock source for SDRAM interface
No
In
-
SPW_CLK
SpaceWire clock
No
In
Low
PROC_ERRORN
Processor error mode indicator
No
Out-Tri
Low
BREAK
Debug Support Unit and watchdog/processor break signal. See description of bootstrap signals.
No
In
High
DSU_EN
Debug Support Unit enable signal
No
In
High
DSU_ACTIVE
Debug Support Unit active signal
No
Out
High
PCIMODE_ENABLE
Enables PCI mode. See description of bootstrap signals
No
In
High
MEM_CLKSEL
Memory interface external clock select signal
No
In
-
MEM_IFWIDTH
Memory interface width select signal
No
In
-
MEM_CLK_OUT
SDRAM clock output
No
Out
-
MEM_CLK_OUT_DIFF_P
SDRAM clock output (differential)
No
Out
-
MEM_CLK_OUT_DIFF_N
SDRAM clock output (differential)
No
Out
-
MEM_CLK_IN
SDRAM clock input
No
In
-
MEM_WEN
SDRAM write enable
No
Out
Low
MEM_SN[1:0]
SDRAM chip select
No
Out
Low
MEM_RASN
SDRAM row address strobe
No
Out
Low
MEM_DQM[11:0]
SDRAM data mask
See 3.2.2
Out
Low
MEM_DQ[95:0]
SDRAM data and checkbit bus
See 3.2.2
BiDir
-
MEM_CKE[1:0]
SDRAM interface clock enable
No
Out
High
MEM_CASN
SDRAM column address strobe
No
Out
Low
MEM_BA[1:0]
SDRAM bank address
No
Out
-
MEM_ADDR[14:0]
SDRAM address and chip select 3,2
No
Out
-
JTAG_TCK
JTAG Clock
No
In
-
JTAG_TMS
JTAG Mode select
No
In
-
JTAG_TDI
JTAG Data in
No
In
-
JTAG_TDO
JTAG Data out
No
Out
-
JTAG_TRST
JTAG Reset
No
In
-
ETH0_TXER
Ethernet port 0, Transmit error
No
Out
High
ETH0_TXD[7:0]
Ethernet port 0, Transmitter output data
No
Out
-
ETH0_TXEN
Ethernet port 0, Transmitter enable
No
Out
High
ETH0_GTXCLK
Ethernet port 0, Gigabit clock
No
In
-
ETH0_TXCLK
Ethernet port 0, Transmitter clock
No
In
-
ETH0_RXER
Ethernet port 0, Receive error
No
In
High
ETH0_RXD[7:0]
Ethernet port 0, Receiver data
No
In
-
ETH0_RXDV
Ethernet port 0, Receive data valid
No
In
High
ETH0_RXCLK
Ethernet port 0, receiver clock
No
In
-
ETH0_MDIO
Ethernet port 0 and 1, Management Interface Data Input/Output
No
BiDir
-
ETH0_MDC
Ethernet port 0 and 1, Management Interface Data Clock
No
Out
-
GR740-UM-DS, Jun 2016, Version 1.4
30
www.cobham.com/gaisler
GR740
Table 28. External signals
Name
Usage
Pin sharing
Direction
Polarity
ETH0_COL
Ethernet port 0, Collision detected
No
In
High
ETH0_CRS
Ethernet port 0, Carrier sense
No
In
High
ETH0_MDINT
Ethernet port 0, Management Interface
Interrupt
No
In
Low
ETH1_TXER
Ethernet port 1, Transmit error
See 3.2.2
Out
High
ETH1_TXD[7:0]
Ethernet port 1, Transmitter output data
See 3.2.2
Out
-
ETH1_TXEN
Ethernet port 1, Transmitter enable
See 3.2.2
Out
High
ETH1_GTXCLK
Ethernet port 1, Gigabit clock
See 3.2.2
In
-
ETH1_TXCLK
Ethernet port 1, Transmitter clock
See 3.2.2
In
-
ETH1_RXER
Ethernet port 1, Receive error
See 3.2.2
In
High
ETH1_RXD[7:0]
Ethernet port 1, Receiver data
See 3.2.2
In
-
ETH1_RXDV
Ethernet port 1, Receive data valid
See 3.2.2
In
High
ETH1_RXCLK
Ethernet port 1, receiver clock
See 3.2.2
In
-
ETH1_COL
Ethernet port 1, Collision detected
See 3.2.2
In
High
ETH1_CRS
Ethernet port 1, Carrier sense
See 3.2.2
In
High
ETH1_MDINT
Ethernet port 1, Management Interface
Interrupt
See 3.2.2
In
Low
SPWD_TXD
SpaceWire Debug Communication Link
transmit data
See 3.2.1
Out
-
SPWD_TXS
SpaceWire Debug Communication Link
transmit strobe
See 3.2.1
Out
-
SPWD_RXD
SpaceWire Debug Communication Link
receive data
See 3.2.1
In
-
SPWD_RXS
SpaceWire Debug Communication Link
receive strobe
See 3.2.1
In
-
SPW_TXD_P[7:0]
SpaceWire router ports 1 - 8, transmit data,
positive
No
Out
-
SPW_TXD_N[7:0]
SpaceWire router ports 1 - 8, transmit data,
negative
No
Out
-
SPW_TXS_P[7:0]
SpaceWire router ports 1 - 8, transmit
strobe, positive
No
Out
-
SPW_TXS_N[7:0]
SpaceWire router ports 1 - 8, transmit
strobe, negative
No
Out
-
SPW_RXD_P[7:0]
SpaceWire router ports 1 - 8, receive data,
positive
No
In
-
SPW_RXD_N[7:0]
SpaceWire router ports 1 - 8, receive data,
negative
No
In
-
SPW_RXS_P[7:0]
SpaceWire router ports 1 - 8, receive strobe,
positive
No
In
-
SPW_RXS_N[7:0]
SpaceWire router ports 1 - 8, receive strobe,
negative
No
In
-
PCI_CLK
PCI clock
See 3.2.2
In
-
PCI_GNT
PCI grant
See 3.2.2
In
Low
PCI_IDSEL
PCI Device select during configuration
See 3.2.2
In
High
PCI_HOSTN
PCI System host. Low = Device will act as
PCI host
See 3.2.2
In
Low
PCI_AD[31:0]
PCI Address and Data bus
See 3.2.2
BiDir
High
PCI_CBE[3:0]
PCI Bus command and byte enable
See 3.2.2
BiDir
Low
GR740-UM-DS, Jun 2016, Version 1.4
31
www.cobham.com/gaisler
GR740
Table 28. External signals
Name
Usage
Pin sharing
Direction
Polarity
PCI_FRAME
PCI Cycle frame
See 3.2.2
BiDir
Low
PCI_IRDY
PCI Initiator ready
See 3.2.2
BiDir
Low
PCI_TRDY
PCI Target ready
See 3.2.2
BiDir
Low
PCI_DEVSEL
PCI Device select
See 3.2.2
BiDir
Low
PCI_STOP
PCI Stop
See 3.2.2
BiDir
Low
PCI_PERR
PCI Parity error
See 3.2.2
BiDir
Low
PCI_SERR
PCI System error
See 3.2.2
BiDir
Low
PCI_PAR
PCI Parity signal
See 3.2.2
BiDir
High
PCI_INTA
PCI Interrupt A
See 3.2.2
BiDir
Low
PCI_INTB
PCI Interrupt B
See 3.2.2
In
Low
PCI_INTC
PCI Interrupt C
See 3.2.2
In
Low
PCI_INTD
PCI Interrupt D
See 3.2.2
In
Low
PCI_REQ
PCI Request signal
See 3.2.2
Out
Low
PCI_M66EN
PCI 66 MHz enable signal
See 3.2.2
In
High
PROM_CEN[1:0]
PROM chip select
See 3.2.1
Out
Low
PROMIO_ADDR[27:0]
PROM/IO address
See 3.2.1
Out
-
PROMIO_OEN
PROM/IO Output Enable
No
Out
Low
PROMIO_WEN
PROM/IO Write Enable
No
Out
Low
PROMIO_BRDYN
PROM/IO Bus ready
No
In
Low
PROMIO_DATA[15:0]
PROM/IO data
See 3.2.1
BiDir
-
IO_SN
PROM/IO chip select
See 3.2.1
Out
Low
WDOGN
Watchdog output
No
Bidir
Low
GPIO[15:0]
General Purpose I/O
See 3.2.1 and
3.1
Bidir
-
GPIO2[21:0]
Second GPIO
See 3.2.1
UART0_TXD
UART 0, transmit data
See 3.2.1
Out
-
UART0_RXD
UART 0, receive data
See 3.2.1
In
-
UART0_RTSN
UART 0, request to sent
See 3.2.1
Out
Low
UART0_CTSN
UART 0, clear to send
See 3.2.1
In
Low
UART1_TXD
UART 1, transmit data
See 3.2.1
Out
-
UART1_RXD
UART 1, receive data
See 3.2.1
In
-
UART1_RTSN
UART 1, request to sent
See 3.2.1
Out
Low
UART1_CTSN
UART 1, clear to send
See 3.2.1
In
Low
GR1553_BUSARXEN
MIL-STD-1553 Bus A receiver enable
See 3.2.1
Out
High
GR1553_BUSARXP
MIL-STD-1553 Bus A receiver positive
input
See 3.2.1
In
High
GR1553_BUSARXN
MIL-STD-1553 Bus A receiver negative
input
See 3.2.1
In
High
GR1553_BUSATXIN
MIL-STD-1553 Bus A transmitter inhibit
No
Out
High
GR1553_BUSATXP
MIL-STD-1553 Bus A transmitter positive
output
See 3.2.1
Out
High
GR1553_BUSATXN
MIL-STD-1553 Bus A transmitter negative
output
See 3.2.1
In
High
GR1553_BUSBRXEN
MIL-STD-1553 Bus B receiver enable
See 3.2.1
Out
High
GR1553_BUSBRXP
MIL-STD-1553 Bus B receiver positive
input
See 3.2.1
In
High
GR740-UM-DS, Jun 2016, Version 1.4
32
www.cobham.com/gaisler
GR740
Table 28. External signals
Name
Usage
Pin sharing
Direction
Polarity
GR1553_BUSBRXN
MIL-STD-1553 Bus B receiver negative
input
See 3.2.1
In
High
GR1553_BUSBTXIN
MIL-STD-1553 Bus B transmitter inhibit
No
Out
High
GR1553_BUSBTXP
MIL-STD-1553 Bus B transmitter positive
output
See 3.2.1
Out
High
GR1553_BUSBTXN
MIL-STD-1553 Bus B transmitter negative
output
See 3.2.1
Out
High
GR1553_CLK
MIL-STD-1553 interface clock
No
In
-
SPI_MISO
SPI controller, master input, slave output
No
BiDir
-
SPI_MOSI
SPI controller, master output, slave input
No
BiDir
-
SPI_SCK
SPI controller, clock
No
BiDir
-
SPI_SEL
SPI controller, SPI select
No
In
Low
SPI_SLVSEL[1:0]
SPI controller, slave select
No
Out
Low
CAN_RXD[1:0]
CAN controller, receive data (shares pin
with PROM/IO interface)
See 3.2.1
In
-
CAN_TXD[1:0]
CAN controller, transmit data (shares pin
with PROM/IO interface)
See 3.2.1
Out
-
TESTEN
Test enable signal
No
In
High
PLL_BYPASS[2:0]
Bypass PLL. See description of bootstrap
signals.
No
In
High
PLL_IGNLOCK
Ignore PLL lock. See description of bootstrap signals.
No
In
High
PLL_LOCKED[5:0]
PLL coarse/fine lock. See description in
clocking section
No
Out
High
GR740-UM-DS, Jun 2016, Version 1.4
33
www.cobham.com/gaisler
GR740
4
Clocking and reset
4.1
Clock inputs
The table below specifies the clock inputs to the device.
Table 29. Clock inputs
Clock input
Description
Recommended frequency
SYS_CLK
System clock input. A clock based on this clock input via PLL
(unless PLL is bypassed) is used to clock the processors, on-chip
buses and on-chip peripherals.
50 MHz
MEM_EXTCLK
Alternative memory interface clock. Clock that either directly, or
50 MHz
through a PLL, provides an alternative clock for the SDRAM memory interface. See description in table 23, section 3.1.
SPW_CLK
SpaceWire clock. Clock that either directly, or through a PLL (recommended operating mode), provides a clock for the SpaceWire
interfaces.
50 MHz
JTAG_TCK
JTAG clock
10 MHz
ETH0_GTXCLK
Ethernet Gigabit MAC 0 clock
125 MHz
ETH0_TXCLK
Ethernet MAC 0 transmit clock
25 MHz
ETH0_RXCLK
Ethernet MAC 0 receive clock
25 MHz (MII)
125 MHz (GMII)
ETH1_GTXCLK
Ethernet Gigabit MAC 1 clock
125 MHz
ETH1_TXCLK
Ethernet MAC 1 transmit clock
25 MHz
ETH1_RXCLK
Ethernet MAC 1 receive clock
25 MHz (MII)
125 MHz (GMII)
PCI_CLK
PCI interface clock
66 or 33 MHz (TBD)
GR1553_CLK
MIL-STD-1553B interface clock
20 MHz
The design makes use of clock multipliers to create the system clock, memory interface clock, and the
SpaceWire transmitter clock.
4.2
Clock loop for SDRAM
Due to the drive strength limitations, the device may not be suitable to feed the clock directly to
SDRAMs at higher speeds. The device therefore implements a clock looping scheme for the SDRAM
clock, where the generated SDRAM clock goes out on either the single-ended MEM_CLK_OUT or
the differential MEM_CLK_OUT_DIFF output, should then on the PCB be split and fed both to the
SDRAM and back to the device’s mem_clk_in input. In the device, the MEM_CLK_IN input clocks
both the SDRAM interfacing registers as well as the SDRAM controller. See figure 1.
Both the differential and single-ended clock outputs are on by default after reset, software can during
boot disable the output that is unused in order to avoid unnecessary switching activity.
While what is described above is the intended usage, technically there is no requirement that the clock
fed to the MEM_CLK_IN input is related in frequency or phase to the clock going out the loop or any
other clock in the system. Other ways of generating the SDRAM clocks such as external PLL:s are
also possible.
Note: The external feedback loop is always required, no matter which clock source that is selected.
The memory controller SDRAM domain is never clocked internally, only through MEM_CLK_IN.
GR740-UM-DS, Jun 2016, Version 1.4
34
www.cobham.com/gaisler
GR740
4.3
Reset scheme
The device has an on-chip reset generator that creates a reset signal that is fed to the rest of the system.
This is asynchronously asserted when the external SYS_RESETN input is asserted and synchronously
deasserted a few cycles after the SYS_RESETN input has been deasserted.
The reset generation also considers the locking status of the PLLs, and will not deassert reset until the
PLL:s have achieved lock. In the event PLL lock is lost, the system will again go into reset. Only the
lock signals of PLLs that are used (not in bypass, or deselected by MEM_CLKSEL) are considered. If
external PLLs are also used on the board, a separate input SYS_EXTLOCK is available to allow also
including the lock status of these PLLs in the reset generation.
Where this default behavior is unwanted, the PLL_IGNLOCK bootstrap signal, when tied HIGH, will
cause the lock statuses of the internal PLLs to be ignored (treated as always in lock) in the reset generation. The SYS_EXTLOCK signal is never ignored. Since all the lock signals are available on package pins, custom lock handling can be implemented on board level.
The bootstrap signal sampling, the general purpose register bank, and the PLL reconfiguration module have separate reset generation that is only reset when the master resetn signal is asserted and will
not be affected by PLL lock status.
The JTAG_TRST input asynchronously resets the JTAG TAP in the device. This can be asserted at
any time while the device is running without affecting device function provided that a JTAG debug
access into the system is not currently in progress. The JTAG_TRST input must be asserted on powerup to ensure that the TAP instruction register can not power-up set to a test command. If JTAG is
unused, JTAG_TRST should be tied low on the board.
Other peripherals, such as Ethernet, SpaceWire and PCI are all reset via internal signals generated
from the SYS_RESETN input and PLL lock signals, as described above. To ensure proper reset of all
the clock domains in the device, care must be taken to ensure that all external clocks for interfaces that
will be used are active and toggling before the interface is enabled and ungated in the clock gating
unit.
GR740-UM-DS, Jun 2016, Version 1.4
35
www.cobham.com/gaisler
GR740
4.4
Clock multiplexing for main system clock, SDRAM and SpaceWire
The diagram below shows how the clocks are multiplexed in the design.
pll_bypass[0]
System clock
1
dsu_en
0
sys_clkin
SYSPLL
Debug bus and debug unit clocks
&
1
Gated CPU, FPU
and peripheral clocks
Clock Gating
Unit
0
pll_locked[1:0]
cpu idle
control registers
pll_bypass[1]
external clock split/loop
mem_extclk
MEMPLL
1
0
0
1
mem_clk_out
mem_clk_out_diff
mem_clk_in
mem_clksel
pll_locked[3:2]
Clock to
SDRAM controller
Clock to
SDRAM devices
pll_bypass[2]
to 1553 codec
gr1553_clk
spw_clk
1
SPWPLL
0
1
0
to TAP and scan chain
jtag_tck
to SPW codec
pll_locked[5:4]
to GRETH0
eth0*clk
Pos clk
mem_ifwidth AND pcimode_enable
Neg clk
mem_dqm[11]
mem_dqm[6]
mem_dqm[10]
mem_dqm[7]
&
to PCI core
&
to GRETH1
&
to GRETH1
&
to GRETH1
mem_ifwidth AND (NOT pcimode_enable)
Figure 1. GR740 clock multiplexing
GR740-UM-DS, Jun 2016, Version 1.4
36
www.cobham.com/gaisler
GR740
4.5
PLL control and configuration
Each PLL is put in power down mode whenever either:
•
The master reset signal SYS_RESETN is asserted
•
The PLL reconfiguration is commanded to reprogram the PLLs
•
The PLL is set to be bypassed using the PLL_BYPASS bootstrap signal
The rest of the PLL configuration is controlled by the PLL reconfiguration unit. When SYS_RESETN
is asserted this will be reset asynchronously to the default configuration. The reconfiguration unit can
then be reprogrammed to other PLL configurations via the general purpose register bank interface
(see section 35.2.4).
The configuration values tabulated below are the only supported configurations, other configurations
are invalid and may lead to malfunction. Note also that when overclocking the device by exceeding
the maximum clock frequencies given in the datasheet, correct functionality is not guaranteed and
power consumption may exceed typical values.
Table 30. Supported SYSPLL configurations
Memory clock
if
MEM_CLKSEL=LOW
SYSPLL
Config word
SYS_CLK
Input range
System clock
000010101
40-85 MHz (50 MHz
nom)
5 x SYS_CLK
1 x SYS_CLK
(250 MHz nom)
000001100
33.3-70 MHz
6 x SYS_CLK
1 x SYS_CLK
000001010
25-53 MHz
8 x SYS_CLK
2 x SYS_CLK
Comment
Default configuration
Table 31. Supported MEMPLL configurations
MEMPLL
Config word
MEM_EXTCLOCK
Input range
Memory clock if
MEM_CLKSEL=HIGH
Comment
000001010
25-53 MHz
2 x MEM_EXTCLK
Default configuration
Table 32. Supported SPWPLL configurations
SPWPLL
Config word
SPW_CLK
Input range
SpaceWire clock
Comment
000010000
25-53 MHz
8 x SPW_CLK
Default configuration
000001100
33.3-70 MHz
6 x SPW_CLK
GR740-UM-DS, Jun 2016, Version 1.4
37
www.cobham.com/gaisler
GR740
4.6
PLL watchdog
An additional watchdog is included in the system to detect if the main system clock stops running due
to PLL malfunction or other unforeseen issue. The PLL watchdog is combined with the regular
watchdog (GPTIMER0 timer 5) status and output on the WDOGN output. The watchdog has no other
effect on the system so if no watchdog functionality is wanted then the wdogn output can be ignored.
The PLL watchdog is clocked by the SYS_CLK input clock, and will trigger after 100 million SYS_CLK cycles (2.0 seconds at the nominal 50 MHz input frequency) unless it is restarted. It is restarted
whenever the GPTIMER0 TCTRL5 register is written (regardless of value written). Since this register
is written as part of the normal system watchdog handling, the PLL watchdog will not need any additional handling by software. The timeout value is fixed and can not be reprogrammed, and the current
status of the PLL watchdog is not accessible from software.
4.7
PCI clock
The PCI clock is taken from the MEM_DQM11 signal when the SDRAM is in half-width and PCI
mode is enabled. The device is capable of 33 MHz and 66 MHz operation (TBC). The input signal
PCI_M66EN must reflect the frequency of the input PCI clock. PCI_M66EN should be HIGH if the
PCI clock is a 66 MHz clock and LOW if the PCI clock frequency is 33 MHz.
4.8
MIL-STD-1553B clock
The 20 MHz clock for the MIL-STD-1553B codec is taken from the dedicated pin GR1553_CLK.
4.9
Clock gating unit
The design has a clock gating unit through which individual cores can have their AHB clocks
enabled/disabled and resets driven. The cores connected to the clock gating unit are listed in the table
below.
Table 33. Devices with gatable clock
Device
State after system reset
Ethernet MAC 0
The Ethernet MAC cores are gated off after reset unless the Debug
Support Unit is enabled via the DSU_EN signal.
Ethernet MAC 1
Ethernet MAC 1 is also disabled whenever mem_ifwidth is LOW
or PCI mode is enabled (PCIMODE_ENABLE = HIGH)
SpaceWire router
The SpaceWire router is disabled after reset unless general purpose
I/O line 11 (GPIO[11]) signals a prom-less system
PCI Target/Initiator and PCI DMA unit
Enabled after reset if PCIMODE_ENABLE=HIGH. Otherwise
disabled.
MIL-STD-1553B interface controller
Disabled after reset
CAN 2.0 controller
Disabled after reset
LEON4 Statistics unit
Disabled after reset
UART 0
Enabled after reset
UART 1
Enabled after reset
SPI controller
Disabled after reset
PROM/IO memory controller
Enabled after reset
The LEON4 processor cores will automatically be clock gated when the processor enters power-down
or halt state. A processor’s floating-point unit (GRFPU) will be clock gated when the corresponding
processor has disabled FPU operations by setting the %psr.ef bit to zero, or when the processor has
entered power-down/halt mode. After reset, processors 1 to 3 will be in power-down mode. Processor
0 will start executing if the BREAK bootstrap signal is LOW. If the BREAK bootstrap signal is HIGH
GR740-UM-DS, Jun 2016, Version 1.4
38
www.cobham.com/gaisler
GR740
then processor 0 will also enter power-down mode. If the device has debug mode enabled via the
DSU_EN signal (=HIGH) then the processors will enter debug mode instead of power-down mode.
For more information see chapter about the clock gating unit, section 30.
4.10
Debug AHB bus clocking
All cores on the Debug AHB bus will be gated off when the DSU_EN signal is low.
4.11
Notes on Ethernet interface clock and mode switch
The Ethernet interface transmit clocks (ETH0_TXCLK and ETH1_TXCLK) are used internally in the
device to clock registers that selects between 10/100 and 1000 Mbit (Gigabit) mode for the respective
Ethernet controller. The default PHY (Ethernet transceiver) behaviour is to enter 10/100 Mbit mode
after reset. When a mode switch is made to 1000 Mbit mode, a signal will change internally in the
Ethernet controller. For this signal change to propagate through the register that selects between 10/
100 and 1000 Mbit mode the corresponding TXCLK must be present. Ethernet PHYs may disable the
TXCLK when entering 1000 Mbit mode and this may cause the internal register value in the GR740
to remain at the 10/100 Mbit value after the PHY has entered 1000 Mbit operation. When this happens
the system will not be able to transmit Ethernet traffic.
When the Ethernet debug communication link (EDCL) is enabled then the Ethernet controller will
automatically try to configure the PHY after reset. If the device is connected to a Gigabit network then
care must be taken to ensure that the TXCLK is available after the PHY has switched to 1000 Mbit
mode.
If the device will only be used on a 10/100 Mbit network then the TXCLK inputs can be connected
directly to the PHY 10/100 transmit clock. If the device will only be used on a 1000 Mbit network
then the Gigabit transmit clock can be connected to the TXCLK input.
If the device should adapt to both 10/100 Mbit networks and 1000 Mbit networks then the TXCLK
input(s) should be connected to the PHY 10/100 transmit clock(s). Software then needs to perform a
special sequence when the PHY has determined that it is connected to a Gigabit network: If the software driver finds that the device is connected to a Gigabit network then software needs to force the
PHY into 10/100 Mbit mode in order to enable the TXCLK. Software can then re-enable 1000 Mbit
operation. Note that this allows the system to adapt to both 10/100 networks and 1000 Mbit networks.
With this configuration, the EDCL will still be unavailable after reset when connected to a Gigabit
network.
GR740-UM-DS, Jun 2016, Version 1.4
39
www.cobham.com/gaisler
GR740
5
Technical notes
5.1
GRLIB AMBA plug&play scanning
The bus structure in this design requires some special consideration with regard to plug&play scanning. The default behavior of GRLIB AMBA plug&play scanning routines is to start scanning at
address 0xFFFFF000. If any AHB/AHB bridges, APB bridges or L2 cache controllers are detected
during the scan, the general scanning routine traverses the bridge and reads the plug&play information from the bus behind the bridge. In this design, the default 0xFFFFF000 address gives plug&play
information for the Processor AHB bus. This plug&play area contains information which allows software to detect all devices on the Processor, Slave I/O, Master I/O and Memory AHB buses.
The plug&play information on the Processor bus does not contain any references to the plug&play
information on the Debug AHB bus, nor can the cores on the Debug AHB bus be accessed from the
Processor AHB bus as the buses are connected using a uni-directional bridge. In order to detect the
cores on the Debug AHB bus, the debug monitor used must be informed about the memory map of the
bus, or be instructed to start plug&play scanning at address 0xEFFFF000 from where all the other
plug&play areas in the system can be found.
Depending on the debug monitor used, the monitor may detect that it connects to a GR740 design and
start scanning on the Debug AHB bus (this applies to GRMON2 from Cobham Gaisler). Otherwise
the address 0xEFFFF000 should be specified to the monitor. In the case where the monitor detects that
it is connected to a GR740 design, it may be necessary to force the monitor to start scanning at the
default address 0xFFFFF000 when connecting with a debug monitor through the Master I/O bus, from
which the Debug AHB bus cannot be accessed (this is not required for GRMON2).
5.2
Processor register file initialisation and data scrubbing
Please refer to section 6.11.
5.3
PROM-less systems and SpaceWire RMAP
The system has support for PROM less operation where system software is uploaded by an external
entity. In order to allow system software to be uploaded via RMAP the bootstrap signal GPIO[11]
should be low in order to not clock gate off the SpaceWire router after system reset. The IOMMU will
be in pass-through after reset allowing an external entity to upload software, change the processor
reset start address, and wake the processors up via the multiprocessor interrupt controller’s register
interface. In order to prevent the processor from starting execution, the external BREAK signal
should be asserted. This will also prevent the timer unit’s watchdog timer from being started. Note
that the PLL watchdog described in section 4.6 will still be active and external units must either pet
this watchdog or have the WDOGN signal disconnected from reset circuitry to prevent reset of the
device.
If the system has a boot PROM available it is recommended to have the SpaceWire router gated off
after reset by setting the bootstrap signal GPIO[11] high during system reset. If router functionality
needs to be immediately available, the designer should consider disabling RMAP or enable IOMMU
protection early in the software boot process so that external entities cannot interfere with system
operation. It takes 20 microseconds for the SpaceWire links to enter run state. Before that, incoming
RMAP traffic cannot enter the system. This leaves time (4000 cycles at 200 MHz system frequency)
for the processors to disable RMAP via a register write, or to set up rudimentary IOMMU protection.
GR740-UM-DS, Jun 2016, Version 1.4
40
www.cobham.com/gaisler
GR740
5.4
System integrity and debug communication links
The debug communication links have unrestricted access to all parts of the system. When the Debug
AHB bus is clock gated off via the external dsu_en signal, all debug communication links will be disabled. However, the Ethernet Debug Communication Links (EDCLs) can still be enabled via the
Ethernet controllers’ register interfaces. Since the Debug AHB bus is gated off, the only path for
EDCL traffic into the system is through the IOMMU. Since EDCL traffic flows through the same
AHB master interface as normal Ethernet traffic the IOMMU may not provide adequate protection. To
ensure that EDCL traffic cannot be harmful, even if accidentally enabled, it is recommended to tie
GPIO[9:8] HIGH during system reset in order to force EDCL traffic onto the gated Debug AHB bus.
5.5
ASMP configurations
The system supports running different OS instances on each of the processor cores. The use of ASMP
configurations is eased by:
•
The multiprocessor interrupt controller that contains four internal interrupt controllers. This
means that each OS (up to four) can have direct access to its own interrupt controller. It is also
possible to run two SMP operating systems simultaneously.
•
The availability of several general purpose timer units allows each OS to have a dedicated timer
unit.
•
All peripheral registers are mapped on 4 KiB address boundaries. This allows using the system’s
memory management units to provide separation between operating systems.
•
The I/O memory management unit (IOMMU) can prevent DMA capable peripheral controllers
belonging to one OS from overwriting memory areas belonging to another OS.
•
The L2 cache supports replacement policies based on AHB master bus index. This means that the
L2 cache can be configured so that one processor cannot evict data allocated by accesses from
another processor.
The system does not provide full separation between operating systems. The main memory interface
and AMBA buses are shared. Since space separation is provided by CPU memory management units,
it is possible for one operating system to disable the memory management unit and access memory
areas assigned to another operating system.
5.6
Clock gating
Some peripherals are clock gated after reset (see section 4.9). Software drivers for LEON systems
generally assume that the peripheral clocks are enabled and the clock gating unit should be configured
by the bootloader or debug tool. The GRMON debugger has support for enabling all clocks when connecting to the device and clocks for specific peripherals can also be enabled via the command line
interface. Please see the GRMON user manual and operating system documentation for more information.
GR740-UM-DS, Jun 2016, Version 1.4
41
www.cobham.com/gaisler
GR740
5.7
Software portability
5.7.1
Instruction set architecture
The LEON4 processor used in this design implements the SPARC V8 instruction set architecture.
This means that any compiler that produces valid SPARC V8 executables can be used. Full instruction
set compatibility is kept with LEON2FT and LEON3FT applications. The LEON4 processor implements the SPARC V9 compare and swap (CAS) instruction. This instruction is not available on
LEON2FT and is optional for LEON3FT implementations. Programs that utilize this instruction may
therefore not be backward compatible with legacy systems. See also information about the memory
map in section 5.7.4 below.
5.7.2
Peripherals
All peripherals in the design are IP cores from Cobham Gaisler’s GRLIB IP library. Standard GRLIB
software drivers can be used.
For software driver development, this document describes the capabilities offered by the GR740 system. In order to write a generic driver for a GRLIB IP core, that can be used on all systems based on
GRLIB, please also refer to the generic IP core documentation. Note, however, that the generic documentation may describe functionality not present in this implementation and that this datasheet supersedes any IP core documentation.
5.7.3
Plug and play
Standard GRLIB AMBA plug&play layout is used (see sections 37 and 38). The same software routines used for typical LEON/GRLIB systems can be used.
5.7.4
Memory map
Many LEON2FT and LEON3FT systems use a memory map with ROM mapped at 0x0 and RAM
mapped at 0x40000000. This design has RAM mapped at 0x0 and ROM mapped at 0xC0000000.
This does in general not affect applications running on an operating system but it has implications for
software running on bare-metal. Please refer to operating system documentation to see if and how
special consideration can be taken for systems with RAM at 0x0 and ROM at 0xC0000000.
Differences in memory map may also mean that prebuilt system software images may not be portable
between systems, and need to be rebuilt, even if software makes use of plug’n’play to discover
peripheral register addresses.
5.8
Level-2 cache
The Level-2 (L2) cache controller is disabled after system reset. From a performance perspective it is
recommended that the L2 cache is enabled as early in the boot process as possible. The L2 cache contents must be invalidated when the cache is enabled, see section 9 for details.
GR740-UM-DS, Jun 2016, Version 1.4
42
www.cobham.com/gaisler
GR740
5.9
Time synchronisation
5.9.1
Overview
The system includes hardware functionality for time synchronization where the system can be configured to save the current time value on certain events. The system also supports toggling GPIO lines on
timer ticks and when the current time value is latched. The event triggering the time latch and GPIO
toggle responses is fully handled in hardware without requiring involvement from software, except
for the need for initial configuration of the peripherals. This section provides an overview of the available resources, please refer to the documentation of the peripherals for further details.
The following events can trigger time latching:
•
Assertion of any of the interrupt lines, includes CAN controller RX and TX events and also
events signalled via GPIO inputs since the GPIO ports can be configured to generate interrupts.
•
SpaceWire Time-Code reception (signalled via router tick outputs 0 - 3, via SpaceWire router
AMBA port interrupts and via TDP controller)
•
MIL-STD-1553B reception of synchronize mode command (when operating as RT).
The following events can trigger a synchronization message or action:
•
GPIO lines can be toggled on GPTIMER 0 timer ticks and all events that trigger time latching
•
The TDP controller can intiate transmission of SpaceWire Time-Codes
•
MIL-STD-1553B message transmission can be triggered by the timer 3 tick on GPTIMER 0
(when operating as BC)
5.9.2
Available timers
The following timers are available in the system:
•
Processor up-counter - The up-counters accessible via internal registers, %asr22 and asr23, in the
processors provide a 63-bit value. All four processors share the same counter. The low part of
this counter is also used for interrupt timestamping, the system’s trace buffers, and performance
counter timestamping.
•
General purpose timer units - Five general purpose timers units (GPTIMER0 - 4) provides 21 32bit timers. GPTIMER0 has five timers where the last timer is used as the system watchdog and
GPTIMER1-4 each has four 32-bit timers. All timers units are capable of latching or setting the
time based on events on the interrupt bus or on separate inputs (called external events).
•
The TDP controller provides basic time keeping functions such as Elapsed Time counter according to the CCSDS Unsegmented Code specification. It provides support for setting and sampling
the Elapsed Time counter. It also includes a frequency synthesizer with which a binary frequency
is generated to drive the Elapsed Time counter. The TDP controller also implements the Time
Distribution Protocol (TDP). The aim of TDP is to distribute and synchronize time across a
SpaceWire network. The TDP controller also provides external datation services, there are four
external datation services implemented which can latch the elapsed time counter when a specified event occurs. All external datation services share the same event inputs. The event on which
time stamp must occur is configurable individually (using mask registers) for all the external
datation services
GR740-UM-DS, Jun 2016, Version 1.4
43
www.cobham.com/gaisler
GR740
5.9.3
Generation of synchronization messages and events
The systems first general purpose I/O port supports toggling external signals GPIO(15:0) based on
on-chip events. The PULSE register controlling this functionality is described in section 27.3.11. The
table below gives an overview of the connections:
Table 34. Events that can invert GPIO output
GPIO lines number
Event
0
GPTIMER 0 tick 0
1
GPTIMER 0 tick 1
2
GPTIMER 0 tick 2
3
GPTIMER 0 tick 3
4
GPTIMER 0 tick 4
5
TDP controller CTICK
A pulse is generated when SpaceWire Time-Code is transmitted
when TDP controller is acting as initiator.
A pulse is also generated when a diagnostic SpaceWire Time-Code
is generated when TDP controller is acting as target.
6
TDP controller JTICK
The incoming SpaceWire Time-Code provides an output pulse
when the TDP controller is acting as target, this output is used to
visualize the jitter in incoming SpaceWire Time-Codes.
7
TDP controller datatation pulse 0
8
TDP controller datatation pulse 1
9
TDP controller datatation pulse 2
10
TDP controller datatation pulse 3
11
GPTIMER 0 latch disable
12
GPTIMER 1 latch disable
13
GPTIMER 2 latch disable
14
GPTIMER 3 latch disable
15
GPTIMER 4 latch disable
Note that the connection of the GPTIMER latch disable events and TDP controller datatation pulses
allow the system to be configured so that any interrupt in the device can invert the value of the corresponding GPIO lines. In this case the TDP controller and timer unit act as filters since they have mask
registers to select which interrupts, or other event sources, that should cause time to be latched.
GR740-UM-DS, Jun 2016, Version 1.4
44
www.cobham.com/gaisler
GR740
6
LEON4 - Fault-tolerant High-performance SPARC V8 32-bit Processor
6.1
Overview
LEON4 is a 32-bit processor core conforming to the IEEE-1754 (SPARC V8) architecture [SPARC]
with a subset of the V8E extensions [V8E]. It is designed for embedded applications, combining high
performance with low complexity and low power consumption.
The LEON4 core has the following main features: 7-stage pipeline with Harvard architecture, separate instruction and data caches, hardware multiplier and divider, on-chip debug support and multiprocessor extensions.
The LEON4 processors in this device have fault-tolerance against SEU errors. The fault-tolerance is
focused on the protection of on-chip RAM blocks, which are used to implement IU/FPU register files
and the L1 cache memory.
4-Port Register File
Trace Buffer
IEEE-754 FPU
7-Stage
Integer pipeline
HW MUL/DIV
I-Cache
ITLB
Debug port
Debug support unit
Interrupt port
Interrupt controller
D-Cache
SRMMU
DTLB
AHB I/F
AMBA AHB Master (128-bit)
Figure 2. LEON4 processor core block diagram
6.1.1
Integer unit
The LEON4 integer unit is implemented according to the SPARC V8 manual [SPARC], including
hardware multiply and divide instructions. The number of register windows is eight. The pipeline
consists of 7 stages with a separate instruction and data cache interface.
6.1.2
Cache sub-system
LEON4 has a cache system consisting of a separate instruction and data cache. Both caches have
four ways, four KiB/way, and 32 bytes per line. The instruction cache maintains one valid bit per
cache line and uses streaming during line-refill to minimize refill latency. The data cache has one
valid bit per cache line, uses write-through policy and implements a double-word write-buffer. Bussnooping on the AHB bus maintains cache coherency for the data cache.
6.1.3
Floating-point unit and co-processor
The LEON4 integer unit provides interfaces for the high-performance GRFPU floating-point
unit.´The floating-point processor executes in parallel with the integer unit, and does not block the
operation unless a data or resource dependency exists. The floating-point controller and floating-point
unit are further describes in sections 7 and 8.
GR740-UM-DS, Jun 2016, Version 1.4
45
www.cobham.com/gaisler
GR740
6.1.4
Memory management unit
Each processor core contains a SPARC V8 Reference Memory Management Unit (SRMMU). The
SRMMU implements the full SPARC V8 MMU specification, and provides mapping between multiple 32-bit virtual address spaces and physical memory. A three-level hardware table-walk is implemented, and the MMU has 16 instruction and 16 data fully associative TLB entries.
6.1.5
On-chip debug support
The LEON4 pipeline includes functionality to allow non-intrusive debugging on target hardware. To
aid software debugging, up to four watchpoint registers can be enabled. Each register can cause a
breakpoint trap on an arbitrary instruction or data address range. When the (optional) debug support
unit is attached, the watchpoints can be used to enter debug mode. Through a debug support interface,
full access to all processor registers and caches is provided. The debug interfaces also allows single
stepping, instruction tracing and hardware breakpoint/watchpoint control. An internal trace buffer can
monitor and store executed instructions, which can later be read out via the debug interface.
6.1.6
Interrupt interface
LEON4 supports the SPARC V8 interrupt model with a total of 15 asynchronous interrupts. The interrupt interface provides functionality to both generate and acknowledge interrupts.
6.1.7
AMBA interface
The cache system implements an AMBA AHB master to load and store data to/from the caches. The
interface is compliant with the AMBA-2.0 standard. During line refill, incremental burst are generated to optimise the data transfer. The AMBA interface makes use of the full width of the 128-bit bus
on cache line fills. The processor also has a snoop AHB slave input port which is used to monitor the
accesses made by other masters on the processor AHB bus.
6.1.8
Power-down mode
The LEON4 processor core implements a power-down mode, which halts the pipeline and caches
until the next interrupt. The processor supports clock gating during the power down period by providing a clock-enable signal to the system’s clock gating unit. A small part of the processor is always
clocked, to check for wake-up conditions and maintain cache coherency.
6.1.9
Multi-processor support
LEON4 is designed to be used in multi-processor systems. Each processor has a unique index to allow
processor enumeration. The write-through caches and snooping mechanism guarantees memory
coherency in shared-memory systems.
GR740-UM-DS, Jun 2016, Version 1.4
46
www.cobham.com/gaisler
GR740
6.2
LEON4 integer unit
6.2.1
Overview
The LEON4 integer unit implements the integer part of the SPARC V8 instruction set. The implementation is focused on high performance and low complexity. The LEON4 integer unit has the following
main features:
•
7-stage instruction pipeline
•
Separate instruction and data cache interface
•
Support for eight register windows
•
Hardware multiplier and Radix-2 divider (non-restoring)
•
Static branch prediction
•
Single-vector trapping for reduced code size
call/branch address
I-cache
+1
Add
‘0’ jmpa tbr
data address
f_pc
Fetch
d_inst
d_pc
r_inst
r_pc
Decode
r_imm
rd
64-bit 4-port register file
rs2
rs1
rs3
Register Access
y, tbr, wim, psr
e_inst
e_pc
rs1
Execute
op2
alu/shift
64
dcache write data
mul/div
y
e pc
m_inst
stdata
m_pc
result
32
dcache address
64
dcache read data
m_y
D-cache
Memory
x_inst
x_pc
xres
x_y
w_inst
w_pc
wres
Y
Exception
Write-back
30
tbr, wim, psr
Figure 3. LEON4 integer unit datapath diagram
6.2.2
Instruction pipeline
The LEON4 integer unit uses a single instruction issue pipeline with 7 stages:
GR740-UM-DS, Jun 2016, Version 1.4
47
www.cobham.com/gaisler
GR740
1.
2.
3.
4.
5.
6.
7.
FE (Instruction Fetch): If the instruction cache is enabled, the instruction is fetched from the instruction cache.
Otherwise, the fetch is forwarded to the memory controller. The instruction is valid at the end of this stage and is
latched inside the IU.
DE (Decode): The instruction is decoded and the CALL/Branch target addresses are generated.
RA (Register access): Operands are read from the register file or from internal data bypasses.
EX (Execute): ALU, logical, and shift operations are performed. For memory operations (e.g., LD) and for JMPL/
RETT, the address is generated.
ME (Memory): Data cache is accessed. Store data read out in the execution stage is written to the data cache at this
time.
XC (Exception) Traps and interrupts are resolved. For cache reads, the data is aligned.
WR (Write): The result of ALU and cache operations are written back to the register file.
Table 35 lists the cycles per instruction (assuming cache hit and no icc or load interlock):
Table 35. Instruction timing
Instruction
Cycles (MMU disabled)
JMPL, RETT
3
SMUL/UMUL
1*
SDIV/UDIV
35
Taken Trap
5
Atomic load/store
5
All other instructions
1
* Multiplication cycle count is 1 clock (1 clock issue rate, 2 clock data latency), for the 32x32 multiplie
Additional conditions that can extend an instructions duration in the pipeline are listed in the table and
text below.
Branch interlock: When a conditional branch or trap is performed 1-2 cycles after an instruction
which modifies the condition codes, 1-2 cycles of delay is added to allow the condition to be computed. If static branch prediction is enabled, this extra delay is incurred only if the branch is not taken.
Load delay: When using data shortly after the load instruction, the second instruction will be delayed
to satisfy the pipeline’s load delay.
Mul latency: For pipelined multiplier implementations there is 1 cycle extra data latency, accessing
the result immediately after a MUL will then add one cycle pipeline delay.
Hold cycles: During cache miss processing or when blocking on the store buffer, the pipeline will be
held still until the data is ready, effectively extending the execution time of the instruction causing the
miss by the corresponding number of cycles. Note that since the whole pipeline is held still, hold
cycles will not mask load delay or interlock delays. For instance on a load cache miss followed by a
data-dependent instruction, both hold cycles and load delay will be incurred.
FPU: The floating-point unit or coprocessor may need to hold the pipeline or extend a specific
instruction.
GR740-UM-DS, Jun 2016, Version 1.4
48
www.cobham.com/gaisler
GR740
Certain specific events that cause these types of locks and their timing are listed in table 36 below.
Table 36. Event timing
Event
Cycles
Instruction cache miss processing, MMU disabled
3 + mem latency
Instruction cache miss processing, MMU enabled
5 + mem latency
Data cache miss processing, MMU disabled (read), L2 hit
3 + mem latency
Data cache miss processing, MMU disabled (write), write-buffer empty
Data cache miss processing, MMU enabled (read)
0
5 + mem latency
Data cache miss processing, MMU enabled (write), write-buffer empty
MMU page table walk
0
10 + 3 * mem latency
Branch prediction miss, branch follows ICC setting
2
Branch prediction miss, one instruction between branch and ICC setting
1
Pipeline restart due to register file or cache error correction
7
6.2.3
SPARC Implementor’s ID
Cobham Gaisler is assigned number 15 (0xF) as SPARC implementor’s identification. This value is
hard-coded into bits 31:28 in the %psr register. The version number for LEON4 is 3 (same as for
LEON3 to provide software compatibility), which is hard-coded in to bits 27:24 of the %psr.
6.2.4
Divide instructions
Full support for SPARC V8 divide instructions is provided (SDIV, UDIV, SDIVCC & UDIVCC). The
divide instructions perform a 64-by-32 bit divide and produce a 32-bit result. Rounding and overflow
detection is performed as defined in the SPARC V8 manual.
6.2.5
Multiply instructions
The LEON processor supports the SPARC integer multiply instructions UMUL, SMUL UMULCC
and SMULCC. These instructions perform a 32x32-bit integer multiply, producing a 64-bit result.
SMUL and SMULCC performs signed multiply while UMUL and UMULCC performs unsigned
multiply. UMULCC and SMULCC also set the condition codes to reflect the result. The multiply
instructions are performed using a 32x32 pipelined hardware multiplier.
6.2.6
Multiply and accumulate instructions
This implementation does not support multiply-and-accumulate (UMAC; SMAC) instructions.
6.2.7
Compare and Swap instruction (CASA)
LEON4 implements the SPARC V9 Compare and Swap Alternative (CASA) instruction. The CASA
operates as described in the SPARC V9 manual. The instruction is privileged, except when setting
ASI = 0xA (user data).
GR740-UM-DS, Jun 2016, Version 1.4
49
www.cobham.com/gaisler
GR740
6.2.8
Branch prediction
Static branch prediction can be optionally be enabled, and reduces the penalty for branches preceded
by an instruction that modifies the integer condition codes. The predictor uses a branch-always strategy, and starts fetching instruction from the branch address. On a prediction hit, 1 or 2 clock cycles
are saved, and there is no extra penalty incurred for misprediction as long as the branch target can be
fetched from cache.
6.2.9
Register file data protection
The integer and FPU register files are protected against soft errors. Data errors will then be transparently corrected without impact at application level. Correction is done for the read data value. The
error remains in the register file and will be corrected on the next write to the register file position.
6.2.10 Hardware breakpoints
The integer unit can supports four hardware breakpoints. Each breakpoint consists of a pair of ancillary state registers (see section 6.10.4). Any binary aligned address range can be watched for instruction or data access, and on a breakpoint hit, trap 0x0B is generated.
6.2.11 Instruction trace buffer
The instruction trace buffer consists of a circular buffer that stores executed instructions. This is
enabled and accessed only through the processor’s debug port via the Debug Support Unit. When
enabled, the following information is stored in real time, without affecting performance:
•
Instruction address and opcode
•
Instruction result
•
Load/store data and address
•
Trap information
•
30-bit time tag
The operation and control of the trace buffer is further described in section 13.4. Note that each processor has its own trace buffer allowing simultaneous tracing of all instruction streams.
The time tag value in the trace buffer has the same time source as the up-counter described in section
6.10.3.
6.2.12 Processor configuration register
The ancillary state register 17 (%asr17) provides information on implementation-specific characteristics for the processor. This can be used to enhance the performance of software. See section 6.10.4 for
layout.
GR740-UM-DS, Jun 2016, Version 1.4
50
www.cobham.com/gaisler
GR740
6.2.13 Exceptions
LEON4 adheres to the general SPARC trap model. The table below shows the implemented traps and
their individual priority. When PSR (processor status register) bit ET=0, an exception trap causes the
processor to halt execution and enter error mode, and the external error signal will then be asserted.
Table 37. Trap allocation and priority
Trap
TT
Pri
Description
Class
reset
0x00
1
Power-on reset
Interrupting
data_store_error
0x2b
2
write buffer error during data store
Interrupting
instruction_access_exception 0x01
3
Error or MMU page fault during instruction fetch
Precise
privileged_instruction
0x03
4
Execution of privileged instruction in user mode
Precise
illegal_instruction
0x02
5
UNIMP or other un-implemented instruction
Precise
fp_disabled
0x04
6
FP instruction while FPU disabled
Precise
cp_disabled
0x24
6
CP instruction while Co-processor disabled
Precise
watchpoint_detected
0x0B
7
Hardware breakpoint match
Precise
window_overflow
0x05
8
SAVE into invalid window
Precise
window_underflow
0x06
8
RESTORE into invalid window
Precise
mem_address_not_aligned
0x07
10
Memory access to un-aligned address
Precise
fp_exception
0x08
11
FPU exception
Deferred
cp_exception
0x28
11
Co-processor exception
Deferred
data_access_exception
0x09
13
Access error during data load, MMU page fault
Precise
tag_overflow
0x0A
14
Tagged arithmetic overflow
Precise
division_by_zero
0x2A
15
Divide by zero
Precise
trap_instruction
0x80 - 0xFF
16
Software trap instruction (TA)
Precise
interrupt_level_15
0x1F
17
Asynchronous interrupt 15
Interrupting
interrupt_level_14
0x1E
18
Asynchronous interrupt 14
Interrupting
interrupt_level_13
0x1D
19
Asynchronous interrupt 13
Interrupting
interrupt_level_12
0x1C
20
Asynchronous interrupt 12
Interrupting
interrupt_level_11
0x1B
21
Asynchronous interrupt 11
Interrupting
interrupt_level_10
0x1A
22
Asynchronous interrupt 10
Interrupting
interrupt_level_9
0x19
23
Asynchronous interrupt 9
Interrupting
interrupt_level_8
0x18
24
Asynchronous interrupt 8
Interrupting
interrupt_level_7
0x17
25
Asynchronous interrupt 7
Interrupting
interrupt_level_6
0x16
26
Asynchronous interrupt 6
Interrupting
interrupt_level_5
0x15
27
Asynchronous interrupt 5
Interrupting
interrupt_level_4
0x14
28
Asynchronous interrupt 4
Interrupting
interrupt_level_3
0x13
29
Asynchronous interrupt 3
Interrupting
interrupt_level_2
0x12
30
Asynchronous interrupt 2
Interrupting
interrupt_level_1
0x11
31
Asynchronous interrupt 1
Interrupting
The prioritization follows the SPARC V8 standard.
The fp_exception trap is deferred. The data_store_error is delivered as a deferred exception but is
non-resumable and therefore classed as interrupting in above table.
6.2.14 Single vector trapping (SVT)
Single-vector trapping (SVT) is an SPARC V8e [V8E] option to reduce code size for embedded applications. When enabled, any taken trap will always jump to the reset trap handler (%tbr.tba + 0). The
GR740-UM-DS, Jun 2016, Version 1.4
51
www.cobham.com/gaisler
GR740
trap type will be indicated in %tbr.tt, and must be decoded by the shared trap handler. SVT is enabled
by setting bit 13 in %asr17.
6.2.15 Address space identifiers (ASI)
In addition to the address, a SPARC processor also generates an 8-bit address space identifier (ASI),
providing up to 256 separate, 32-bit address spaces. During normal operation, the LEON4 processor
accesses instructions and data using ASI 0x8 - 0xB as defined in the SPARC standard. Using the
LDA/STA instructions, alternative address spaces can be accessed. The different available ASIs are
described in section 6.9.
6.2.16 Partial WRPSR
Partial write %PSR (WRPSR) is a SPARC V8e option that allows WRPSR instructions to only affect
the %PSR.ET field. If the WRPSR instruction’s rd field is non-zero, then the WRPSR write will only
update ET.
6.2.17 Power-down
The processor has a power-down feature to minimize power consumption during idle periods. The
power-down mode is entered by performing a WRASR instruction to %asr19:
wr %g0, %asr19
During power-down, the pipeline is halted until the next interrupt occurs. Signals inside the processor
pipeline and caches are then static, reducing power consumption from dynamic switching. The default
setting of the clock-gating unit is to also disable the processor and FPU clock when the processor
enters this idle mode
Note: %asr19 must always be written with the data value zero to ensure compatiblity with future
extensions.
Note: This instruction must be performed in supervisor mode with interrupts enabled.
When resuming from power-down, the pipeline will be re-filled from the point of power-down and the
first instruction following the WRASR instruction will be executed prior to taking the interrupt trap.
Up to six instructions after the WRASR instruction will be fetched (possibly with cache miss if they
are not in cache) prior to fetching the trap handler.
6.2.18 Processor reset operation
The following table indicates the reset values of the registers which are affected by system reset. See
also reset values specified for other registers, such as the cache control register in sections 6.9 and
6.10. All other registers maintain their value or are undefined.
Table 38. Processor reset values
Register
Reset value
Trap Base Register
Trap Base Address field reset to 0xC0000000
PC (program counter)
0xC0000000
nPC (next program counter)
0xC0000004
PSR (processor status register)
ET=0, S=1
By default, the execution will start from address 0xC0000000. This can be overridden by setting the
reset start address register on the interrupt controller.
GR740-UM-DS, Jun 2016, Version 1.4
52
www.cobham.com/gaisler
GR740
6.2.19 Multi-processor systems
The LEON4 processor supports symmetric multi-processing (SMP) and asymmetric multi-processing
(ASMP) configurations. The ID of the processor on which the code is executing can be read out by
reading the index field of the LEON4 configuration register.
After system reset, only the first processor will start (note that this depends on the value of the external signal BREAK. If BREAK is high after system reset. The first processor will either be halted or go
into debug mode, depending on the value of external signal DSU_EN. All other processors will
remain halted in power-down mode.
After the system has been initialized, the remaining processors can be started by writing to the MP
status register, located in the multi-processor interrupt controller. The halted processors start execuing
from the reset address. Note that if the reset start address is changed (via the interrupt controller) then
the processors must be started via the interrupt controller’s Processor boot register.
6.3
Cache system
6.3.1
Overview
The LEON4 processor pipeline implements a Harvard architecture with separate instruction and data
buses, connected to two separate cache controllers. As long as the execution does not cause a cache
miss, the cache controllers can serve one beat of an instruction fetch and one data load/store per cycle,
keeping the pipeline running at full speed.
On cache miss, the cache controller will assert a hold signal freezing the IU pipeline, and after delivering the data the hold signal is again lifted so execution continues. For accessing the bus, the cache
controllers share the same AHB connection to the on-chip bus. Certain parts of the MMU (table walk
logic) are also shared between the two caches.
Another important component included in the data cache is the write buffer, allowing stores to proceed in parallel to executing instructions.
Cachability (memory areas that are cachable) for both caches is described in section 6.7.2.
6.3.2
Cache operation
Each cache controller has two main memory blocks, the tag memory and the data memory. At each
address in the tag memory, a number of cache entries, ways, are stored for a certain set of possible
memory addresses. The data memory stores the data for the corresponding ways.
For each way, the tag memory contains the following information:
•
Valid bits saying if the entry contains valid data or is free. Both caches have a single valid bit for
each cache line.
•
The tag, all bits of the cached memory address that are not implied by the set
•
If MMU is enabled, the context ID of the cache entry
•
Check bits for detecting errors
When a read from cache is performed, the tags and data for all cache ways of the corresponding set
are read out in parallel, the tags and valid bits are compared to the desired address and the matching
way is selected. In the hit case, this is all done in the same cycle to support the full execution rate of
the processor.
In the miss case, the cache will at first deliver incorrect data. However on the following cycle, a hold
signal will be asserted to prevent the processor from proceeding with that data. After the miss has
been processed, the correct data is injected into the pipeline using a memory data strobe (mds) signal,
and afterwards the hold signal can be released. If the missed address is cacheable, then the data read in
from the cache miss will be stored into the cache, possibly replacing one of the existing ways.
GR740-UM-DS, Jun 2016, Version 1.4
53
www.cobham.com/gaisler
GR740
In the instruction streaming case, the processor pipeline is stepped one step for every received instruction. If the processor needs extra pipeline cycles to stretch a multi-cycle instruction or due to an interlock condition (see section 6.2), or if the processor jumps/branches away, then the instruction cache
will hold the pipe, fetch the remainder of the cache line, and the pipeline will then proceed normally.
6.3.3
Address mapping
The addresses seen by the CPU are divided into tag, index and offset bits. The index is used to select
the set in the cache, therefore only a limited number of cache lines with the same index part can be
stored at one time in the cache. The tag is stored in the cache and compared upon read.
4 KiB way, 32 bytes/line
31
12 11
Tag
5
4
Index
0
Offset
Figure 4. Cache address mapping
6.3.4
Data cache policy
The data cache employs a write-through policy, meaning that every store made on the CPU will propagate, via the write buffer, to the bus and there are no “dirty” lines in the cache that has not yet been
written out apart from what is in the buffer. The store will also update the cache if the address is present, however a new line will not be allocated in that case.
Table 39. LEON4 Data caching behavior
Operation
In cache
Cacheable
Bus action
Cache action
Load data
Data load
No
No
Read
No change
Bus
No
Yes
Read
Line allocated/replaced
Bus
Yes
-
None
No change
Cache
Data load with
forced cache
miss (ASI 1)
No
No
Read
No change
Bus
No
Yes
Read
Line allocated/replaced
Bus
Yes
-
Read
Data updated
Bus
Data load with
MMU bypass
(ASI 0x1C)
-
-
Read (phys addr)
No change
Bus
Data store
No
No
Write (via buffer)
No change
(N/A)
No
Yes
Write (via buffer)
No change
(N/A)
Yes
-
Write (via buffer)
Data updated
(N/A)
-
-
Write (via buffer,
phys addr)
No change
(N/A)
Data store with
MMU bypass
(ASI 0x1C)
6.3.5
Write buffer
The data cache contains a write buffer able to hold a single 8,16,32, or 64-bit write. For half-word or
byte stores, the stored data replicated into proper byte alignment for writing to a word-addressed
device. The write is processed in the background so the system can keep executing while the write is
being processed. However, any following instruction that requires bus access will block until the write
buffer has been emptied. Loads served from cache will however not block, due to the cache policy
GR740-UM-DS, Jun 2016, Version 1.4
54
www.cobham.com/gaisler
GR740
used there can not be a mismatch between cache data and store buffer (the effect of this behavior on
SMP systems is discussed in section 6.7).
Since the processor executes in parallel with the write buffer, a write error will not cause an exception
to the store instruction. Depending on memory and cache activity, the write cycle may not occur until
several clock cycles after the store instructions has completed. If a write error occurs, the currently
executing instruction will take trap 0x2b. This trap can be disabled using the DWT configuration (see
section 6.10.2).
Note: a 0x2b trap handler should flush the data cache, since a write hit would update the cache while
the memory would keep the old value due the write error
6.3.6
Operating with MMU
When MMU is enabled, the virtual addresses seen by the running code no longer correspond directly
to the physical addresses on the AHB bus. The cache uses tags based on the virtual addresses, as this
avoids having to do any additional work to translate the address in the most timing-critical hit case.
However, any time a bus access needs to be made, a translation request has to be sent to the MMU to
convert the virtual address to a physical address. For the write buffer, this work is included in the
background processing of the store. The translation request to the MMU may result in memory
accesses from the MMU to perform table walk, depending on the state of the MMU.
The MMU context ID is included in the cache tags in order to allow switching between multiple
MMU contexts mapping the same virtual address to different physical addresses. Note that the cache
does not detect aliases to the same physical address so in that case the same physical address may be
cached in multiple ways (also see snooping below).
6.3.7
Snooping
The data cache supports AHB bus snooping. The AHB bus the processor is connected to, is monitored
for writes from other masters to an address which is in the cache. If a write is done to a cached
address, that cache line is marked invalid and the processor will be forced to fetch the (new) data from
memory the next time it is read.
For using snooping together with the MMU, an extra tag memory storing physical tags is used to
allow comparing with the physical address on the AHB bus.
The processor can snoop on itself and invalidate any other cache lines aliased to the same physical
address in case there are multiple virtual mappings to the same physical address that is being written.
However, note that this does not happen until the write occurs on the bus so the other virtual aliases
will return the old data in the meantime.
6.3.8
Enabling and disabling cache
Both I and D caches are disabled after reset. They are enabled by writing to the cache control register
(see 6.10.5). Before enabling the caches after a reset they must be flushed to ensure that all tags are
marked invalid.
6.3.9
Cache freeze
Each cache can be in one of three modes: disabled, enabled and frozen. If disabled, no cache operation
is performed and load and store requests are passed directly to the memory controller. If enabled, the
cache operates as described above. In the frozen state, the cache is accessed and kept in sync with the
main memory as if it was enabled, but no new lines are allocated on read misses.
If the DF or IF bit is set, the corresponding cache will be frozen when an asynchronous interrupt is
taken. This can be beneficial in real-time system to allow a more accurate calculation of worst-case
execution time for a code segment. The execution of the interrupt handler will not evict any cache
lines and when control is returned to the interrupted task, the cache state is identical to what it was
GR740-UM-DS, Jun 2016, Version 1.4
55
www.cobham.com/gaisler
GR740
before the interrupt. If a cache has been frozen by an interrupt, it can only be enabled again by
enabling it in the CCR. This is typically done at the end of the interrupt handler before control is
returned to the interrupted task.
6.3.10 Flushing
Both instruction and data cache are flushed either by executing the FLUSH instruction, setting the FI/
FD bits in the cache control register, or by writing to certain ASI address spaces.
Cache flushing takes one clock cycle per cache set, during which the IU will not be halted, but during
which the caches are disabled. When the flush operation is completed, the cache will resume the state
(disabled, enabled or frozen) indicated in the cache control register. Diagnostic access to the cache is
not possible during a flush operation and will cause a data exception (trap=0x09) if attempted.
Note that while the SPARC V8 specifies only that the instructions pointed to by the FLUSH argument
will be flushed, the LEON4 will additionally flush the entire I and D cache (which is permitted by the
standard as the additional flushing only affects performance and not operation). While the LEON4
currently ignores the address argument, it is recommended for future compatibility to only use the
basic flush %g0 form if you want the full flush behavior.
6.3.11 Locking
Cache line locking is not supported by LEON4.
6.3.12 Diagnostic access
The cache tag and data contents can be directly accessed for diagnostics and for locking purposes via
various ASI:s, see section 6.9.5.
6.3.13 Local scratch pad RAM
Local scratch pad RAM is not supported by LEON4.
6.3.14 Fault tolerance support
The cache memories (tags and data) are protected agains soft errors using byte-parity codes. On a
detected parity error, the corresponding cache (I or D) will be flushed and the data will be refetched
from external memory. This is done transparently to software execution.
6.4
Memory management unit
6.4.1
Overview
The memory-management unit is compatible with the SPARC V8 reference MMU (SRMMU) architecture described inthe SPARC V8 manual, appendix H.
The MMU provides address translation of both instructions and data via page tables stored in memory.When needed, the MMU will automatically access the page tables to calculate the correct physical
address. The latest translations are stored in a special cache called the translation lookaside buffer
(TLB), also referred to as Page Descriptor Cache (PDC) in the SRMMU specification. The MMU also
provides access control, making it possible to “sandbox” unpriviledged code from accessing the rest
of the system.
6.4.2
MMU/Cache operation
When the MMU is disabled, the MMU is bypassed and the caches operate with physical address mapping. When the MMU is enabled, the caches tags store the virtual address and also include an 8-bit
context field. Both the tag address and context field must match to generate a cache hit. If cache
GR740-UM-DS, Jun 2016, Version 1.4
56
www.cobham.com/gaisler
GR740
snooping is used, physical tags must be enabled for it to work when address translation is used, see
section 6.3.7.
Because the cache is virtually tagged, no extra clock cycles are needed in case of a cache load or
instruction cache hit. In case of miss or write buffer processing, a translation is required which might
add extra latency to the processing time, depending on if there is a TLB miss. TLB lookup is done at
the same time as tag lookup and therefore add no extra clock cycles.
If there is a TLB miss the page table must be traversed, resulting in up to four AMBA read accesses
and one possible writeback operation. See the SRMMU specification for the exact format of the page
table.
An MMU page fault will generate trap 0x09 for the D-cache and trap 0x01 for the I cache, and update
the MMU status registers according to table 40 and the SRMMU specification. In case of multiple
errors, they fault type values are prioritized as the SRMMU specification requires. The cache and
memory will not be modified on an MMU page fault.
Table 40. LEON4 MMU Fault Status Register, fault type values
Fault type
SPARC V8 ref
Priority
Condition
6
Internal error
1
Never issued by LEON SRMMU
4
Translation error
2
AHB error response while performing table walk. Translations errors as defined in SPARC V8 manual. A translation
error caused by an AMBA ERROR response will overwrite all other errors. Other translation errors do no overwrite existing translation errors when FAV = 1.
1
Invalid address error
3
Page table entry for address was marked invalid
3
Prilege violation error
4
2
Protection error
5
Access denied based on page table and su status (see
SRMMU spec for how privilege and protection error are
prioritized)
0
None
-
6.4.3
No error (inside trap this means the trap occurred when
fetching the actual data)
Translation look-aside buffer (TLB)
The MMU has separate TLBs for instructions and data. The number of TLB entries (for each implemented TLB) is 16. The organisation of the TLB and number of entries is not visible to the software
and does thus not require any modification to the operating system. The TLB can be flushed using an
STA instruction to ASI 0x18, see section 6.9.6.
6.5
Floating-point unit
The high-performance GRFPU operates on single- and double-precision operands, and implements all
SPARC V8 FPU operations including square root and division. The FPU is interfaced to the LEON4
pipeline using a LEON4-specific FPU controller (GRFPC) that allows FPU instructions to be executed simultaneously with integer instructions. Only in case of a data or resource dependency is the
integer pipeline held. The GRFPU is fully pipelined and allows the start of one instruction each clock
cycle, with the exception is FDIV and FSQRT which can only be executed one at a time. The FDIV
and FSQRT are however executed in a separate divide unit and do not block the FPU from performing
all other operations in parallel.
All instructions except FDIV and FSQRT has a latency of three cycles, but to improve timing, the
LEON4 FPU controller inserts an extra pipeline stage in the result forwarding path. This results in a
GR740-UM-DS, Jun 2016, Version 1.4
57
www.cobham.com/gaisler
GR740
latency of four clock cycles at instruction level. The table below shows the GRFPU instruction timing
when used together with GRFPC:
Table 41. GRFPU instruction timing with GRFPC
Instruction
Throughput
Latency
FADDS, FADDD, FSUBS, FSUBD,FMULS, FMULD, FSMULD, FITOS, FITOD,
FSTOI, FDTOI, FSTOD, FDTOS, FCMPS, FCMPD, FCMPES. FCMPED
1
4
FDIVS
14
16
FDIVD
15
17
FSQRTS
22
24
FSQRTD
23
25
The GRFPC controller implements the SPARC deferred trap model, and the FPU trap queue (FQ) can
contain up to 7 queued instructions when an FPU exception is taken. The version field in %fsr has the
value of 2 to signal that the processor is implemented with the GRFPU.
The GRFPU does not handle denormalized numbers as inputs and will in that case cause an fp_exception with the FPU trap type set to unfinised_FPOP (tt=2). There is a non-standard mode in the FPU
that will instead replace the denormalized inputs with zero and thus never create this condition.
6.6
Co-processor interface
The coprocessor interface is unused and disabled in this device.
6.7
AMBA interface
6.7.1
Overview
The LEON4 processor has one AHB master interface. The types of AMBA accesses supported and
performed by the processor depend on the accessed memory area’s cachability, if the corresponding
cache is enabled, and if the accessed memory area has been marked as being on the wide bus.
Cacheable instructions are fetched with a burst of two 128-bit accesses.
The HPROT signals of the AHB bus are driven to indicate if the accesses is instruction or data, and if
it is a user or supervisor access.
Table 42. HPROT values
Type of access
User/Super
HPROT
Instruction
User
1100
Instruction
Super
1110
Data
User
1101
Data
Super
1111
MMU
Any
1101
In case of atomic accesses, a locked access will be made on the AMBA bus to guarantee atomicity as
seen from other masters on the bus.
6.7.2
Cachability
The processor treats the memory areas 0x00000000 - 0x7FFFFFFF and 0xC0000000 - 0xCFFFFFFF
as cacheable. The test of the physical address space is treated as uncached.
GR740-UM-DS, Jun 2016, Version 1.4
58
www.cobham.com/gaisler
GR740
6.7.3
AMBA access size
Cacheable data is fetched in a burst 128-bit accesses Data access to uncacheable areas may only be
done with 8-, 16- and 32-bit accesses, i.e. the LDD and STD instructions may not be used. If an area is
marked as cacheable then the data cache will automatically try to use 128-bit accesses. This means
that if 128-bit accesses are unwanted and a memory area is mapped as cacheable then software should
only perform data accesses with cache bypass (ASI 0x1C) and no 64-bit loads (LDD) when accessing
the slave. One example of how to use forced cache miss for loads is given by the following function:
static inline int load(int addr)
{
int tmp;
asm volatile(" lda [%1]0x1c, %0 "
: "=r"(tmp)
: "r"(addr)
);
return tmp;
}
In the GR740 device, this may primarily be of interest when accessing the PROM area (base address
at 0xC0000000) and possibly also for using the processor to test word and sub-word accesses to the
Level-2 cache and memory controller (memory area 0x00000000 - 0x7FFFFFFF).
The processor only supports using wide accesses to memory areas that are marked as cached. This
means that LDD shall not be used for peripheral register areas.
Store instructions result in a AMBA access with size corresponding to the executed instruction, 64-bit
store instructions (STD) are always translated to 64-bit accesses (never converted into two 32-bit
stores as is done for LEON3). The table below indicates the access types used for instruction and data
accesses depending on cachability and cache configuration.

Processor
operation
Area not
Area is cacheable1
cacheable1
Cache enabled2
Instruction
fetch
Burst of 32-bit read accesses
Burst of 128-bit accesses
Data load <=
32-bit
Read access with size specified by load
instruction
Burst of 128-bit accesses
Data load 64bit (LDD)
Illegal3
Single 64-bit access will be performed
Burst of 128-bit accesses
Data store <=
32-bit
Store access with size specified by store instruction.
Data store 64bit (STD)
Illegal (64-bit store to 32-bit area)
64-bit store access will be performed.
Single accesses can be performed via ASI 0x1C.
Cache disabled
Read access with size specified by load instruction
Single 64-bit read access
64-bit store access
1
Cached memory regions are 0x00000000 - 0x7FFFFFFF and 0xC0000000 - 0xCFFFFFFF.
Bus accesses for reads will only be made on L1 cache miss or on load with forced cache miss.
3 Data accesses to uncached areas may only be done with 8-, 16- and 32-bit accesses.
2
GR740-UM-DS, Jun 2016, Version 1.4
59
www.cobham.com/gaisler
GR740
6.7.4
Error handling
An AHB ERROR response received while fetching instructions will normally case an instruction
access exception (tt=0x1). However if this occurs during streaming on an address that is not needed,
the I cache controller will just not set the corresponding valid bit in the cache tag. If the IU later
fetches an instruction from the failed address, a cache miss will occur, triggering a new access to the
failed address.
An AHB ERROR response while fetching data into the data cache will normally trigger a data_access_exception trap (tt=0x9). If the error was for a part of the cache line other than what was currently
being requested by the pipeline, a trap is not generated and the valid bit for that line is not set.
An ERROR response during an MMU table walk will lead the MMU to set the fault type to Internal
error (1) and generate an instruction or data access exception, depending on which type of access that
caused the table walk.
6.8
Multi-processor system support
This section gives an overview of issues when using the LEON4 in multi-processor configuration.
6.8.1
Start-up
Only the first processor will start after reset, assuming that the BREAK bootstrap signal is low, and all
other processors will remain halted in power-down mode. After the system has been initialized, the
remaining processors can be started by writing to the ‘multiprocessor status register’, located in the
multiprocessor interrupt controller. The halted processors start executing from the reset address (see
section 6.2.18).
An application in a multiprocessor system can determine which processor it is executing on by checking the processor index field in the LEON4 configuration register (%asr17). As all processors typically have the same reset start address value, boot software must check the processor index and
perform processor specific setup (e.g. initialization of stack pointer) based on the value of the processor index.
It is only possible for a processor to wake other processors up via the ‘multiprocessor status register’.
Once a processor is running it cannot be reset via the interrupt controller. If software detects that one
processor is unresponsive and needs to restart the processor then the full system should be reset, for
example by triggering the system’s watchdog. In order for software to monitor that all processors in a
system are up and running it is recommended to implement a heartbeat mechanism in software.
6.8.2
Shared memory model
Each processor core has it’s own separate AHB master interface and the AHB controller will arbitrate
between them to share access to the on-chip bus.
If caches are not used, the processors will form a sequentially consistent (SC) system, where every
processor will execute it’s loads, stores and atomics to memory in program order on the AHB bus and
the different processors operations will be interleaved in some order through the AHB arbitration. The
shared memory controller AHB slave is assumed to not reorder accesses so a read always returns the
latest written value to that location on the bus.
When using caches with snooping (and with physical tags if using the MMU), the shared memory will
act according to the slightly weaker SPARC Total Store Order (TSO) model. The TSO model is close
to SC, except that loads may be reordered before stores coming from the same CPU. The stores and
atomics are conceptually placed in a FIFO (see the diagrams in the SPARC standard) and the loads are
allowed to bypass the FIFO if they are not to the same address as the stores. Loaded data from other
addresses may therefore be either older or newer, with respect to the global memory order, than the
stores that have been performed by the same CPU.
GR740-UM-DS, Jun 2016, Version 1.4
60
www.cobham.com/gaisler
GR740
In the LEON4 case this happens because cache hits are served without blocking even when there is
data in the write buffer. The loaded data will always return the stored data in case of reading the same
address, because if it is cached, the store updates the cache before being put in the write buffer, and if
it was not in cache then the load will result in a miss which waits for the write buffer to complete.
Loaded data from a different address can be older than the store if it is served by cache before the
write has completed, or newer if it results in a cache miss or if there is a long enough delay for the
store to propagate to memory before reading.
See relevant literature on shared memory systems for more information. These details are mainly of
concern for complex applications using lock-free data structures such as the Linux kernel, the recommendation for applications is to instead avoid concurrent access to shared structures by using
mutexes/semaphores based on atomic instructions, or to use message passing schemes with one-directional circular buffers.
6.8.3
Memory-mapped hardware
Hardware resource (peripheral registers) are memory mapped on uncacheable address spaces. They
will be accessible from all the CPU:s in a sequentially consistent manner. Since software drivers usually expect to be “alone” accessing the peripheral and the peripheral’s register interfaces are not
designed for concurrent use by multiple masters, using a bare-C application designed for single-processor usage on multiple cores at the same time will generally not work. This can be solved by partitioning the applications so that each peripheral is only accessed by one of the CPU:s. This partitioning
also need to be done between the interrupts so the peripheral’s interrupts will be received by the correct processor.
6.9
ASI assignments
6.9.1
Summary
The table shows the ASI usage for LEON.
Table 43. ASI usage
ASI
Usage
0x01
Forced cache miss.
0x02
System control registers (cache control register)
0x08, 0x09, 0x0A, 0x0B
Normal cached access (replace if cacheable)
0x0C
Instruction cache tags
0x0D
Instruction cache data
0x0E
Data cache tags
0x0F
Data cache data
0x10
Flush instruction cache (and also data cache when system is implemented with MMU)
0x11
Flush data cache
0x13
MMU only: Flush instruction and data cache
0x14
MMU only: MMU diagnostic D context cache access (deprecated, do not use in new SWapplications)
0x15
MMU only: MMU diagnostic I cache context access (deprecated, do not use in new SW applications)
0x18
MMU only: Flush TLB and I/D cache
0x19
MMU only: MMU registers
0x1C
MMU only: MMU and cache bypass
0x1D
MMU only: MMU diagnostic access (deprecated, do not use in new SW applications)
0x1E
MMU only: MMU snoop tags diagnostic access
GR740-UM-DS, Jun 2016, Version 1.4
61
www.cobham.com/gaisler
GR740
The processor implements SPARC V8E nonprivileged ASI access, accesses to ASI 0x80 - 0xFF do
not require supervisor privileges. No registers are mapped at ASI 0x80 - 0xFF and the instructions
used to access these areas can be used as trace points for software tracing. Trace filtering (see section
13.4) allows filtering of these instructions.
6.9.2
ASI 0x1, Forced cache miss
ASI 1 is used for systems without cache coherency, to load data that may have changed in the background, for example by DMA units. It can also be used for other reasons, for example diagnostic purposes, to force a AHB load from memory regardless of cache state.
The address mapping of this ASI is matched with the regular address space, and if MMU is enabled
then the address will be translated normally. Stores to this ASI will perform the same way as ordinary
data stores.
For situations where you want to guarantee that the cache is not modified by the access, the MMU and
cache bypass ASI, 0x1C, can be used instead. However this is only available when MMU is implemented.
GR740-UM-DS, Jun 2016, Version 1.4
62
www.cobham.com/gaisler
GR740
6.9.3
ASI 0x2, System control registers
ASI 2 contains a few control registers that have not been assigned as ancillary state registers. These
should only be read and written using 32-bit LDA/STA instructions.
All cache registers are accessed through load/store operations to the alternate address space (LDA/
STA), using ASI = 2. The table below shows the register addresses:
Table 44. ASI 2 (system registers) address map
Address
Register
0x00
Cache control register
0x04
Reserved
0x08
Instruction cache configuration register
0x0C
Data cache configuration register
6.9.4
ASI 0x8-0xB, Data/Instruction
These ASIs are assigned by the SPARC standard for normal data and instruction fetches.
Accessing the instruction ASIs explicitly via LDA/STA instructions is not supported in the LEON4
implementation. Using LDA/STA with the user/supervisor data ASI will behave as the affect the
HPROT signal emitted by the processor according to section 6.7.1, but MMU access control will still
be done according to the super-user state of the %psr register.
6.9.5
ASI 0xC-0xF, ICache tags/data, DCache tags/data
ASI 0xC-0xF provide diagnostic access to the instruction cache memories. These ASIs should only be
accessed by 32-bit LDA/STA instructions. These ASIs can not be used while a cache flush is in progress.
The same address bits used normally as index are used to index the cache also in the diagnostic
access. For a multi-way cache, the lowest bits above the index part, the lowest bits that would normally be used as tag, are used to select which way to read/write. The remaining address bits are don’t
cares, leading the address map to wrap around.
The tag parity and context bits can also be read out through these ASIs by setting the PS bit in the
cache configuration register. When this bit is set, the parity data is read instead of the ordinary data.
When writing the tag bits, the context bits will always be written with the current context in the MMU
control register. The parity to be written is calculated based on the supply write-value and the context
ID in the MMU control register. The parity bits can be modified via the TB field in the cache control
register.
GR740-UM-DS, Jun 2016, Version 1.4
63
www.cobham.com/gaisler
GR740
Example for 4 KiB way, 32 bytes/line, 4 ways
Tag diagnostic ASIs (ASI 0xC,E):
31
15 14
Addr:
12 11
Way
(don’t care)
31
10
Data:
ATAG
31
Parity:
23
Reserved
5
4
Index
9
8
0
0
(don’t care)
7
0
VALID
16 15
4
CTXID
0
3
Reserved
0
TPAR
Data diagnostic ASIs (ASI 0xD,F):
31
15 14
Addr:
(don’t care)
12 11
Way
5
4
Index
0
Offset
31
0
Data:
Cached data word
4
31
Parity:
Reserved
3
0
DPAR
Figure 5. ASI 0xC-0xF address mapping and data layout
Field Definitions:
•
Address Tag (ATAG) - Contains the tag address of the cache line.
•
Valid (V) - When set, the cache line contains valid data. The LEON4 caches only have one valid
bit per cache line which is replicated for the whole 8-bit diagnostic field to keep software backward compatibility.
•
CTXID - Context ID, used when MMU is enabled
•
TPAR - Byte-wise parity of tag bits, context ID parity is XOR:ed into bit 3.
•
DPAR - Byte-wise parity of data bits
6.9.6
ASI 0x10, 0x11, 0x13, 0x18 - Flush
For historical reasons there are multiple ASIs that flush the cache in different ways.
Writing to ASI 0x10 will flush the entire instruction cache. If MMU is implemented in the core, both
instruction and data cache will be flushed.
Writing to ASI 0x11 will flush the data cache only.
Writing to ASI 0x13 will flush the instruction cache and data cache. Only available when MMU is
implemented.
Writing to ASI 0x18, which is available only if MMU is implemented, will flush both the MMU TLB,
the I-cache, and the D-cache. This will block execution for a few cycles while the TLB is flushed and
then continue asynchronously with the cache flushes continuing in the background.
GR740-UM-DS, Jun 2016, Version 1.4
64
www.cobham.com/gaisler
GR740
6.9.7
ASI 0x19 - MMU registers
This ASI provides access to the MMU:s control and status registers. The following MMU registers
are implemented:
Table 45. MMU registers (ASI = 0x19)
Address
Register
0x000
MMU control register
0x100
Context pointer register
0x200
Context register
0x300
Fault status register
0x400
Fault address register
6.9.8
ASI 0x1C - MMU and cache bypass
Performing an access via ASI 0x1C will act as if MMU and cache were disabled. The address will not
be translated and the cache will not be used or updated by the access.
6.9.9
ASI 0x1E - MMU snoop tags diagnostic access
If the MMU has been configured to use separate snoop tags, they can be accessed via ASI 0x1E. This
is primarily useful for RAM testing, and should not be performed during normal operation. This ASI
is addressed the same way as the regular diagnostic ASI:s 0xC, 0xE, and the read/written data has the
layout as shown below:
31
12
ATAG
11
2
“0000”
1
PAR
0
IV
Figure 6. Snoop cache tag layout
[31:10] Address tag. The physical address tag of the cache line.
[1]:
Parity. The odd parity over the data tag. Only used when processor is implemented with fault-tolerance features.
[0]:
Invalid. When set, the cache line is not valid and will cause a cache miss if accessed by the processor. Only present
if fast snooping is enabled.
GR740-UM-DS, Jun 2016, Version 1.4
65
www.cobham.com/gaisler
GR740
6.10
Configuration registers
6.10.1 PSR, WIM, TBR registers
The %psr, %wim, %tbr registers are implemented as required by the SPARC V8 standard.
Table 46. %psr- Processor state register
31
28 27
24 23
20 19
14 13 12 11
EC EF
8
7
PIL
S
6
5
IMPL
VER
ICC
RESERVED
0b1111
0b0011
0
0b000000
0
0
0x0
1
r
r
r
r
r
rw
rw
rw rw rw
4
0
PS ET
1
CWP
0
0b00000
rw
31: 28
Implementation ID (IMPL), read-only hardwired to “1111” (15)
27: 24
Implementation version (VER), read-only hardwired to “0011” (3) for LEON3/LEON4.
23: 20
Integer condition codes (ICC), see sparcv8 for details
19: 14
Reserved
13
Enable coprocessor (EC) - read-only
12
Enable floating-point (EF)
11 8
Processor interrupt level (PIL) - controls the lowest IRQ number that can generate a trap
7
Supervisor (S)
6
Previous supervisor (PS), see SPARC V8 manual for details
5:
Enable traps (ET)
4: 0
Current window pointer (CWP)
Table 47. %wim - Window Invalid Mask
31
8
7
0
RESERVED
WIM
0
NR
r
rw
31: 8
RESERVED
7: 0
Window Invalid Mask (WIM)
Table 48. %tbr - Trap Base Register
31
12 11
4
3
0
TBA
TT
R
Taken from interrupt controller. Default is 0xC0000
0
0
rw
r
r
31: 12
Trap base address (TBA) - Top 20 bits used for trap table address
11: 4
Trap type (TT) - Last taken trap type.
3: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
66
www.cobham.com/gaisler
GR740
6.10.2 ASR17, LEON4 configuration register
The ancillary state register 17 (%asr17) provides information on how various configuration options
were set during synthesis. This can be used to enhance the performance of software, or to support enumeration in multi-processor systems. There are also a few bits that are writable to configure certain
aspects of the processor.
Table 49. %asr17 - LEON4 configuration register
31
28 27 26 25 24 23
INDEX
0-3
r
31: 28
D
B
P
R1
0
0
D R2
B
P
M
18 17 16 15 14 13 12 11 10
CS
CF
0
0
0b00
rw rw rw rw
r
r
1
RESERVED
DW SV LD
0
8
7
6
5
4
0
FPU
M V8
NWP
NWIN
0
0
0b01
0
1
0b100
0x7
rw rw
r
r
r
r
r
r
Processor index (INDEX) - Each LEON core gets a unique index to support enumeration. The processors are numbered 0 - 3.
27
Disable Branch Prediction (DBP) - Disables branch prediction when set to ‘1’.
26
Reserved field (R1) - The setting of this field affects timing of SWAPand LDST instructions. It must
always be set to ’0’.
25
Disable Branch Prediction on instruction cache misses (DBPM) - When set to ‘1’ this avoids instruction cache fetches (and possible MMU table walk) for predicted instructions that may be annullated.
24
Reserved field (R2) - The setting of this field affects lock generation used for atomic bus accesses
(SWAP, LDST, CASA). This field must always be set to ’0’.
23: 18
Reserved for future implementations
17
Clock switching (CS) - This field is 0 to signify that this implementation does not support clock
switching.
16: 15
CPU clock frequency (CF) - This field is 0 to signify that the CPU runs at the same frequency as the
AMBA bus.
14
Disable write error trap (DWT) - When set, a write error trap (tt = 0x2b) will be ignored. Set to zero
after reset.
13
Single-vector trapping (SVT) enable - If set, will enable single-vector trapping.
12
Load delay (LD) - 0 to signify that a 1-cycle load delay i s used.
11: 10
FPU option (FPU) - "01” = GRFPU.
9
Multiply and accumulate (M) - 0 to signify that (MAC) instructions are unsupported.
8
SPARC V8 (V8) - Set to 1, to signify that the SPARC V8 multiply and divide instructions are available.
7: 5
Number of implemented watchpoints (NWP) - Value is 4.
4: 0
Number of register windows (NWIN) - Number of implemented registers windows corresponds to
NWIN+1. Field has value 7.
GR740-UM-DS, Jun 2016, Version 1.4
67
www.cobham.com/gaisler
GR740
6.10.3 ASR22-23 - Up-counter
The ancillary state registers 22 and 23 (%asr22-23) contain a internal up-counter that can be read by
software without causing any access on the on-chip bus. The number of available bits in the counter is
56 and corresponds to the DSU time tag counter. %ASR23 contains the least significant part of the
counter value and %ASR22 contains the most significant part. In case the implementation does not
contain a debug support unit connected to the processor then the up-counter is not available (value is
always zero).
The time tag value accessible in these registers is the same time tag value used for the system’s trace
buffers (if implemented) and for all processors connected to the same debug support unit. The time
tag counter will increment when any of the trace buffers is enabled, or when the time tag counter is
forced to be enabled via the DSU register interface, or when any processor has its %ASR22 Disable
Up-counter (DUCNT) field set to zero.
The up-counter value will increment even if all processors have entered power-down mode.
Table 50. %asr22 - LEON4 Up-counter MSbs
31 30
24 23
D
U
C
N
T
RESERVED
1
0
0
UPCNT(55:32)
*
rw
r
31
Disable Up-counter (DUCNT) - Disable upcounter. When set to ‘1’ the up-counter may be disabled.
The value for the up-counter in each processor is taken from a shared timer. The shared counter is
stopped when all processors have DUCNT set to one and the time tag counter in the DSU is disabled.
When cleared, the counter will increment each processor clock cycle. Default (reset) value is ‘1’.
30: 0
Counter value (UPCNT(62:32)) - Most significant bits of internal up-counter. Counter is reset to 0 at
reset but may start counting due to conditions described for the DUCNT field.
Table 51. %asr23 - LEON4 Up-counter LSbs
31
0
UPCNT(31:0)
*
r
31: 0
Counter value (UPCNT(31:0)) - Least significant bits of internal up-counter. Counter is reset to 0 at
reset but may start counting due to conditions described for the DUCNT field in %asr22.
GR740-UM-DS, Jun 2016, Version 1.4
68
www.cobham.com/gaisler
GR740
6.10.4 ASR24-31, Hardware watchpoint/breakpoint registers
Each breakpoint consists of a pair of ancillary state registers (%asr24/25, %asr26/27, %asr28/29 and
%asr30/31) registers; one with the break address and one with a mask:
Table 52. %asr24, %asr26, %asr28, %asr30 - Watchpoint address register(s)
31
1
0
WADDR[31:2]
2
R
IF
NR
0
0
rw
r
rw
31: 2
Watchpoint address (WADDR) - Address to compare against
1
RESERVED
0
Break on instruction fetch (IF) - Break on instruction fetch from the specified address/mask combination
Table 53. %asr25, %asr27, %asr29, %asr31 - Watchpoint mask register(s)
31
2
WMASKR[31:2]
31: 2
1
0
DL DS
NR
0
rw
rw rw
0
Watchpoint mask (WMASK) - Bit mask controlling which bits to check (1) or ignore (0) for match
1
Break on data load (DL) - Break on data load from the specified address/mask combination
0
Break on data store (DS) - Break on data store to the specified address/mask comination
Note: Setting IF=DL=DS=0 disables the breakpoint
When there is a hardware watchpoint match and DL or DS is set then trap 0x0B will be generated.
Hardware watchpoints can be used with or without the LEON4 debug support unit (DSU) enabled.
GR740-UM-DS, Jun 2016, Version 1.4
69
www.cobham.com/gaisler
GR740
6.10.5 Cache control register
The cache control register located at ASI 0x2, offset 0, contains control and status registers for the I
and D cache.
Table 54. ASI 0x2, 0x00 - CCR - Cache control register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
9
R
DTE
S
T
E
R PS
TB
0 NR
0
0
0x0
0
r
r
rw
rw
rw rw rw
rw
DS FD FI
31
Reserved
0
0
ITE
IDE
8
7
6
DDE
5
4
DF IF
3
2
DCS
1
0
FT
R
ST
R
IP DP
ICS
0b01
0
1
0
0
0
0
0
0
0
0
0
0
0
r
r
r
r
r
r
rw
rw
rw
rw
rw rw
rw
rw
30
Snoop Tag Flag (STE) - Set when parity error is detected in the data physical (snoop) tags.
29
Reserved
28
Parity Select (PS) - if set diagnostic read will return 4 check bits in the lsb bits, otherwise tag or data
word is returned.
27: 24
Test Bits (TB) - if set, check bits will be xored with test bits TB during diagnostic write.
23
Data cache snoop enable (DS) - if set, will enable data cache snooping.
22
Flush data cache (FD). If set, will flush the instruction cache. Always reads as zero.
21
Flush Instruction cache (FI). If set, will flush the instruction cache. Always reads as zero.
20: 19
FT scheme (FT) - “01” = 4-bit checking implemented
18
Reserved for future implementations
17
Separate snoop tags (ST). Has value 1.
16
Reserved
15
Instruction cache flush pending (IP). This bit is set when an instruction cache flush operation is in
progress
14
Data cache flush pending (DP). This bit is set when an data cache flush operation is in progress.
13: 12
Instruction Tag Errors (ITE) - Number of detected parity errors in the instruction tag cache.
11: 10
Instruction Data Errors (IDE) - Number of detected parity errors in the instruction data cache.
9: 8
Data Tag Errors (DTE) - Number of detected parity errors in the data tag cache.
7:6
Data Data Errors (DDE) - Number of detected parity errors in the data data cache.
5
Data Cache Freeze on Interrupt (DF) - If set, the data cache will automatically be frozen when an
asynchronous interrupt is taken.
4
Instruction Cache Freeze on Interrupt (IF) - If set, the instruction cache will automatically be frozen
when an asynchronous interrupt is taken.
3:2
Data Cache state (DCS) - Indicates the current data cache state according to the following: X0= disabled, 01 = frozen, 11 = enabled.
1:0
Instruction Cache state (ICS) - Indicates the current data cache state according to the following: X0=
disabled, 01 = frozen, 11 = enabled.
GR740-UM-DS, Jun 2016, Version 1.4
70
www.cobham.com/gaisler
GR740
6.10.6 I-cache and D-cache configuration registers
The configuration of the two caches if defined in two registers: the instruction and data configuration
registers. These registers are read-only, except for the REPL field that can be written, and indicate the
size and configuration of the caches. They are located under ASI 2 at offset 8 and 12.
Table 55. ASI 0x2, 0x08 and 0x09C - CCFG - Cache configration registers
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
CL R
0
0
r
r
REPL
rw
9
8
7
6
5
4
3
2
1
SN
WAYS
WSIZE
LR
LSIZE
RESERVED
1
0b011
0x2
0
0b100
0x000
1
0b000
r
r
r
r
r
r
r
r
0
M RESERVED
31
Cache locking (CL) - Set if cache locking is implemented (always zeo).
30
RESERVED
29: 28
Cache replacement policy (REPL) - 00 - no replacement policy (direct-mapped cache), 01 - least
recently used (LRU), 10 - least recently replaced (LRR), 11 - random.
This field is writable, default (reset) value is "01" = LRU.
27
Cache snooping (SN) - Value 1 to signify that snooping is supported.
26: 24
Cache associativity (WAYS) - "011" - 4-way associative.
23: 20
Way size (WSIZE) - Indicates the size (KiB) of each cache way. This field has value 2. Size =
2SIZE=4 KiB way size.
19
Local ram (LR) - 0 in this implementation to signify that local RAMis not implemented.
18: 16
Line size (LSIZE) - Indicated the size (words) of each cache line. Set to 2. Line size = 2LSIZE= 4
words = 32 bytes.
15: 4
RESERVED
3
MMU present (M) - This bit is set to ‘1’ to signify that an MMU is present.
2: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
71
www.cobham.com/gaisler
GR740
6.10.7 MMU control register
The MMU control register is located in ASI 0x19 offset 0.
Table 56. ASI 0x19, 0x00- MMUCTRL - MMU control register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
IMPL
VER
ITLB
DTLB
R
TD ST
0x0
0x0
0b100
0b100
0
NR
1
r
r
r
r
r
rw
r
9
RESERVED
8
7
6
5
4
3
1
0
NF
E
0
0
0
r
rw rw
P
S
O
RESERVED
0
0
r
rw
2
31: 28
MMU Implementation ID (IMPL) - Hardcoded to “0000”
27: 24
MMU Version ID (VER) - Hardcoded to “0000”.
23: 21
Number of ITLB entries (ITLB) - The number of ITLB entries is calculated as 2ITLB = 24 = 16.
20: 18
Number of DTLB entries (DTLB) - The number of DTLB entries is calculated as 2DTLB = 24 = 16.
17: 16
RESERVED
15
TLB disable (TD) - When set to 1, the TLB will be disabled and each data access will generate an
MMU page table walk.
14
Separate TLB (ST) - This bit is set to 1 to signify that separate instruction and data TLBs are implemented
13: 8
RESERVED
7
Partial Store Ordering (PSO) - This field is writable but does not have an effect on processor operation.
6: 2
RESERVED
1
No Fault (NF) - When NF= 0, any fault detected by the MMU causes FSR and FAR to be updated
and causes a fault to be generated to the processor. When NF= 1, a fault on an access to ASI 9 is handled as when NF= 0; a fault on an access to any other ASI causes FSR and FAR to be updated but no
fault is generated to the processor.
0
Enable MMU (E) - 0 = MMU disabled, 1 = MMU enabled.
GR740-UM-DS, Jun 2016, Version 1.4
72
www.cobham.com/gaisler
GR740
6.10.8 MMU context pointer and context registers
The MMU context pointer register is located in ASI 0x19 offset 0x100 and the MMU context register
is located in ASI 0x19 offset 0x200. They together determine the location of the root page table
descriptor for the current context. Their definition follow the SRMMU specification in the SPARC V8
manual with layouts shown below.
Table 57. ASI 0x19, offset 0x100 - MMUCTXP - MMU context pointer register
31
2
CONTEXT TABLE POINTER
1
0
R
NR
0
rw
rr
31: 2
Context Table Pointer (CONTEXT TABLE POINTER) - Context table pointer, physical address bits
35:6 (note address is shifted 4 bits)
1: 0
Reserved
Table 58. ASI 0x19, offset 0x200 - MMUCTX - MMU context register
31
8
7
0
RESERVED
CONTEXT
0
0x00
r
rw
31: 8
RESERVED
7: 0
Current contect ID (CONTEXT)
In the LEON4, the context bits are OR:ed with the lower MMU context pointer bits when calculating
the address, so one can use less context bits to reduce the size/alignment requirements for the context
table.
GR740-UM-DS, Jun 2016, Version 1.4
73
www.cobham.com/gaisler
GR740
6.10.9 MMU fault status register
The MMU fault status register is located in ASI 0x19 offset 0x300, and the definition is based on the
SRMMU specification in the SPARC V8 manual. The SPARC V8 specifies that the fault status register should be cleared on read, on the LEON4 only the FAV bit is cleared on read. The FAV bit is
always set on error in the LEON4 implementation, so it can be used as a valid bit for the other fields.
Table 59. ASI 0x19, offset 0x300 - FSR - MMU Fault Status Register
31
1
0
RESERVED
18 17
EBE
10
9
L
8
7
AT
FT
F
A
V
O
W
0
0
0
0
0
0
0
r
r
r
r
r
r
r
31: 18
RESERVED
17: 10
External bus error (EBE) - Never set on the LEON4
9: 8
Level (L) - Level of page table entry causing the fault
7: 5
Access type (AT) - See V8 standard
4: 2
Fault type (FT) - See table 40.
1
Fault address valid (FAV) - Cleared on read, always written to 1 on fault
0
Overwrite (W) - Multiple faults of the same priority encountered
5
4
2
6.10.10 MMU fault address register
The MMU fault address register is located in ASI 0x19 offset 0x400, and the definition follows the
SRMMU specification in the SPARC V8 manual..
Table 60. ASI 0x19, offset 0x400 - FAR - MMU Fault Address Register
31
12 11
0
RESERVED
6.11
NR
0
r
r
31: 12
Fault Address (FAULT ADDRESS) - Top bits of virtual address causing translation fault
11: 0
RESERVED
Software considerations
6.11.1 Register file initialization on power up
It is recommended that the boot code for the processor writes all registers in the IU and FPU register
files before launching the main application. This allows software to be portable to both FT and nonFT versions of the LEON3 and LEON4 processors.
GR740-UM-DS, Jun 2016, Version 1.4
74
www.cobham.com/gaisler
GR740
6.11.2 Start-up
After reset, the caches are disabled and the cache control register (CCR) is 0. Before the caches may
be enabled, a flush operation must be performed to initialized (clear) the tags and valid bits. A suitable
assembly sequence could be:
flush
set 0x81000f, %g1
sta %g1, [%g0] 2
6.11.3 Data scrubbing
There is generally no need to perform data scrubbing on either IU/FPU register files or the cache
memory. During normal operation, the active part of the IU/FPU register files will be flushed to memory on each task switch. This will cause all registers to be checked and corrected if necessary. Since
most real-time operating systems performs several task switches per second, the data in the register
files will be frequently refreshed.
The similar situation arises for the cache memory. In most applications, the cache memory is significantly smaller than the full application image, and the cache contents is gradually replaced as part of
normal operation. For very small programs, the only risk of error build-up is if a part of the application is resident in the cache but not executed for a long period of time. In such cases, executing a
cache flush instruction periodically (e.g. once per minute) is sufficient to refresh the cache contents.
GR740-UM-DS, Jun 2016, Version 1.4
75
www.cobham.com/gaisler
GR740
7
Floating-point Control Unit
The GRFPU Control Unit (GRFPC) is used to attach the GRFPU to the LEON integer unit (IU).
GRFPC performs scheduling, decoding and dispatching of the FP operations to the GRFPU as well as
managing the floating-point register file, the floating-point state register (FSR) and the floating-point
deferred-trap queue (FQ). Floating-point operations are executed in parallel with other integer instructions, the LEON integer pipeline is only stalled in case of operand or resource conflicts.
Each of the four LEON4 processor cores in the system integrates a GRFPU control unit that connects
to one GRFPU unit per processor. Each processor has its own dedicated FPU.
7.1
Floating-Point register file
The GRFPU floating-point register file contains 32 32-bit floating-point registers (%f0-%f31). The
register file is accessed by floating-point load and store instructions (LDF, LDDF, STD, STDF) and
floating-point operate instructions (FPop).
7.2
Floating-Point State Register (FSR)
The GRFPC manages the floating-point state register (FSR) containing FPU mode and status information. All fields of the FSR register as defined in SPARC V8 specification are implemented and
managed by the GRFPU conforming to the SPARC V8 specification and the IEEE-754 standard.
Implementation-specific parts of the FSR managing are the NS (non-standard) bit and ftt field.
If the NS (non-standard) bit of the FSR register is set, all floating-point operations will be performed
in non-standard mode as described in section 8.2.6. When the NS bit is cleared all operations are performed in standard IEEE-compliant mode.
Following floating-point trap types never occur and are therefore never set in the ftt field:
- unimplemented_FPop: all FPop operations are implemented
- hardware_error: non-resumable hardware error
- invalid_fp_register: no check that double-precision register is 0 mod 2 is performed
GRFPU implements the qne bit of the FSR register which reads 0 if the floating-point deferred-queue
(FQ) is empty and 1 otherwise.
The FSR is accessed using LDFSR and STFSR instructions.
7.3
Floating-Point Exceptions and Floating-Point Deferred-Queue
GRFPU implements the SPARC deferred trap model for floating-point exceptions (fp_exception). A
floating-point exception is caused by a floating-point instruction performing an operation resulting in
one of following conditions:
•
an operation raises IEEE floating-point exception (ftt = IEEE_754_exception) e.g. executing
invalid operation such as 0/0 while the NVM bit of the TEM field id set (invalid exception
enabled).
•
an operation on denormalized floating-point numbers (in standard IEEE-mode) raises unfinished_FPop floating-point exception
•
sequence error: abnormal error condition in the FPU due to the erroneous use of the floatingpoint instructions in the supervisor software.
The trap is deferred to one of the floating-point instructions (FPop, FP load/store, FP branch) following the trap-inducing instruction (note that this may not be next floating-point instruction in the program order due to exception-detecting mechanism and out-of-order instruction execution in the
GRFPC). When the trap is taken the floating-point deferred-queue (FQ) contains the trap-inducing
instruction and up to seven FPop instructions that were dispatched in the GRFPC but did not complete.
GR740-UM-DS, Jun 2016, Version 1.4
76
www.cobham.com/gaisler
GR740
After the trap is taken the qne bit of the FSR is set and remains set until the FQ is emptied. The
STDFQ instruction reads a double-word from the floating-point deferred queue, the first word is the
address of the instruction and the second word is the instruction code. All instructions in the FQ are
FPop type instructions. The first access to the FQ gives a double-word with the trap-inducing instruction, following double-words contain pending floating-point instructions. Supervisor software should
emulate FPops from the FQ in the same order as they were read from the FQ.
Note that instructions in the FQ may not appear in the same order as the program order since GRFPU
executes floating-point instructions out-of-order. A floating-point trap is never deferred past an
instruction specifying source registers, destination registers or condition codes that could be modified
by the trap-inducing instruction. Execution or emulation of instructions in the FQ by the supervisor
software gives therefore the same FPU state as if the instructions were executed in the program order.
GR740-UM-DS, Jun 2016, Version 1.4
77
www.cobham.com/gaisler
GR740
8
High-performance IEEE-754 Floating-point Unit
8.1
Overview
GRFPU is a high-performance FPU implementing floating-point operations as defined in the IEEE
Standard for Binary Floating-Point Arithmetic (IEEE-754) and the SPARC V8 standard (IEEE-1754).
Supported formats are single and double precision floating-point numbers. The advanced design combines two execution units, a fully pipelined unit for execution of the most common FP operations and
a non-blocking unit for execution of divide and square-root operations.
The logical view of the GRFPU is shown in figure 7.
clk
Pipelined execution
unit
reset
GRFPU
start
opcode
9
ready
opid
6
allow
3
operand1
64
resid
6
operand2
64
result
64
round
2
except
6
Iteration unit
cc
flush
flushid
2
6
nonstd
Figure 7. GRFPU Logical View
8.2
Functional description
8.2.1
Floating-point number formats
GRFPU handles floating-point numbers in single or double precision format as defined in the IEEE754 standard with exception for denormalized numbers. See section 8.2.5 for more information on
denormalized numbers.
8.2.2
FP operations
GRFPU supports four types of floating-point operations: arithmetic, compare, convert and move. The
operations implement all FP instructions specified by SPARC V8 instruction set, and most of the
operations defined in IEEE-754. All operations are summarized in table 61.
GR740-UM-DS, Jun 2016, Version 1.4
78
www.cobham.com/gaisler
GR740
Table 61. : GRFPU operations
Operation
OpCode[8:0]
Op1
Op2
Result
Exceptions
Description
SP
DP
SP
DP
SP
DP
UNF, NV,
OF, UF, NX
Addition
SP
DP
SP
DP
SP
DP
UNF, NV,
OF, UF, NX
Subtraction
SP
DP
SP
SP
DP
SP
SP
DP
DP
UNF, NV,
OF, UF, NX
Multiplication, FSMULD gives
exact double-precision product of
two single-precision operands.
SP
DP
SP
DP
SP
DP
UNF, NV,
OF, UF, NX,
DZ
Division
-
-
SP
DP
SP
DP
UNF, NV,
NX
Square-root
-
INT
SP
DP
NX
-
Integer to floating-point conversion
-
SP
DP
INT
UNF, NV,
NX
Floating-point to integer conversion.
The result is rounded in round-tozero mode.
-
SP
DP
INT
UNF, NV,
NX
Floating-point to integer conversion.
Rounding according to RND input.
-
SP
DP
DP
SP
UNF, NV
UNF, NV,
OF, UF, NX
Conversion between floating-point
formats
SP
DP
SP
DP
CC
NV
Floating-point compare. Invalid
exception is generated if either operand is a signaling NaN.
SP
DP
SP
DP
CC
NV
Floating point compare. Invalid
exception is generated if either operand is a NaN (quiet or signaling).
000001001
-
SP
SP
-
Absolute value.
FNEGS
000000101
-
SP
SP
-
Negate.
FMOVS
000000001
SP
SP
-
Move. Copies operand to result output.
Arithmetic operations
FADDS
FADDD
001000001
FSUBS
FSUBD
001000101
FMULS
FMULD
FSMULD
001001001
FDIVS
FDIVD
001001101
FSQRTS
FSQRTD
000101001
001000010
001000110
001001010
001101001
001001110
000101010
UNF, NV,
OF, UF, NX
UNF, NV,
OF, UF
Conversion operations
FITOS
FITOD
011000100
FSTOI
FDTOI
011010001
FSTOI_RND
FDTOI_RND
111010001
FSTOD
FDTOS
011001001
011001000
011010010
111010010
011000110
Comparison operations
FCMPS
FCMPD
001010001
FCMPES
FCMPED
001010101
001010010
001010110
Negate, Absolute value and Move
FABSS
SP - single precision floating-point number
CC - condition codes INT - 32 bit integer
DP - double precision floating-point number
UNF, NV, OF, UF, NX - floating-point exceptions, see section 8.2.3
Arithmetic operations include addition, subtraction, multiplication, division and square-root. Each
arithmetic operation can be performed in single or double precision formats. Arithmetic operations
have one clock cycle throughput and a latency of four clock cycles, except for divide and square-root
operations, which have a throughput of 16 - 25 clock cycles and latency of 16 - 25 clock cycles (see
GR740-UM-DS, Jun 2016, Version 1.4
79
www.cobham.com/gaisler
GR740
table 62). Add, sub and multiply can be started on every clock cycle, providing high throughput for
these common operations. Divide and square-root operations have lower throughput and higher
latency due to complexity of the algorithms, but are executed in parallel with all other FP operations
in a non-blocking iteration unit. Out-of-order execution of operations with different latencies is easily
handled through the GRFPU interface by assigning an id to every operation which appears with the
result on the output once the operation is completed.
Table 62. : Throughput and latency
Operation
Throughput
Latency
FADDS, FADDD, FSUBS, FSUBD, FMULS, FMULD, FSMULD
1
4
FITOS, FITOD, FSTOI, FSTOI_RND, FDTOI, FDTOI_RND, FSTOD,
FDTOS
1
4
FCMPS, FCMPD, FCMPES, FCMPED
1
4
FDIVS
16
16
FDIVD
16.5 (15/18)*
16.5 (15/18)*
FSQRTS
24
24
FSQRTD
24.5 (23/26)*
24.5 (23/26)*
* Throughput and latency are data dependant with two possible cases with equal statistical possibility.
Conversion operations execute in a pipelined execution unit and have throughput of one clock cycle
and latency of four clock cycles. Conversion operations provide conversion between different floating-point numbers and between floating-point numbers and integers.
Comparison functions offering two different types of quiet Not-a-Numbers (QNaNs) handling are
provided. Move, negate and absolute value are also provided. These operations do not ever generate
unfinished exception (unfinished exception is never signaled since compare, negate, absolute value
and move handle denormalized numbers).
8.2.3
Exceptions
GRFPU detects all exceptions defined by the IEEE-754 standard. This includes detection of Invalid
Operation (NV), Overflow (OF), Underflow (UF), Division-by-Zero (DZ) and Inexact (NX) exception conditions. Generation of special results such as NaNs and infinity is also implemented. Overflow (OF) and underflow (UF) are detected before rounding. If an operation underflows the result is
flushed to zero (GRFPU does not support denormalized numbers or gradual underflow). A special
Unfinished exception (UNF) is signaled when one of the operands is a denormalized number which is
not handled by the arithmetic and conversion operations.
8.2.4
Rounding
All four rounding modes defined in the IEEE-754 standard are supported: round-to-nearest, round-to+inf, round-to--inf and round-to-zero.
8.2.5
Denormalized numbers
Denormalized numbers are not handled by the GRFPU arithmetic and conversion operations. A system (microprocessor) with the GRFPU could emulate rare cases of operations on denormals in software using non-FPU operations. A special Unfinished exception (UNF) is used to signal an arithmetic
or conversion operation on the denormalized numbers. Compare, move, negate and absolute value
operations can handle denormalized numbers and do not raise the unfinished exception. GRFPU does
not generate any denormalized numbers during arithmetic and conversion operations on normalized
numbers. If the infinitely precise result of an operation is a tiny number (smaller than minimum value
representable in normal format) the result is flushed to zero (with underflow and inexact flags set).
GR740-UM-DS, Jun 2016, Version 1.4
80
www.cobham.com/gaisler
GR740
8.2.6
Non-standard Mode
GRFPU can operate in a non-standard mode where all denormalized operands to arithmetic and conversion operations are treated as (correctly signed) zeroes. Calculations are performed on zero operands instead of the denormalized numbers obeying all rules of the floating-point arithmetics including
rounding of the results and detecting exceptions.
8.2.7
NaNs
GRFPU supports handling of Not-a-Numbers (NaNs) as defined in the IEEE-754 standard. Operations on signaling NaNs (SNaNs) and invalid operations (e.g. inf/inf) generate the Invalid exception
and deliver QNaN_GEN as result. Operations on Quiet NaNs (QNaNs), except for FCMPES and
FCMPED, do not raise any exceptions and propagate QNaNs through the FP operations by delivering
NaN-results according to table 63. QNaN_GEN is 0x7fffe00000000000 for double precision results
and 0x7fff0000 for single precision results.
Table 63. : Operations on NaNs
Operand 2



Operand 1
FP
QNaN2
SNaN2
none
FP
QNaN2
QNaN_GEN
FP
FP
QNaN2
QNaN_GEN
QNaN1
QNaN1
QNaN2
QNaN_GEN
SNaN1
QNaN_GEN
QNaN_GEN
QNaN_GEN
GR740-UM-DS, Jun 2016, Version 1.4
81
www.cobham.com/gaisler
GR740
9
Level 2 Cache controller
9.1
Overview
The L2 cache works as an AHB to AHB bridge, caching the data that is read or written via the bridge.
The cache is a unified cache and data may exist in both the processor Level-1 caches and the Level-2
cache, or only in a Level-1 or the Level-2 cache. A front-side AHB interface is connected to the Processor AHB bus, while a backend AHB interface is connected to the Memory AHB bus. Figure 8
shows a system block diagram for the cache controller.
Note that the L2 cache is disabled after reset and should be enabled by boot software.
CPU
CPU
Processor AHB bus
L2C
Memory AHB bus
Memory
Controller
Figure 8. Block diagram
9.2
Operation
The Level-2 cache is implemented as a multi-way cache with an associativity of four. The replacement policy can be configured as: LRU (least-recently-used), pseudo-random or master-index (where
the way to replace is determine by the master index). The way size is 512 KiB with a line size of 32
bytes.
9.2.1
Replacement policy
The cache supports three different replacement policies: LRU (least-recently-used), (pseudo-) random
and master-index. The reset value for replacement policy is LRU.
With the master-index replacement policy, master 0 would replace way 1, master 1 would replace way
2 and so on. For master indexes corresponding to a way number larger than the number of implemented ways there are two options to determine which way to replace. One option is to map all these
master index to a specific way. This is done by specifying this way in the index-replace field in the
control register and selecting this option in the replacement policy field also located in the control register. It is not allowed to select a locked way in the index-replace field. The second option is to replace
way = ((master index) modulus (number of ways)). This option can be selected in the replacement
policy field.
9.2.2
Write policy
The cache can be configured to operate as write-through or copy-back cache. Before changing the
write policy to write-through, the cache has to be disabled and flushed (to write back dirty cache lines
to memory). This can be done by setting the Cache disable bit when issue a flush all command. The
GR740-UM-DS, Jun 2016, Version 1.4
82
www.cobham.com/gaisler
GR740
write policy is controlled via the cache control register. More fine-grained control can also be
obtained by enabling the MTRR registers (see text below).
9.2.3
Memory type range registers
The memory type range registers (MTRR) are used to control the cache operation with respect to the
address. Each MTRR can define an area in memory to be uncached, write-through or write-protected.
Each MTRR register consist of a 14-bit address field, a 14-bit mask and two 2-bit control fields. The
address field is compared to the 14 most significant bits of the cache address, masked by the mask
field. If the unmasked bits are equal to the address, an MTRR hit is declared. The cache operation is
then performed according to the control fields (see register descriptions). If no hit is declared or if the
MTRR is disabled, cache operation takes place according to the cache control register. The number of
implemented MTRRs is sixteen. When changing the value of any MTRR register, the cache must be
disabled and flushed (this can be done by setting the Cache disable bit when issuing a flush all command).
Note that the write-protection provided via the MTRR registers is enforced even if the cache is disabled.
9.2.4
Cachability
The cache considers the address range 0x00000000 - 0x7FFFFFFF to be cachable. The cache can also
be configured to use the HPROT signal to override the default cachable area. An access can only be
redefined as non-cachable by the HPROT signal. See table 64 for information on how HPROT can
change the access cachability within a cachable address area. The AMBA AHB signal HPROT[3]
defines the access cacheable when active high and the AMBA AHB signal HPROT[2] defines the
access as bufferable when active high.
Table 64. Access cachability using HPROT.
HPROT:
non-cachable, non-bufferable
non-cachable, bufferable
cacheable
Read hit
Cache access*
Cache access
Cache access
Read miss
Memory access
Memory access
Cache allocation and Memory access
Write hit
Cache and Memory access
Cache access
Cache access
Write miss
Memory access
Memory access
Cache allocation
* When the HPROT-Read-Hit-Bypass bit is set in the cache control register this will generate a Memory access.
9.2.5
Cache tag entry
Table 65 shows the different fields of the cache tag entry for a cache with a way size of 512 KiB.
Table 65. L2C Cache tag entry
31
19
18
TAG
31 : 19
10
000000
9
8
Valid
7
6
Dirty
5
RES
4
0
LRU
Address Tag (TAG) - Contains the address of the data held in the cache line.
GR740-UM-DS, Jun 2016, Version 1.4
83
www.cobham.com/gaisler
GR740
9:8
9.2.6
Table 65. L2C Cache tag entry
Valid bits. When set, the corresponding sub-block of the cache line contains valid data. Valid bit 0
corresponds to the lower 16 bytes sub-block (with offset 1) in the cache line and valid bit 1 corresponds to the upper 16 bytes sub-block (with offset 0) in the cache line.
7:6
Dirty bits When set, this sub-block contains modified data.
5
RESERVED
4:0
LRU bits
AHB address mapping
The AHB slave interface occupies three AHB address ranges. The first AHB memory bar is used for
memory/cache data access and is mapped at 0x00000000 - 0x7FFFFFFF. The second AHB memory
bar is used for access to configuration registers and the diagnostic interface and is mapped at
0xF0800000 - 0xF08FFFFF. The last AHB memory bar is used to map the IO area of the backend
AHB bus (to access the plug&play information on that bus) and maps the Memory AHB bus area
0xFFE00000 - 0xFFEFFFFF.
9.2.7
Memory protection and Error handling
The L2 cache provides Error Detection And Correction (EDAC) protection for the data and tag memory. One error can be corrected and two error can be detected with the use of a (39, 32, 7) BCH code.
The EDAC functionality can dynamically be enabled or disabled. Before being enabled the cache
should be flushed. The dirty and valid bits fore each cache line is implemented with TMR. When
EDAC error or backend AHB error or write-protection hit in a MTRR register is detected, the error
status register is updated to store the error type. The address which caused the error is also saved in
the error address register. The error types is prioritised in the way that a uncorrected EDAC error will
overwrite any other previously stored error in the error status register. In all other cases, the error status register has to be cleared before a new error can be stored. Each error type (correctable-, uncorrectable EDAC error, write-protection hit, backend AHB error) has a pending register bit. When set
and this error is unmasked, a interrupt is generated. When an uncorrectable error is detected in the
read data, the cache will respond with an AHB error. AHB error responses can also be enabled for
access that match a stored error in the error status register. Error detection is done per cache line. The
cache also provides a correctable error counter accessible via the error status register. After power-up
the error status register needs to be cleared before any valid data can be read out.
GR740-UM-DS, Jun 2016, Version 1.4
84
www.cobham.com/gaisler
GR740
Table 66. Cache action on detected EDAC error
Access/Error type
Cache-line not dirty
Read, Correctable
Tag error
Tag is corrected before read is handled, Error sta- Tag is corrected before read is handled, Error
tus is updated with a corretable error.
status is updated with a corretable error.
Read, Uncorrectable
Tag error
Cache-line invalidated before read is handled,
Error status is updated with a corretable error.
Write, Correctable
Tag error
Tag is corrected before write is handed, Error sta- Tag is corrected before write is handled, Error
tus is updated with a corretable error.
status is updated with a corretable error.
Write, Uncorrectable Tag error
Cache-line invalidated before write is handled,
Error status is updated with a correctable error.
Cache-line invalidated before write is handled,
Error status is updated with a uncorrectable
error. Cache data is lost.
Read, Correctable
Data error
Cache-data is correted and updated, Error status
is updated with a correctable error. AHB access
is not affected.
Cache-data is correted and updated, Error status is updated with a correctable error. AHB
access is not affected.
Read, Uncorrectable
Data error
Cache-line is invalidated, Error status is updated
with a correctable error. AHB access is terminated with retry.
Cache-line is invalidated, Error status is
updated with a uncorrectable error. AHB
access is terminated with error.
Write (<32-bit), Correctable Data error
Cache-data is correted and updated, Error status
is updated with a correctable error. AHB access
is not affected.
Cache-data is correted and updated, Error status is updated with a correctable error. AHB
access is not affected.
Write (<32-bit),
Uncorrectable Data
error
Cache-line is re-fetched from memory, Error sta- Cache-line is invalidated, Error status is
tus is updated with a correctable error. AHB
updated with a uncorrectable error. AHB
access is not affected.
access write data and cache data is lost.
9.2.8
Cache-line dirty
Cache-line invalidated before read is handled,
Error status is updated with a uncorrectable
error. Cache data is lost.
Scrubber
The cache is implemented with an internal memory scrubber to prevent build-up of errors in the cache
memories. The scrubber is controlled via two registers in the cache configuration interface. To scrub
one specific cache line the index and way of the line is set in the scrub control register. The scrub
operation is started by setting the the pending bit to 1. The scrubber can also be configured to continuously loop through and scrub each cache line by setting the enabled bit to 1. In this mode, the delay
between the scrub operation on each cache line is determine by the scrub delay register (in clock
cycles).
9.2.9
Locked way
One or more ways can be configured to be locked (not replaced). The number of ways that should be
locked is configured by the locked-way field in the control register. The way to be locked is starting
with the uppermost way (for a 4-way associative cache way 4 is the first locked way, way 3 the second, and so on). After a way is locked, the cache-way has to be flushed with the “way flush” function
to update the tag to match the desired locked address. During this “way flush” operation, the data can
also be fetched from memory.
9.3
Operation
9.3.1
Read
A cachable read access to the cache results in a tag lookup to determine if the requested data is located
in the cache memory. For a hit (requested data is in the cache) the data is read from the cache and no
read access is issued to the memory. If the requested data is not in the cache (cache miss), the cache
controller issues a read access to the memory controller to fetch the cache line containing the
requested data. The replacement policy determines which cache line in a multi-way configuration that
GR740-UM-DS, Jun 2016, Version 1.4
85
www.cobham.com/gaisler
GR740
should be replaced and its tag is updated. If the replaced cache line is modified (dirty) this data is
stored in a write buffer and after the requested data is fetched from memory the replaced cache line is
written to memory.
For a non-cachable read access to the cache, the cache controller can issue a single read access or a
burst read access to fetch the data from memory. The access type is determine by how the cache is
configured regarding hprot support and bypass line fetch in the access control register. The data is
stored in a read buffer and the state of the cache is not modified in any way.
The cache will insert wait-states until the read access is determined to be a cache hit or miss. For a
cache hit the data is then delivered. For a miss the cache can insert wait-states during the memory
fetch or issue a AMBA SPLIT (depending on how the cache is configured).
9.3.2
Write
A cachable write access to the cache results in a tag lookup to determine if the cache line is present in
the cache. For a hit the cache line is updated. No access is issued to the memory for a copy-back configuration. When the cache is configured as a write-through cache, each write access is also issued
towards memory. For a miss, the replacement policy determines which cache line in a multi-way configuration that should be replaced and updates its tag. If the replaced cache line is dirty, it is stored in
a write buffer to be written back to the memory. The new cache line is updated with the data from the
write access and for a non-128-bit access the rest of the cache line is fetched from memory. Last the
replaced cache line is written to memory (when copy-back policy is used and the replaced cache line
was marked dirty). When the cache is configured as a write-through cache, no cache lines are marked
as dirty and no cache line needs to be written back to memory. Instead the write access is issued
towards the memory as well. A new cache line is allocated on a miss for a cacheable write access
independent of write policy (copy-back or write-through).
For a non-cachable write access to the cache, the data is stored in a write buffer and the cache controller issue single write accesses to write the data to memory. The state of the cache is unmodified during
this access.
The cache can accept a non sub-word write hit access every clock cycle. When the cache is unable to
accept a new write access the cache inserts wait-states or issue a AMBA SPLIT response depending
on how the cache is configured.
9.3.3
Cache flushing
The cache can be flushed by accessing a cache flush register. There are three flush modes: invalidate
(reset valid bits), write back (write back dirty cache lines to memory, but no invalidation of the cache
content) and flush (write back dirty cache lines to memory and invalidate the cache line). The flush
command can be applied to the entire cache, one way or to only one cache line. The cache line to be
flushed can be addresses in two ways: direct address (specify way and line address) and memory
address (specify which memory address that should be flushed in the cache. The controller will make
a cache lookup for the specified address and on a hit, flush that cache line). When the entire cache is
flushed the Memory Address field should be set to zero. To invalidate a cache line takes 5 clock
cycles. If the cache line needs to be written back to memory one additional clock cycle is needed plus
the memory write latency. When the whole cache is flushed the invalidation of the first cache line
takes 5 clock cycles, after this one line can be invalidated each clock cycle. When a cache line needs
to be written back to memory this memory access will be stored in an access buffer. If the buffer is
full, the invalidation of the next cache line will stall until a slot in the buffer has opened up. If the
cache also should be disabled after the flush is complete, it is recommended to set the cache disable
bit together with the flush command in the Fush set/index register instead of writing ‘0’ to the cache
enable bit in the cache control register.
Note that after a processor (or any other AHB master) has initiated a flush the processor is not blocked
by the flush unless it writes or requests data from the Level-2 cache. The cache blocks all accesses
GR740-UM-DS, Jun 2016, Version 1.4
86
www.cobham.com/gaisler
GR740
(responds with AMBA SPLIT or wait-states depending on cache configuration) until the flush is complete.
9.3.4
Disabling Cache
To be able to safely disable the cache when it is being accessed, the cache need to be disabled and
flushed at the same time. This is accomplished by setting the cache disable bit when issue the flush
command.
9.3.5
Diagnostic cache access
The diagnostic interface can be used for RAM block testing and direct access to the cache tag, cache
data content and EDAC check bits. The read-check-bits field in the error status/control register selects
if data content or the EDAC check bits should be read out. On writes, the EDAC check bits can be
selected from the data-check-bit or tag-check-bit register. These register can also be XOR:ed with the
correct check bits on a write. See the error status/control register for how this is done.
9.3.6
Error injection
Error injection can be performed for data and tag lines either by modifying the value or the checkbits.
The checkbits can be modified by defining a mask that will be XOR:ed with the generated checkbits
or by defining the full checkbits to be written via the tag-check-bit register or data-check-bit-registers.
The value can be modified by performing a diagnostic access while keeping the existing checkbits.
EDAC checkbits can be modified on a regular cache access by setting the xor-check-bit field in the
error status/control register the data EDAC check bits will be XOR:ed with the data-check-bit register
on the next write, or the tag EDAC check bits will be XOR:ed with the tag-check-bit register on the
next tag replacement. The tag check bit manipulation is only done if the tag-check-bit register is not
zero. The xor-check-bit is reset on the next tag replacement or data write. Errors can also be injected
by writing an address together with the inject bit to the “Error injection” register. This will XOR the
check-bits for the specified address with the data-check-bit register. If the specified address in not
cached, the cache contents will be unchanged.
9.3.7
AHB slave interface
The cache can accept 8-bit (byte), 16-bit (half word), 32-bit (word), 64-bit, and 128-bit single
accesses and also 32-bit, 64-bit, and 128-bit burst accesses. For an access during a flush operation, the
cache will respond with an AHB SPLIT response or with wait-states. For an uncorrectable error or a
backend AHB error on a read access, the cache will respond with an AMBA ERROR response. For a
correctable data error which require a cache line to be re-fetched from memory the cache will respond
with a AMBA RETRY response.
9.3.8
AHB master interface
The master interface is the cache’s connection to the memory controller. During cache line fetch, the
controller can issue either a 32-bit, 64-bit or 128-bit burst access. For a non cachable access and in
write-through mode the cache can also issue a 8-bit (byte), 16-bit (half word), 32-bit (word), 64-bit, or
128-bit single write access.
9.3.9
Cache status
The cache controller has a status register that provides information on the cache configuration (multiway configuration and set size). The cache also provides access, hit and error correction counters via
the LEON4 statistics unit (see section 31).
GR740-UM-DS, Jun 2016, Version 1.4
87
www.cobham.com/gaisler
GR740
9.4
Registers
The cache is configured via registers mapped into the AHB memory address space.
Table 67. L2C: AHB registers
AHB address offset
Register
0x00
Control register
0x04
Status register
0x08
Flush (Memory address)
0x0C
Flush (set, index)
0x10 - 0x1C
Reserved
0x20
Error status/control
0x24
Error address
0x28
TAG-check-bit
0x2C
Data-check-bit
0x30
Scrub Control/Status
0x34
Scrub Delay
0x38
Error injection
0x3C
Access control
0x80 - 0xFC
MTRR registers
0x80000 - 0x8FFFC
Diagnostic interface (Tag)
0x80000: Tag 1, way-1
0x80004: Tag 1, way-2
0x80008: Tag 1, way-3
0x8000C: Tag 1, way-4
0x80010: Tag check-bits way-0,1,2,3 (Read only)
bit[31] = RESERVED
bit[30:24] = check-bits for way-1.
bit[23] = RESERVED
bit[22:16] = check-bits for way-2.
bit[15] = RESERVED
bit[14:8] = check-bits for way-3.
bit[7] = RESERVED
bit[6:0] = check-bits for way-4.
0x80020: Tag 2, way-1
0x80024: ...
0x200000 - 0x3FFFFC
Diagnostic interface (Data)
0x200000 - 0x27FFFC: Data or check-bits way-1
0x280000 - 0x2FFFFF: Data or check-bits way-2
0x300000 - 0x27FFFC: Data or check-bits way-3
0x380000 - 0x3FFFFF: Data or check-bits way-4
When check-bits are read out:
Only 32-word at offset 0x0, 0x10, 0x20,... are valid check-bits.
bit[31] = RESERVED
bit[30:24] = check-bits for data word at offset 0x0.
bit[23] = RESERVED
bit[22:16] = check-bits for data word at offset 0x4.
bit[15] = RESERVED
bit[14:8] = check-bits for data word at offset 0x8.
bit[7] = RESERVED
bit[6:0] = check-bits for data word at offset 0xc.
GR740-UM-DS, Jun 2016, Version 1.4
88
www.cobham.com/gaisler
GR740
9.4.1
Control register
Table 68. 0x00 - L2CC - L2C Control register
31
29 28 27
16 15
EN ED REPL
AC
0
RESERVED
12 11
INDEX-WAY
8
7
6
LOCK
RES
5
4
3
2
1
0
HP HP UC HC WP HP
RH B
B
0
0
0
0
0
0
0
rw rw
rw
r
rw
rw
r
rw rw rw rw rw rw
9.4.2
0
0
0
31
Cache enable (EN) - When set, the cache controller is enabled. When disabled, the cache is
bypassed.
30
EDAC enable (EDAC)
29: 28
Replacement policy (REPL) - 
00: LRU
01: (pseudo-) random
10: Master-index using index-replace field
11: Master-index using the modulus function
0
0
27: 16
RESERVED
15: 12
Master-index replacement (INDEX-WAY) - Way to replace when Master-index replacement policy
and master index is larger than number of ways in the cache.
11: 8
Locked ways (LOCK) - Number of locked ways.
7: 6
RESERVED
5
HPROT read hit bypass (HPRHB) - When set, a non-cacheable and non-bufferable read access will
bypass the cache on a cache hit and return data from memory. Only used with HPROT support.
4
HPROT bufferable (HPB) - When HPROT is used to determine cachability and this bit is set, all
accesses is marked bufferable.
3
Bus usage status mode (UC) - 0 = wrapping mode, 1 = shifting mode.
2
Hit rate status mode (HC) - 0 = wrapping mode, 1 = shifting mode.
1
Write policy (WP) - When set, the cache controller uses the write-through write policy. When not
set, the write policy is copy-back.
0
HPROT enable (HP) - When set, use HPROT to determine cachability.
Status register
Table 69. 0x04 - L2CS - L2C Status register
31
25 24 23 22 21
RESERVED
LS AT MP
16 15
13 12
2
1
0
MTRR
BBUS-W
WAY-SIZE
WAY
0
0
0
1
16
1
*
3
r
r
r
r
r
r
r
r
31: 25
RESERVED
24
Cache line size (LS) - 1 = 64 bytes, 0 = 32 bytes.
23
Access time (AT) - Access timing not simulated.
22
Memory protection (MP) - implemented
21: 16
Memory Type Range Registers (MTRR) - Number of MTRR registers implemented (16)
15: 13
Backend bus width (BBUS-W) Set to 1 = 128-bit.
12: 2
Cache way size (WAY-SIZE) - Size in kBytes
1: 0
Multi-Way configuration (WAY)
Set to “11“: 4-way
GR740-UM-DS, Jun 2016, Version 1.4
89
www.cobham.com/gaisler
GR740
9.4.3
Flush memory address register
Table 70. 0x08 - L2CFMA - L2C Flush (Memory address) register
31
4
3
Memory Address (ADDR)
5
R
DI
2
0
NR
0
0
0
rw
r
w
rw
FMODE
31: 5
Memory Address (ADDR) - (For flush all cache lines, this field should be set to zero)
4
RESERVED
3
Cache disable (DI) - Setting this bit to ‘1’ is equal to setting the Cache enable bit to ‘0’ in the Cache
Control register
2: 0
Flush mode (FMODE) -
“001“: Invalidate one line, “010”: Write-back one line, “011“: Invalidate & Write-back one line.
“101“: Invalidate all lines, “110”: Write-back all lines, “111“: Invalidate & Write-back all lines.
Only dirty cache lines are written back to memory.
Flush set/index register
9.4.4
Table 71. 0x0C - L2CFSI - L2C Flush (Set, Index) register
31
16
INDEX / TAG
10
9
8
7
6
FL VB DB R
0
5
4
WAY
3
2
1
0
DI WF FMODE
NR
0
0
0
0
0
0
0
rw
rw rw rw
r
rw
w
rw
rw
31: 16
Cache line index (INDEX) - used when a specific cache line is flushed
31: 10
(TAG) - used when “way flush” is issued. If a specific cache line is flushed, bit should be set to zero.
When a way flush is issued, the bits in this field will be written to the TAGs for the selected cache
way.
9
Fetch Line (FL) - If set to ‘1’ data is fetched form memory when a “way flush” is issued. If a specific
cache line is flushed, this bit should be set to zero
8
Valid bit (VB) - used when “way flush” is issued. If a specific cache line is flushed, this bit should be
set to zero.
7
Dirty bit (DB) - used when “way flush” is issued. If a specific cache line is flushed, this bit should be
set to zero
6
RESERVED
5: 4
Cache way (WAY) -
3
Cache disable (DI) - Setting this bit to ‘1’ is equal to setting the Cache enable bit to ‘0’ in the Cache
Control register.
2
Way-flush (WF) - When set one way is flushed, If a specific cache line should be flushed, this bit
should be set to zero
1: 0
Flush mode (FMODE) line flush:
“01“: Invalidate one line
“10”: Write-back one line (if line is dirty)
“11“: Invalidate & Write-back one line (if line is dirty).
way flush:
“01“: Update Valid/Dirty bits according to register bit[8:7] and TAG according to register
bits[31:10]
“10”: Write-back dirty lines to memory
“11“: Update Valid/Dirty bits according to register bits [8:7] and TAG according to register
bits[31:10], and Write-back dirty lines to memory.
GR740-UM-DS, Jun 2016, Version 1.4
90
www.cobham.com/gaisler
GR740
9.4.5
Error status/control register
Table 72. 0x20 - L2CERR - L2CError status/control register
31
28 27 26
24 23 22 21 20 19 18
AHB
master
index
S
C
R
U
B
TYPE
NR
NR
NR
r
r
r
T
A
G
/
D
A
T
A
C
O
R
/
U
C
O
R
M
U
L
T
I
V
A
L
I
D
NR NR NR NR
r
r
r
r
16 15
12 11
8
7
IRQ
mask
6
D
I
S
E
R
E
S
P
Correctable
error
counter
IRQ
pending
Select
CB
0
NR
NR
0
0
rw
r
r
rw
rw
4
3
2
1
0
Select
TCB
5
X
C
B
R
C
B
C
O
M
P
R
S
T
0
0
0
0
0
rw
rw rw rw
w
31: 28
AHB master that generated the access
27
Scrub error (SCRUB) - Indicates that the error was trigged by the scrubber.
26: 24
Access/Error Type: (TYPE) -
000: cache read, 001: cache write, 010: memory fetch, 011: memory write,
100: Write-protection hit, 101: backend read AHB error, 110: backend write AHB error
23
Tag or data Error - 0 tag error, 1: data error
22
Correctable or uncorrectable error - 0: correctable error, 1: uncorrectable error
21
Multiple error (MULTI) - set when multiple errors has been detected.
20
Error status valid (VALID) - register contains valid error status.
19
Disable error responses for uncorrectable EDAC error (DISERESP).
18: 16
Correctable error counter
15: 12
Interrupt pending
bit3: Backend AHB error
bit2: Write-protection hit
bit1: Uncorrectable EDAC error
bit0: Correctable EDAC error
11: 8
Interrupt mask (if set this interrupt is unmasked)
bit3: Backend AHB error
bit2: Write-protection hit
bit1: Uncorrectable EDAC error
bit0: Correctable EDAC error
7: 6
Selects (CB) - data-check-bits for diagnostic data write:
00: use generated check-bits
01: use check-bits in the data-check-bit register
10: XOR check-bits with the data-check-bit register
11: use generated check-bits
Note: If this field is set to "01" or "10" then check-bits are overridden for all accesses. To get controlled error injection, the internal scrubber should be disabled and no accesses should be made to
the Level-2 cache.
5: 4
Selects (TCB) - tag-check-bits for diagnostic tag write:
00: use generated check-bits
01: use check-bits in the tag-check-bit register
10: XOR check-bits with the tag-check-bit register
11: use generated check-bits
Note: If this field is set to "01" or "10" then check-bits are overridden for all accesses. To get controlled error injection, the internal scrubber should be disabled and no accesses should be made to
the Level-2 cache.
3
Xor check-bits (XOR) - If set, the check-bits for the next data write or tag replace will be XOR:ed
withe the check-bit register. Default value is 0.
2
Read check-bits (RCB) - If set, a diagnostic read to the cache data area will return the check-bits
related to that data.When this bit is set, check bits for the data at offset 0x0 - 0xc can be read at offset
0x0, the check bits for data at offset 0x10 - 0x1c can be read at offset 0x10, ...
GR740-UM-DS, Jun 2016, Version 1.4
91
www.cobham.com/gaisler
GR740
Table 72. 0x20 - L2CERR - L2CError status/control register
1
Compare error status (COMP) - If set, a read access matching a uncorrectable error stored in the
error status register will generate a AHB error response. Default value is 0.
0
Resets (RST) - clear the status register to be able to store a new error. After power up the status register needs to be cleared before any valid data can be read out.
GR740-UM-DS, Jun 2016, Version 1.4
92
www.cobham.com/gaisler
GR740
9.4.6
Error address register
Table 73. 0x24 - L2CERRA - L2C Error address register
31
0
Error Address (EADDR)
NR
r
31: 0
9.4.7
Error Address (EADDR)
TAG check bits register
Table 74. 0x28 - L2CTCB - L2C TAG-Check-Bits register
31
7
6
0
RESERVED
TCB
0
0
r
rw
31: 7
RESERVED
6: 0
TAG Check-bits (TCB) - Check-bits which can be selected by the “Select check-bit“ field in the
error status/control register for TAG updates
Data check bits register
9.4.8
Table 75. 0x2C - L2CCB - L2C Data-Check-Bits register
31
28 27
0
RESERVED
CB
0
0
r
rw
31: 28
RESERVED
27: 0
Data Check-bits (CB) - Check-bits which can be selected by the “Select check-bit“ field in the error
status/control register for TAG updates
Scrub control/status register
9.4.9
Table 76. 0x30 - L2CSCRUB - L2C Scrub control/status register
31
16 15
4
3
2
WAY
1
0
INDEX
RESERVED
PE EN
N
0
0
0
0
rw
r
rw
rw rw
0
31: 16
Scrub Index (INDEX) - Index for the next line scrub operation
15: 4
RESERVED
3: 2
Scrub Way (WAY) - Way for the next line scrub operation
1
Scrub Pending (PEN) - Indicates when a line scrub operation is pending. When the scrubber is disabled, writing ‘1’ to this bit scrubs one line.
0
Scrub Enable (EN) - Enables / disables the automatic scrub functionality.
9.4.10 Scrub delay register
Table 77. 0x34 - L2CSDEL - L2C Scrub delay register
31
16 15
0
RESERVED
DEL
0
0
GR740-UM-DS, Jun 2016, Version 1.4
93
www.cobham.com/gaisler
GR740
Table 77. 0x34 - L2CSDEL - L2C Scrub delay register
r
rw
31: 16
RESERVED
15: 0
Scrub Delay (DEL) - Delay the scrubber waits before issue the next line scrub operation
9.4.11 Error injection register
Table 78. 0x38 - L2CEINJ - L2C Error injection register
31
2
ADDR
0
0
0
0
rw
r
rw
31: 2
Error Inject address (ADDR)
1:
RESERVED
0
Inject error (INJ) - Set to ‘1’ to inject a error at “address”.
GR740-UM-DS, Jun 2016, Version 1.4
1
R INJ
94
www.cobham.com/gaisler
GR740
9.4.12 Access control register
Table 79. 0x3C - L2CACCC - L2C Access control register
31
11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
SP NH BE OA FLI DB 128 R DB SP R
LIT M RR PM NE PF WF
PW LIT
Q
S
0
0
0
0
0
0
0
r
rw rw rw rw rw rw rw
0
0
0
0
0
r
rw rw
r
31: 11
RESERVED
10
SPLIT queue write order (SPLITQ) When set, all write accesses (except locked) will be placed in the
split queue when the split queue is not empty
9
No hit for cache misses (NHM) - When set, the unsplited read access for a read miss will not trig the
access/hit counters.
8
Bit error status (BERR) - When set, the error status signals will represent the actual error detected
rather then if the error could be corrected by refetching data from memory.
7
One access/master (OAPM) - When set, only one ongoing access per master is allowed to enter the
cache. A second access would receive a SPLIT response
6
(FLINE) - When set, a cache line fetched from memory can be replaced before it has been read out
by the requesting master.
5
Disable bypass prefetching (DBPF) - When set, bypass accesses will be performed as single accesses
towards memory.
4
128-bit write line fetch (128WF) - When set, a 128-bit write miss will fetch the rest of the cache
from memory.
3
RESERVED
2
Disable wait-states for discarded bypass data (DBPWS) - When set, split response is given to a
bypass read access which data has been discarded and needs to refetch data from memory.
1
Enabled SPLIT response (SPLIT) - When set the cache will issue a AMBA SPLIT response on
cache miss
0
RESERVED
9.4.13 Memory type range registers
Table 80. 0x80-FC - L2CMTRR - L2C Memory type range register
31
18 17 16 15
ADDR
ACC
2
MASK
1
0
WP AC
0
0
0
0
rw
rw
rw
rw rw
31: 18
Address field (ADDR) - to be compared to the cache address [31:18]
17: 16
Access field (ACC) - 00: uncached, 01: write-through
15: 2
Address mask (MASK) - Only bits set to 1 will be used during address comparison
1
Write-protection (WP) - 0: disabled, 1: enabled
0
Access control field (AC) -. 0: disabled, 1: enabled
GR740-UM-DS, Jun 2016, Version 1.4
95
0
www.cobham.com/gaisler
GR740
10
SDRAM Memory Controller with Reed-Solomon EDAC
10.1
Overview
The SDRAM memory controller is a 64+32-bit memory controller which is divided into a front-end
and a back-end part.
AHB Front-end
with EDAC
AHB slave I/F
Read/Write
data buffers
SDR back-end
to SDRAM
mem_ifwidth
mem_ifwidth
Figure 9. Memory controller connected to AMBA bus and SDRAM
10.2
Operation
10.2.1 Memory data width
The controller supports a full-width and a half-width mode, selected via the MEM_IFWIDTH input
signal. In full-width mode, the memory bus has 64 data bits, and 0,16 or 32 check bits depending on
EDAC configuration. In half-width mode, the memory bus has 32 data bits, plus 0,8 or 16 check bits.
10.2.2 Memory access
When an AHB access is done to the controller, the corresponding request is sent to the memory backend which performs the access. For read bursts, the controller streams the read data so each burst item
is delivered to the bus as soon as it arrives and wait states are added as needed between each part of
the burst.
The controller has a write buffer holding one write access in EDAC configuration, and two write
accesses in non-EDAC configuration. Each write access can be up to the configured burst length in
size. The controller will mask the write latency by storing the data into the write buffer and releasing
the AHB bus immediately. The latency will be seen however if a read access is done before the writes
have completed or an additional write access is made when all buffers are used.
Writes of 32 bits or less will result in a read-modify-write cycle to update the checkbits (this is done
even if EDAC has been disabled in the control register). In this case, the memory controller generates
wait states on the AHB bus until the read part of the cycle has completed.
10.3
Limitations
The AHB front-end with EDAC is optimized for 64/128-bit masters and does not handle 32-bit bursts
efficiently, each access will result in a RMW cycle in the write case, and a read cycle in the read case.
10.4
SDRAM back-end operation
10.4.1 General
Synchronous dynamic RAM (SDRAM) access is supported to two banks of PC100/PC133 compatible devices. The controller supports 64M, 256M and 512M devices with 8 - 12 column-address bits,
GR740-UM-DS, Jun 2016, Version 1.4
96
www.cobham.com/gaisler
GR740
up to 13 row-address bits, and 4 banks. The size of each of the two banks can be programmed in
binary steps between 4 Mbyte and 512 Mbyte. The operation of the SDRAM controller is controlled
through the configuration register SDCFG (see section 10.6).
10.4.2 Initialization
When the SDRAM controller is enabled, it automatically performs the SDRAM initialization
sequence of PRECHARGE, 8x AUTO-REFRESH and LOAD-MODE-REG on both banks simultaneously. The controller programs the SDRAM to use page burst on read accesses and single location
access on write accesses.
10.4.3 Read and write cycles
A read cycle is started by performing an ACTIVATE command to the desired bank and row, followed
by a READ command with data read after the programmed CAS delay. The read cycle is terminated
with a PRE-CHARGE command, no banks are left open between two accesses.
Write cycles are performed similarly to read cycles, with the difference that WRITE commands are
issued after activation.
10.4.4 Configurable SDRAM timing parameters
To provide optimum access cycles for different SDRAM devices (and at different frequencies), three
SDRAM parameters can be programmed through the memory configuration register (SDCFG):
TCAS, TRP and TRFCD. The value of these fields affect the SDRAM timing as described in table 81.
Table 81. SDRAM programmable minimum timing parameters
SDRAM timing parameter
Minimum timing (clocks)
CAS latency, RAS/CAS delay (tCAS, tRCD)
TCAS + 2
Precharge to activate (tRP)
TRP + 2
Auto-refresh command period (tRFC)
TRFC + 3
Activate to precharge (tRAS)
TRFC + 1
Activate to Activate (tRC)
TRP + TRFC + 4
If the TCAS, TRP and TRFC are programmed such that the PC100/133 specifications are fulfilled,
the remaining SDRAM timing parameters will also be met. The table below shows typical settings for
100 and 133 MHz operation and the resulting SDRAM timing (in ns):
Table 82. SDRAM example programming
SDRAM settings
tCAS
tRC
tRP
tRFC
tRAS
100 MHz, CL=2; TRP=0, TCAS=0, TRFC=4
20
80
20
70
50
100 MHz, CL=3; TRP=0, TCAS=1, TRFC=4
30
80
20
70
50
133 MHz, CL=2; TRP=1, TCAS=0, TRFC=6
15
82
22
67
52
133 MHz, CL=3; TRP=1, TCAS=1, TRFC=6
22
82
22
67
52
10.4.5 Refresh
The SDRAM controller contains a refresh function that periodically issues an AUTO-REFRESH
command to both SDRAM banks. The period between the commands (in clock periods) is programmed in the refresh counter reload field in the SDCFG register. Depending on SDRAM type, the
GR740-UM-DS, Jun 2016, Version 1.4
97
www.cobham.com/gaisler
GR740
required period is typically 7.8 or 15.6 s (corresponding to 780 or 1560 clocks at 100 MHz). The
generated refresh period is calculated as (reload value+1)/(memory clock period). The refresh function is enabled by setting bit 31 in the SDCFG register.
10.4.6 2T signaling mode
An alternative mode is supported in the controller where all address and control signals except for the
chip select is set for an extra cycle with chip select deasserted before the command is issued. This
improves the board timing analysis for these signals since an extra cycle of setup time is achieved.
The price of this is one extra SDRAM cycle of latency for the ACTIVATE command, resulting in one
cycle extra read latency, and also one extra cycle for the REFRESH. Also due to internal design reasons, one extra cycle before PRECHARGE after a read burst (but not after writing) is also inserted,
however this does not affect latency as it is done in parallel with the transfer of read data to the AHB
domain.
The 2T signaling mode is activated by setting the EN2T bit in the SDRAM configuration register 2
SDCFG2) register.
In order for the 2T signaling mode to work, the mode register needs to be programmed for length-2
read and length-2 write bursts. Therefore, after changing the EN2T setting, the LOAD-MODE-REGISTER command must be reissued by writing to the SDRAM command configuration register field
before proceeding with additional accesses.
10.4.7 Double chip select mode
As the 2T signaling mode improves analog setup time for all control signals except chip select, an
additional setting exists to group the four chip select outputs into two chip selects, where each chip
select signal is driven identically on two outputs in parallel. On the board, the two copies can each
then be routed to half of the SDRAM devices and thereby reduce the capacitive load on each output.
Table 83. Mapping of chip selects to I/Os
Pin
Function DCS=0
Function DCS=1
mem_sn(0)
CS(0)
CS(0)
mem_sn(1)
CS(1)
CS(0)
mem_addr(13)
CS(2)
CS(1)
mem_addr(14)
CS(3)
CS(1)
10.4.8 SDRAM commands
The controller can issue four SDRAM commands by writing to the SDRAM command field in the
SDRAM Configuration register: PRE-CHARGE, AUTO-REFRESH and LOAD-MODE-REG
(LMR). The command field will be cleared after a command has been executed. Note that when
changing the value of the CAS delay or enabling/disabling 2T signaling, a LOAD-MODE-REGISTER command should be generated at the same time to update the mode register. The mode register is
programmed as shown in table 84.
Table 84. SDRAM controller mode register programming
Mode register bit
Controller setting
12 .. 10
9
8
(reserved)
WB
Op Mode
CAS Latency
0
EN2T
0
0
GR740-UM-DS, Jun 2016, Version 1.4
7
6
0
98
5
1
4
CL
3
2
BT
Burst length
1
0
0
0
0
EN2T
www.cobham.com/gaisler
GR740
10.4.9 Address bus connection
The SDRAM address bus should be connected to mem_addr[12:0], the bank address to mem_ba[1:0],
and the data bus to mem_dq[63:0] or mem_dq[31:0] if a 32-bit SDRAM data bus is used. Checkbits
should be tied to mem_dq[95:64] in full-width mode and mem_dq[79:64] in half-width mode.
10.4.10 Command sequences
Below illustrates the sequences of commands issued by the SDRAM controller for read and write
bursts at the minimum delay settings. CS and DQM in the table refers to the CS and DQM belonging
to the selected address and size, unused banks and byte lanes are kept at 1 throughout the access.
Table 85. Length-4 read access command sequence
Cycle
2T signaling disabled
2T signaling enabled
CSn CMD
DQM
BA
Addr
CSn
CMD
DQM
BA
Addr
-
1
(NOP)
(1)
-
-
1
(NOP)
1
-
-
0
0
RAS
(1)
Bank
Row
1
(RAS)
1
(Bank)
(Row)
1
0
NOP
1
(Bank)
(Row)
0
RAS
1
Bank
Row
2
0
READ
0
Bank
Col
1
(READ)
1
(Bank)
(Col)
3
0
READ
0
Bank
Col+1
0
READ
0
Bank
Col
4
0
READ
0
Bank
Col+2
1
(READ)
0
(Bank)
(Col+2)
5
0
READ
0
Bank
Col+3
0
READ
0
Bank
Col+2
6
0
READ*
0
Bank
Col+4
1
(READ)
0
(Bank)
(Col+4)
7
0
READ*
0
Bank
Col+5
0
READ*
0
Bank
Col+4
8
0
READ*
0
Bank
Col+6
1
(READ)
0
(Bank)
Col+6
9
0
PCH
1
Bank
(Col+6)
0
READ*
0
Bank
Col+6
10
1
(NOP)
1
(Bank)
(Col+6)
1
(PCH)
1
(Bank)
(Col+8)
11
1
(NOP)
1
(Bank)
(Col+6)
0
PCH
1
Bank
(Col+8)
12 ...
1
(NOP)
1
(Bank)
(Col+6)
1
NOP
1
(Bank)
(Col+8)
Note: * Dummy read while waiting for end of data, data not used
Table 86. Length-4 write access command sequence
Cycle
2T signaling disabled
2T signaling enabled
CSn CMD
DQM
BA
Addr
CSn
CMD
DQM
BA
Addr
-
1
(NOP)
1
-
-
1
(NOP)
1
-
-
0
0
RAS
1
Bank
Row
1
(RAS)
1
Bank
(Row)
1
0
NOP
1
(Bank)
(Row)
0
RAS
1
Bank
Row
2
0
WRITE
0
Bank
Col
1
(WRITE) 1
Bank
(Col)
3
0
WRITE
0
Bank
Col+1
0
WRITE
0
Bank
Col
4
0
WRITE
0
Bank
Col+2
1
(WRITE) 0
Bank
(Col+2)
5
0
WRITE
0
Bank
Col+3
0
WRITE
0
Bank
Col+2
6
0
WRITE
1
Bank
Col+4
1
(WRITE) 0
Bank
(Col+4)
7
0
PCH
1
Bank
Col+5
1
(PCH)
1
Bank
(Col+4)
8
1
NOP
1
(Bank)
Col+6
0
PCH
1
Bank
Col+6
9
1
NOP
1
(Bank)
(Col+6)
1
NOP
1
Bank
Col+6
GR740-UM-DS, Jun 2016, Version 1.4
99
www.cobham.com/gaisler
GR740
10.5
Fault-tolerant operation
10.5.1 Overview
For FT operation, the external memory interface data bus is widened and the extra bits are used to
store 16 or 32 checkbits corresponding to each 64 bit data word. The variant to be used can be configured at run-time depending on the connected data width and the desired level of fault tolerance.
When writing, the controller generates the check bits and stores them along with the data. When reading, the controller will transparently correct any correctable bit errors and provide the corrected data
on the AHB bus. However, the corrected bits are not written back to the memory so external scrubbing is necessary to avoid uncorrectable errors accumulating over time.
An extra corrected error output signal is asserted when a correctable read error occurs, at the same
cycle as the corrected data is delivered. This signal is connected to the memory scrubber. In case of
uncorrectable error, this is signaled by giving an AHB ERROR response.
10.5.2 Error-correction properties
The memory controller uses an interleaved error correcting code which works on nibble (4-bit) units
of data. The codec can be used in two interleaving modes, mode A and mode B.
In mode A, the basic code has 16 data bits, 8 check bits and can correct one nibble error. This code is
interleaved by 4 using the pattern in table 87 to create a code with 64 data bits and 32 check bits.
This code can tolerate one nibble error in each of the A,B,C,D groups shown below. This means that
we can correct 100% of single errors in two adjacent nibbles, or in any 8/16-bit wide data bus lane,
that would correspond to a physical memory device. The code can also correct 18/23=78% of all possible random two-nibble errors.
This interleaving pattern was designed to also provide good protection in case of reduced (32/16-bit)
bus width with the same data-checkbit relation, so software will see the exact same checkbits on diagnostic reads.
In mode B, the basic code has 32 data bits, 8 check bits and can correct one nibble error. This code is
then interleaved by a factor of two to create a code with 64 data bits and 16 check bits.
Table 87. Mode Ax4 interleaving pattern (64-bit data width)
63:60
59:56
55:52
51:48
47:44
43:40
39:36
35:32
31:28
27:24
23:20
19:16
15:12
11:8
7:4
C
D
A
B
A
B
C
D
B
A
D
C
D
C
B
A
95:88
87:80
79:72
71:64
Ccb
Dcb
Acb
Bcb
3:0
127:120 119:112 111:104 103:96
Ccb
Dcb
Acb
Bcb
3:0
Table 88. Mode Bx2 interleaving pattern (64-bit data width)
63:60
59:56
55:52
51:48
47:44
43:40
39:36
35:32
31:28
27:24
23:20
19:16
15:12
11:8
7:4
A
B
A
B
A
B
A
B
B
A
B
A
B
A
B
A
95:88
87:80
79:72
71:64
Acb
Bcb
Acb
Bcb
Table 89. Mode Ax4 interleaving pattern (32-bit data width)
95:80
79:76
75:72
71:68
67:64
63:32
31:28
27:24
23:20
19:16
15:12
11:8
7:4
3:0
-
Ccb
Dcb
Acb
Bcb
-
C
D
A
B
A
B
C
D
-
Bcb
Acb
Dcb
Ccb
-
B
A
D
C
D
C
B
A
95:80
79:76
75:72
71:68
67:64
63:32
31:28
27:24
23:20
19:16
15:12
11:8
7:4
3:0
-
-
-
Acb
Bcb
-
A
B
A
B
A
B
A
B
-
-
-
Bcb
Acb
-
B
A
B
A
B
A
B
A
Table 90. Mode Bx2 interleaving pattern (32-bit data width)
GR740-UM-DS, Jun 2016, Version 1.4
100
www.cobham.com/gaisler
GR740
10.5.3 Data transfers
There is no extra time penalty in the case data is corrected compared to the error-free case.
Only writes of 64 bit width or higher will translate directly into write cycles to the external memory.
Other types of write accesses will generate a read-modify-write (RMW) cycle in order to correctly
update the check-bits. In the special case where an uncorrectable error is detected while performing
the RMW cycle, the write is aborted and the incorrect checkbits are left unchanged so they will be
detected upon the next read.
10.5.4 Configuration
Checkbits are always generated when writing even if EDEN is disabled. Which type of code, A or B,
that is used can be controlled by the CODE field in the FTCFG register. If the code is changed during
operation, you will need to re-initialize the memory to regenerate the check-bits with the new code.
One way to do this is to clear EDEN and then read and rewrite the memory contents.
Code checking on read is disabled on reset and is enabled by setting the EDEN bit in the FTCFG register. Before enabling this, the code to be used should be set in the CODE field and the memory contents should be (re-)initialized.
10.5.5 Diagnostic checkbit access
The checkbits and data can be accessed directly for testing and fault injection. This is done by writing
the address of into the FTDA register. The check-bits and data can then be read and written via the
FTDC and FTDD registers. Note that for checkbits the FTDA address is 64-bit aligned, while for data
it is 32-bit aligned.
When the FTDC or FTDD registers are accessed, the corresponding access to the address configured
in the FTDA register will be performed to the SDRAM and the read data is returned as if it was the
contents of the register. The access will block with wait states until the access has completed, and
unlike regular accesses so also writes will always block.
After the diagnostic data register has been read, the FT control register bits 31:19 can be read out to
see if there were any correctable or uncorrectable errors detected, and where the correctable errors
were located. There is one bit per byte lane describing where any correctable errors occurred. Note
that the location mask is not valid if there were uncorrectable errors.
10.5.6 Code boundary
The code boundary feature allows you to gradually switch the memory from one interleaving mode to
the other and regenerate the checkbits without stopping normal operation. This can be used when
recovering from memory faults, as explained further below.
If the boundary address enable (BAEN) control bit is set, the controller will look at the address of
each access, and use the interleaving mode selected in the CODE field for memory accesses above or
equal to the boundary address, and the opposite code for memory accesses below to the boundary
address.
If the boundary address update (BAUPD) control bit is also set, the controller will shift the boundary
upwards whenever the the address directly above the boundary is written to. Since the written data is
now below the boundary, it will be written using the opposite code. The write can be done with any
size supported by the controller.
10.5.7 Data multiplexing
When code B is used instead of code A, the upper half of the checkbits become unused. The controller
supports switching in this part of the data bus to replace another faulty part of the bus. To do this, one
GR740-UM-DS, Jun 2016, Version 1.4
101
www.cobham.com/gaisler
GR740
sets the DATAMUX field to a value between 1-4 to replace a quarter of the data bus, or to 5 to replace
the active checkbit half. When writing, the selected part will also be written into the top bits, and
when reading the top bits will be copied over into the selected part.
Table 91. DATAMUX configurations
mem_ifwidth
DATAMUX
Top bits swapped in
-
0
No swapping
0
1
mem_dq(95:80)
1
Replaced "faulty" slice
mem_dq(15:0)
2
mem_dq(31:16)
3
mem_dq(47:32)
4
mem_dq(63:48)
5
mem_dq(79:64)
1
mem_dq(79:72)
mem_dq(7:0)
2
mem_dq(15:8)
3
mem_dq(23:16)
4
mem_dq(31:24)
5
mem_dq(71:64)
10.5.8 Memory fault recovery
The above features are designed to make the system capable to deal with a permanent fault in an
external memory chip.
A basic sequence of events is as follows:
1.
The system is running correctly with EDAC enabled and the larger code A is used.
2. A memory chip gets a fault making the SDRAM deliver incorrect data on one byte lane. The
memory controller keeps delivering error-free data but reports a correctable error on every read
access.
3. A logging device (the memory scrubber) registers the high frequency of correctable errors and
signals an interrupt.
4. The CPU performs a probe using the FT diagnostic registers to confirm that the error is permanent and on which physical lane the error is.
5. After determining that a permanent fault has occurred, the CPU reconfigures the memory controller as follows (all configuration register fields changed with a single register write):
The data multiplexing control field is set so the top checkbit half replaces the failed part of the
data bus.
The code boundary register is set to the lowest memory address.
The boundary address enable and boundary address update enable bits are set.
The mask correctable error bit is set
6. The memory data and checkbits are now regenerated using locked read-write cycles to use the
smaller code and replace the broken data with the upper half of the checkbit bus. This can be done in
hardware using the memory scrubber.
7. After the whole memory has been regenerated, the CPU disables the code boundary, changes the
code selection field to code B, and unsets the mask correctable error bit.
After this sequence, the system is now again fully operational, but running with the smaller code and
replacement chip and can again recover from any single-nibble error. Note that during this sequence,
it is possible for the system to operate and other masters can both read and write to memory while the
regeneration is ongoing.
GR740-UM-DS, Jun 2016, Version 1.4
102
www.cobham.com/gaisler
GR740
10.6
Registers
The controller has a register area accessible via the AHB slave interface. The registers should be
accessed with 32-bit reads and writes. The registers are tabulated below.
Table 92. MMCTRL Registers
Offset
Register
0x00
SDRAM configuration register 1(SDCFG1)
0x04
SDRAM configuration register 2 (SDCFG2)
0x08 - 0x1C Reserved
0x20
Mux Configuration Register (MUXCFG)
0x24
Mux Diagnostic Address register (FTDA)
0x28
FT Diagnostic Checkbit register (FTDC)
0x2C
FT Diagnostic Data register (FTDD)
0x30
FT Code Boundary Register (FTBND)
0x34 - 0xFF
GR740-UM-DS, Jun 2016, Version 1.4
Reserved
103
www.cobham.com/gaisler
GR740
10.6.1 SDRAM configuration register 1
Table 93. 0x00 - SDCFG1 - SDRAM configuration register 1
31 30 29
RF tRP
0
27 26 25
tRFC
1
rw rw
23 22 21 20
18 17 16 15 14
COLSZ COMMAND
R MS 64
0
tC
BANKSZ
RFLOAD
0b111
1
0b000
10
0
0
0
*
NR
rw
rw
rw
rw
rw
r
r
r
rw
31
SDRAM refresh (RF) - If set, the SDRAM refresh will be enabled.
30
SDRAM tRP timing (tRP) - tRP will be equal to 2 or 3 system clocks (0/1). When mobile SDRAM
support is enabled, this bit also represent the MSB in the tRFC timing.
29: 27
SDRAM tRFC timing (tRFC) - tRFC will be equal to 3 + field-value system clocks. When mobile
SDRAM support is enabled, this field is extended with the bit 30.
26
SDRAM CAS delay (tC) - Selects 2 or 3 cycle CAS delay (0/1). When changed, a LOAD-COMMAND-REGISTER command must be issued at the same time. Also sets RAS/CAS delay (tRCD).
25: 23
SDRAM banks size (BANKSZ) - Defines the decoded memory size for each SDRAM chip select:
“000”= 4 Mbyte, “001”= 8 Mbyte, “010”= 16 Mbyte .... “111”= 512 Mbyte.
22: 21
SDRAM column size (COLSZ) - “00”=256, “01”=512, “10”=1024, “11”=4096 when bit[25:23]=
“111”, 2048 otherwise.
20: 18
SDRAM command (COMMAND) - Writing a non-zero value will generate an SDRAM command:
“010”=PRECHARGE, “100”=AUTO-REFRESH, “110”=LOAD-COMMAND-REGISTER. The
field is reset after command has been executed.
17
RESERVED
16
Mobile SDRAM support (MS) - Disabled
15
64-bit data bus (64) - Reads ‘1’ if memory controller is configured for 64-bit data bus, otherwise ‘0’.
Read-only. Affected by value of MEM_IFWIDTH bootstrap signal.
14: 0
Refresh counted reload value (RFLOAD) - The period between each AUTO-REFRESH command Calculated as follows: tREFRESH = ((reload value) + 1) / SYSCLK
10.6.2 SDRAM configuration register 2
Table 94. 0x04 - SDCFG2 - SDRAM configuration register 2
31 30 29
R CE
24
16 15 14 13
RESERVED
0
E
N
2
T
D
C
S
RESERVED
0
1
0
0
0
0
r
rw
r
rw rw
r
31
RESERVED
30
Clock enable (CE) - This value is driven on the CKE inputs of the SDRAM. Should be set to ‘1’ for
correct operation..
29: 16
RESERVED
15
Enable 2T signaling (EN2T)
14
Double chip select mode (DCS)
13: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
104
www.cobham.com/gaisler
GR740
10.6.3 Mux configuration register
Table 95. 0x20 - MUXCFG - Mux configuration register
31
20 19 18
16 15
12 11
8
ERRLOC
D
D
E
R
R
DWIDTH
BEID
RESERVED
0x000
0
*
0b0001
0
r
r
r
r
r
5
4
3
2
1
0
DATAMUX
7
C
E
M
B
A
U
P
D
B
A
E
N
C
O
D
E
E
D
E
N
0
0
0
0
0
0
rw
rw rw rw rw rw
31: 20
Diag data read error location (ERRLOC) - Bit field describing location of corrected errors for last
diagnostic data read. One bit per byte lane in 64+32-bit configuration.
19
Set high if last diagnostic data read contained an uncorrectable error (read-only)
18: 16
Data width (DWIDTH) - 010=32+16, 011=64+32 bits
15: 12
Back-end identifier (BEID) - “0001” - SDRAM
11: 8
RESERVED
7: 5
Data mux control (DATAMUX) - setting this nonzero switchess in the upper checkbit half with
another data lane.
For 64-bit interface
000 = no switching
001 = Data bits 15:0, 010 = Data bits 31:16, 011: Data bits 47:32, 100: Data bits 63:48,
101 = Checkbits 79:64, 110,111 = Undefined
4
Correctable error masking (CEM) - If set to 1, the correctable error signal is masked out.
3
Enable automatic boundary shifting on write (BAUPD)
2
Enable the code boundary (BAEN)
1
Code selection (CODE) - 0=Code A (64+32/32+16/16+8), 1=Code B (64+16/32+8) (FT only)
0
EDAC Enable (EDEN) - Set to 1 to enable EDAC
10.6.4 FT diagnostic address register
Table 96. 0x24 - FTDA - FT diagnostic address register
31
2
1
0
FTDA
RES
0
0
rw
r
31: 2
Address to memory location for checkbit read/write (FTDA) - 64/32-bit aligned for checkbits/data
1: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
105
www.cobham.com/gaisler
GR740
10.6.5 FT diagnostic checkbits register
Table 97. 0x28 - FTDC - FT diagnostic checkbits register
31
24 23
CBD
16 15
CBC
8
7
CBB
0
CBA
*
*
*
*
rw
rw
rw
rw
31: 24
Checkbits for part D of 64-bit data word (CBD) - (undefined for code B)
23: 16
Checkbits for part C of 64-bit data word (CBC) - (undefined for code B)
15: 8
Checkbits for part B of 64-bit data word (CBB)
7: 0
Checkbits for part A of 64-it data word. (CBA)
Note that this is a "virtual" register backed by memory, an access to it will result in a corresponding access to the
memory location configured in the FTDA register.
10.6.6 FT diagnostic data register
Table 98. 0x2C - FTDD - FT diagnostic data register
31
0
DATA
*
r
31: 0
Uncorrected data (DATA) - For 32-bit address set in FTDA register
Note that this is a "virtual" register backed by memory, an access to it will result in a corresponding access to the
memory location configured in the FTDA register.
10.6.7 FT boundary address register
Table 99. 0x30 - FTBND - FT boundary address register
31
3
2
0
FTBND(31:3)
RESERVED
0
0
rw
0
31: 3
Code boundary address (FTBND) - 64-bit aligned. Field contains address bits 31:3. Bits 2:0 are
always zero.
2: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
106
www.cobham.com/gaisler
GR740
11
Memory Scrubber and AHB Status Register
11.1
Overview
The memory scrubber monitors the Memory AHB bus for accesses triggering an AMBA ERROR
response, and for correctable errors signaled from fault tolerant memory controllers on the same bus.
The memory scrubber can be programmed to scrub a memory area by reading through the memory
and writing back the contents using a locked read-write cycle whenever a correctable error is detected.
It can also be programmed to initialize a memory area to known values.
The memory scrubber register interface is largely backward compatible with the AHB status register.
AMBA AHB
Scrubber DMA
Registers
Memory with EDAC
AHB Error monitor
ce
Figure 10. Memory scrubber block diagram
11.2
Operation
11.2.1 Errors
All AMBA AHB bus transactions are monitored and current HADDR, HWRITE, HMASTER and
HSIZE values are stored internally. When an error response (HRESP = “01”) is detected, an internal
counter is increased. When the counter exceeds a user-selected threshold, the status and address register contents are frozen and the New Error (NE) bit is set to one. At the same time an interrupt is generated, as described hereunder.
The default threshold is zero and enabled on reset so the first error on the bus will generate an interrupt.
The fault-tolerant memory controllers signal an un-correctable error as an AMBA error response, so
that it can be detected by the processor as described above.
11.2.2 Correctable errors
Not only AMBA ERROR responses on the AHB bus can be detected. The memory controllers on the
Memory AHB bus have a correctable error signal that is asserted each time a correctable error is
detected. When such an error is detected, the effect will be the same as for an AHB error response.
The only difference is that the Correctable Error (CE) bit in the status register is set to one when a correctable error is detected. Correctable and uncorrectable errors use separate counters and threshold
values.
When the CE bit is set, the interrupt routine can acquire the address containing the correctable error
from the failing address register and correct it. When it is finished it resets the CE bit and the monitoring becomes active again. Interrupt handling is described in detail hereunder.
GR740-UM-DS, Jun 2016, Version 1.4
107
www.cobham.com/gaisler
GR740
11.2.3 Scrubbing
The memory scrubber can be commanded to scrub a certain memory area, by writing a start and end
address to the scrubber’s start/end registers, followed by writing “00” to the scrub mode field and ‘1’
to the scrub enable bit in the scrubber control register.
After starting, the memory scrubber will proceed to read the memory region in bursts. The burst size
is fixed to eight 32-bit words. When a correctable error is detected, the scrubber performs a locked
read-write cycle to correct the error, and then resumes the scrub operation.
If a correctable error detected is in the middle of a burst, the following read in the burst is completed
before the read-write cycle begins. The memory scrubber can handle the special case where that
access also had a correctable error within the same locked scrub cycle.
If an uncorrectable error is detected, that location is left untouched.
Note that the status register functionality is running in parallel with the scrubber, so correctable and
uncorrectable errors will be logged as usual. To prevent double logging, the memory scrubber masks
out the (expected) correctable error arising during the locked correction cycle.
To allow normal access to the bus, the memory scrubber sleeps for a number of cycles between each
burst. The number of cycles can be adjusted in the config register.
If the ID bit is set in the config register, the memory scrubber will interrupt when the complete scrub
is done.
11.2.4 Scrubber error counters
The memory scrubber keeps track of the number of correctable errors detected during the current
scrub run and the number of errors detected during processing of the current “count block”. The size
of the count block is a fixed power of two equal or larger than the burst length (set to eight 32-bit
words).
The memory scrubber can be set up to interrupt when the counters exceed given thresholds. When this
happens, the NE bit, plus one of the SEC/SBC bits, is set in the status register.
11.2.5 External start
If the ES bit is set in the config register, the scrub enable bit is set automatically when the start input
signal goes high. This can be used to set up periodic scrubbing. The start input signal is connected to
the tick output of timer four on the system’s first general purpose timer unit (GPTIMER 0). The tick
output will be high for one clock cycle when the fourth timer underflows.
11.2.6 Memory regeneration
The regeneration mode performs the same basic function as the scrub mode, but is optimised for the
case where many (or all) locations have correctable errors.
In this mode, the whole memory area selected is scrubbed using locked read/write bursts.
If an uncorrectable error is encountered during the read burst, that burst block is processed once again
using the regular scrub routine, and the regeneration mode resumes on the following block. This
avoids overwriting uncorrectable error locations.
11.2.7 Initialization
The scrubber can be used to write a pre-defined pattern to a block of memory. This is often necessary
on EDAC memory before it can be used.
Before running the initialization, the pattern to be written to memory should be written into the scrubber initialization data register. The pattern has the same size as the burst length, so the corresponding
number of writes to the initialization data register must be made.
GR740-UM-DS, Jun 2016, Version 1.4
108
www.cobham.com/gaisler
GR740
11.2.8 Interrupts
After an interrupt is generated, either the NE bit or the DONE bit in the status register is set, to indicate which type of event caused the interrupt.
The normal procedure is that an interrupt routine handles the error with the aid of the information in
the status registers. When it is finished it resets the NE bit in the AHB status register or the DONE bit
in the scrubber status register, and the monitoring becomes active again. Error interrupts can be generated for both AMBA ERROR responses and correctable errors as described above.
11.2.9 Mode switching
Switching between scrubbing and regeneration modes can be done on the fly during a scrub by modifying the MODE field in the configuration register. The mode change will take effect on the following
scrub burst.
If the address range needs to be changed, then the memory scrubber should be stopped before updating the registers. This is done by clearing the SCEN bit, and waiting for the ACTIVE bit in the status
register to go low. An exception is when making the range larger (i.e. increasing the end address or
decreasing the start address), as this can be done on the fly.
11.2.10 Dual range support
The scrubber can work over two non-overlapping memory ranges. This feature is enabled by writing
the start/end addresses of the second range into the scrubber’s second range start/end registers and setting the SERA bit in the configuration register. The two address ranges should not overlap.
GR740-UM-DS, Jun 2016, Version 1.4
109
www.cobham.com/gaisler
GR740
11.3
Registers
The memory scrubber is programmed through registers mapped into an I/O region in the AHB
address space. Only 32-bit accesses are supported
.
Table 100.Memory scrubber registers
AHB address offset
Registers
0x00
AHB Status register
0x04
AHB Failing address register
0x08
AHB Error configuration register
0x0C
Reserved
0x10
Status register
0x14
Configuration register
0x18
Range low address register
0x1C
Range high address register
0x20
Position register
0x24
Error threshold register
0x28
Initialization data register
0x2C
Second range start address register
0x30
Second range end address register
GR740-UM-DS, Jun 2016, Version 1.4
110
www.cobham.com/gaisler
GR740
11.3.1 AHB status register
Table 101.0x00 - AHBS - AHB Status register
31
22 21
13
12
11
10
UECNT
DONE
RES
SEC
SBC
0
0
0
0
0
0
0
rw
rw
r
r
rw
rw
rw rw
CECNT
14
9
8
CE NE
0
7
6
3
2
0
HWRITE
HMASTER
HSIZE
NR
NR
NR
r
r
r
31: 22
Global correctable error count (CECNT) - Global correctable error count
21: 14
Global uncorrectable error count (UECNT) - Global uncorrectable error count
13
Task Completed (DONE): Task completed.
This is a read-only copy of the DONE bit in the status register.
12
RESERVED
11
Scrubber error counter threshold exceeded (SEC) - Scrubber error counter threshold exceeded.
Asserted together with NE.
10
Scrubber block error counter threshold exceeded (SBC) - Scrubber block error counter threshold
exceeded. Asserted together with NE.
9
Correctable Error (CE) - Correctable Error. Set if the detected error was caused by a correctable error
and zero otherwise.
8
New Error (NE) - Deasserted at start-up and after reset. Asserted when an error is detected. Reset by
writing a zero to it.
7
AMBA write signal (HWRITE) - The HWRITE signal of the AHB transaction that caused the error.
6: 3
AMBA master signal (HMASTER) - The HMASTER signal of the AHB transaction that caused the
error.
2: 0
AMBA size signal (HSIZE) - The HSIZE signal of the AHB transaction that caused the error
11.3.2 AHB failing address register
Table 102.0x04 - AHBFAR - AHB Failing Address Register
31
0
AHB FAILING ADDRESS
NR
r
31: 0
AHB failing address (AHB FAILING ADDRESS) - The HADDR signal of the AHB transaction that
caused the error.
GR740-UM-DS, Jun 2016, Version 1.4
111
www.cobham.com/gaisler
GR740
11.3.3 AHB error configuration register
Table 103.0x08 - AHBERC - AHB Error configuration register
31
22 21
CECNTT
14 13
2
UECNTT
RESERVED
0
0
0
rw
rw
1
0
C
E
C
T
E
U
E
C
T
E
0
0
rw rw
31: 22
Interrupt threshold value for global correctable error count (CECNTT) - Interrupt threshold value for
global correctable error count
21: 14
Interrupt threshold value for global uncorrectable error count (UECNTT) - Interrupt threshold value
for global uncorrectable error count
13: 2
RESERVED
1
Correctable error count threshold enable (CECTE) - Correctable error count threshold enable
0
Uncorrectable error count threshold enable (UECTE) - Uncorrectable error count threshold enable
11.3.4 Status register
Table 104.0x10 - STAT - Status register
31
22 21
14 13 12
RUNCOUNT
BLKCOUNT
0
r
5
4
2
1
0
D
O
N
E
RESERVED
BURSTLEN
A
C
T
I
V
E
0
0
0
0x1
0
r
wc
r
r
r
31: 22
Run error count (RUNCOUNT) - Number of correctable errors in current scrub run
21: 14
Block error count (BLKCOUNT) - Number of correctable errors in current block
13
Task completed (DONE) - Task completed.
Needs to be cleared (by writing zero) before a new task completed interrupt can occur.
12: 5
RESERVED
4: 1
Burst length (BURSTLEN) - 2-log of AHB bus cycles; “0001”=2
0
Current state (ACTIVE) - 0=Idle, 1=Running
GR740-UM-DS, Jun 2016, Version 1.4
112
www.cobham.com/gaisler
GR740
11.3.5 Configuration register
Table 105.0x14 - CONFIG - Configuration register
31
16 15
RESERVED
8
DELAY
7
6
5
4
3
I
R
Q
D
R
S
E
R
A
L
O
O
P
MODE ES
2
1
0
S
C
E
N
0
0
0
0
0
0
0
0
r
rw
rw
r
rw rw
rw
rw rw
31: 16
RESERVED
15: 8
Delay time between processed blocks (DELAY) - Defines delay in cycles
7
Interrupt when DONE (IRQD) - Interrupt when task has completed
6
RESERVED
5
Second memory range enable (SERA) - Enables second memory range
4
Loop mode (LOOP) - Restart scrubber when run finishes
3: 2
Operation Mode (MODE) - 00=Scrub, 01=Regenerate, 10=Initialize, 11=Undefined
1
External start enable (ES) - If set to ’1’ then external start is enabled.
0
Scrubber enable (SCEN) - Enables scrubber
0
11.3.6 Range low address register
Table 106.0x18 - RANGEL - Range low address register
31
5
4
0
RLADDR
31: 0
0
0b00000
rw
r
Scrubber range low address (RLADDR) - The lowest address in the range to be
scrubbed
The address bits below the burst size alignment are constant ‘0’
11.3.7 Range high address register
Table 107.0x1C - RANGEH - Range high address register
31
5
4
0
RHADDR
31: 0
0
0b11111
rw
r
Scrubber range high address (RHADDR) - The highest address in the range to be
scrubbed
The address bits below the burst size alignment are constant ‘1’
GR740-UM-DS, Jun 2016, Version 1.4
113
www.cobham.com/gaisler
GR740
11.3.8 Position register
Table 108.0x20 - POS - Position register
31
5
4
0
POSITION
31: 0
0
0b00000
rw
r
Scrubber position (POSITION) - The current position of the scrubber while active,
otherwise zero.
The address bits below the burst size alignment are constant ‘0’
11.3.9 Error threshold register
Table 109.0x24 - ETHRES - Error threshold register
31
22 21
RECT
31: 22
14 13
2
1
0
R
E
C
T
E
B
E
C
T
E
0
0
0
r
rw rw
BECT
RESERVED
0
0
rw
rw
Interrupt threshold value for current scrub run correctable error count (RECT)
21: 14
Interrupt threshold value for current scrub block correctable error count (BECT)
13: 2
RESERVED
1
Scrub run correctable error count threshold enable (RECTE)
0
Scrub block correctable error count threshold enable (BECTE)
11.3.10 Initialisation data register
Table 110.0x28 - INIT - Initialisation data register
31
22 21
14 13
2
1
0
DATA
w
31: 0
Initialisation data (DATA) - Part of data pattern to be written in initialisation mode. A write operation assigns the first part of the buffer and moves the rest of the words in the buffer one step.
GR740-UM-DS, Jun 2016, Version 1.4
114
www.cobham.com/gaisler
GR740
11.3.11 Second range low address register
Table 111.0x2C - RANGEL2 - Second range low address register
31
5
4
0
RLADDR
31: 0
0
0b00000
rw
r
Scrubber range low address (RLADDR) - The lowest address in the range to be
scrubbed (if CONFIG.SERA = 1)
The address bits below the burst size alignment are constant ‘0’
11.3.12 Second range high address register
Table 112.0x30 - RANGEH2 - Second range high address register
31
5
4
0
RHADDR
31: 0
0
0b11111
rw
r
Scrubber range high address (RHADDR) - The highest address in the range to be
scrubbed (if CONFIG.SERA = 1)
The address bits below the burst size alignment are constant ‘1’
GR740-UM-DS, Jun 2016, Version 1.4
115
www.cobham.com/gaisler
GR740
12
Bridge connecting Debug AHB bus to Processor AHB bus
12.1
Overview
The Debug AHB bus is connected to the Processor AHB bus via a uni-directional AHB/AHB bridge.
The bridge provides:
12.2
•
Propagation of single and burst AHB transfers
•
Data buffering in internal FIFOs
•
Efficient bus utilization through use of AMBA SPLIT response and data prefetching
•
Posted writes
•
Read and write combining, improves bus utilization and allows connecting cores with differing
AMBA access size restrictions.
Operation
12.2.1 General
For AHB write transfers write data is always buffered in an internal FIFO implementing posted
writes. For AHB read transfers the bridge uses AMBA Plug&Play information to determine whether
the read data will be prefetched and buffered in an internal FIFO. If the target address for an AHB
read burst transfer is a prefetchable location the read data will be prefetched and buffered.
An AHB master initiating a read transfer to the bridge is always splitted on the first transfer attempt to
allow other masters to use the slave bus while the bridge performs read transfer on the master bus.
12.2.2 AHB read transfers
When a read transfer is registered on the slave interface the bridge gives a SPLIT response. The master that initiated the transfer will be de-granted allowing other bus masters to use the slave bus while
the bridge performs a read transfer on the master side. The master interface then requests the bus and
starts the read transfer on the master side. Single transfers on the slave side are normally translated to
single transfers with the same AHB address and control signals on the master side, however read combining can translate one access into several smaller accesses. Translation of burst transfers from the
slave to the master side depends on the burst type, burst length and access size.
If the transfer is a burst transfer to a prefetchable location, the master interface will prefetch data in
the internal read FIFO. If the splitted burst on the slave side was an incremental burst of unspecified
length (INCR), the length of the burst is unknown. In this case the master interface performs an incremental burst up to a 32-byte address boundary. When the burst transfer is completed on the master
side, the splitted master that initiated the transfer (on the slave side) is allowed in bus arbitration by
asserting the appropriate HSPLIT signal to the AHB controller. The splitted master re-attempts the
transfer and the bridge will return data with zero wait states.
If the burst is to non-prefetchable area, the burst transfer on the master side is performed using
sequence of NONSEQ, BUSY and SEQ transfers. The first access in the burst on the master side is of
NONSEQ type. Since the master interface can not decide whether the splitted burst will continue on
the slave side or not, the master bus is held by performing BUSY transfers. On the slave side the splitted master that initiated the transfer is allowed in bus arbitration. The first access in the transfer is
completed by returning read data. The next access in the transfer on the slave side is extended by
asserting HREADY low. On the master side the next access is started by performing a SEQ transfer
(and then holding the bus using BUSY transfers). This sequence is repeated until the transfer is ended
on the slave side.
GR740-UM-DS, Jun 2016, Version 1.4
116
www.cobham.com/gaisler
GR740
In case of an ERROR response on the master side the ERROR response will be given for the same
access (address) on the slave side. SPLIT and RETRY responses on the master side are re-attempted
until an OKAY or ERROR response is received.
12.2.3 AHB write transfers
The AHB/AHB bridge implements posted writes. Writes are accepted with zero wait states if the
bridge is idle. During the AHB write transfer on the slave side the data is buffered in the internal write
FIFO and the transfer is completed on the slave side by always giving an OKAY response. The master
interface requests the bus and performs the write transfer when the master bus is granted. If the burst
transfer crosses the 32-byte write burst boundary, a SPLIT response is given. When the bridge has
written the contents of the FIFO out on the master side, the bridge will allow the master on the slave
side to perform the remaining accesses of the write burst transfer.
12.2.4 Read and write combining
Read and write combining allows the bridge to assemble or split AMBA accesses from the Debug
AHB bus into one or several accesses on the Processor AHB bus. This functionality can improve bus
utilization and also allows cores that have differing AMBA access size restrictions to communicate
with each other. The effects of read and write combining is shown in the table below.
Table 113.Read and write combining
Access on slave interface
Resulting access(es) on master interface
BYTE or HALF-WORD single read
access to any area
Single access of same size
BYTE or HALF-WORD read burst to
prefetchable area
Incremental read burst of same access size as on slave interface, the
length is the same as the number of 32-bit words in the read buffer, but
will not cross the read burst boundary.
BYTE or HALF-WORD read burst to
non-prefetchable area
Incremental read burst of same access size as on slave interface, the
length is the same as the length of the incoming burst. The master
interface will insert BUSY cycles between the sequential accesses.
BYTE or HALF-WORD single write
Single access of same size
BYTE or HALF-WORD write burst
Incremental write burst of same size and length, the maximum length
is the number of 32-bit words in the write FIFO.
Single read access to any area
Single access of same size
Read burst to prefetchable area
Burst of 128-bit up to 32-byte address boundary.
Read burst to non-prefetchable area
Incremental read burst of same access size as on slave interface, the
length is the same as the length of the incoming burst. The master
interface will insert BUSY cycles between the sequential accesses.
Single write
Single write access of same size
Write burst
Burst write of maximum possible size. The bridge will use the maximum size (up to 128-bit) that it can use to empty the writebuffer.
Read and write combining is disabled for accesses to the area 0xF0000000 - 0xFFFFFFFF to prevent
accesses wider than 32 bits to register areas.
GR740-UM-DS, Jun 2016, Version 1.4
117
www.cobham.com/gaisler
GR740
12.2.5 Core latency
The delay incurred when performing an access over the core depends on several parameters such as
operating frequency of the AMBA buses and memory access patterns. Table 114 below shows core
behavior for a single read operation initiated while the bridge is idle.
Table 114.Example of single read
Clock cycle
Core slave side activity
Core master side activity
0
Discovers access and transitions from idle state
Idle
1
Slave side waits for master side, SPLIT response
is given to incoming access, any new incoming
accesses also receive SPLIT responses.
Discovers slave side transition. Master interface output
signals are assigned.
2
3
If bus access is granted, perform address phase. Otherwise wait for bus grant.
Register read data and transition to data ready state.
4
Discovers that read data is ready, assign read
data output and assign SPLIT complete
5
SPLIT complete output is HIGH
6
Typically a wait cycle for the SPLIT:ed master to
be allowed into arbitration. Core waits for master
to return. Other masters receive SPLIT
responses.
7
Master has been allowed into arbitration and performs address phase. Core keeps HREADY high
8
Access data phase. Core has returned to idle
state.
Idle
While the transitions shown in table 114 are simplified they give an accurate view of the core delay. If
the master interface needs to wait for a bus grant or if the read operation receives wait states, these
cycles must be added to the cycle count in the tables.
Table 115 below lists the delays incurred for single operations that traverse the bridge while the bridge
is in its idle state. The second column shows the number of cycles it takes the master side to perform
the requested access, this column assumes that the master slave gets access to the bus immediately
and that each access is completed with zero wait states. The table only includes the delay incurred by
traversing the core. For instance, when the access initiating master reads the core’s prefetch buffer,
each additional read will consume one clock cycle. However, this delay would also have been present
if the master accessed any other slave.
Write accesses are accepted with zero wait states if the bridge is idle, this means that performing a
write to the idle core does not incur any extra latency. However, the core must complete the write
operation on the master side before it can handle a new access on the slave side. If the core has not
transitioned into its idle state, pending the completion of an earlier access, the delay suffered by an
access be longer than what is shown in the tables in this section.
Since the core has been implemented to use AMBA SPLIT responses there will be an additional delay
where, typically, one cycle is required for the arbiter to react to the assertion of HSPLIT and one clock
cycle for the repetition of the address phase. Also, since the core has support for read and/or write
combining, the number of cycles required for the master will change depending on the access size and
length of the incoming burst access.
Table 115.Access latencies
Access
Master acc. cycles Slave cycles
Delay incurred by performing access over core
Single read
3
3
6 * clk
Burst read with prefetch
2 + (burst length)x
4
(6 + burst length)* clk
GR740-UM-DS, Jun 2016, Version 1.4
118
www.cobham.com/gaisler
GR740
Table 115.Access latencies
Access
Master acc. cycles Slave cycles
Delay incurred by performing access over core
Single write
(2)
0
0
Burst writexx
(2 + (burst length))
0
0
xx
x A prefetch
xx The
12.3
operation ends at the address boundary defined by the prefetch buffer’s size
core implements posted writes, the number of cycles taken by the master side can only affect the next access.
Registers
The core does not implement any registers accessible over AMBA AHB or APB.
GR740-UM-DS, Jun 2016, Version 1.4
119
www.cobham.com/gaisler
GR740
13
LEON4 Hardware Debug Support Unit
13.1
Overview
To simplify debugging on target hardware, the LEON4 processor implements a debug mode during
which the pipeline is idle and the processor is controlled through a special debug interface. The
LEON4 Debug Support Unit (DSU4) is used to control the processor during debug mode. The DSU
acts as an AHB slave and can be accessed by all AHB masters on the Debug AHB bus. An external
debug host can therefore access the DSU through several different interfaces.
Debug I/F
LEON4
Processor(s)
Debug Support
Unit
Connection to
system via bridge
AHB Slave I/F
Debug AHB Bus
SpaceWire
Ethernet
JTAG
DEBUG HOST
Figure 11. LEON4/DSU Connection
13.2
Operation
Through the DSU AHB slave interface, any AHB master on the Debug AHB bus can access the processor registers and the contents of the instruction trace buffer. The DSU control registers can be
accessed at any time, while the processor registers, caches and trace buffer can only be accessed when
the processor has entered debug mode. In debug mode, the processor pipeline is held and the processor state can be accessed by the DSU. Entering the debug mode can occur on the following events:
•
executing a breakpoint instruction (ta 1)
•
integer unit hardware breakpoint/watchpoint hit (trap 0xb)
•
rising edge of the external break signal (BREAK)
•
setting the break-now (BN) bit in the DSU control register
•
a trap that would cause the processor to enter error mode
•
occurrence of any, or a selection of traps as defined in the DSU control register
•
after a single-step operation
•
one of the processors in a multiprocessor system has entered the debug mode
•
DSU AHB breakpoint or watchpoint hit
GR740-UM-DS, Jun 2016, Version 1.4
120
www.cobham.com/gaisler
GR740
The debug mode can only be entered when the debug support unit is enabled through an external signal (DSU_EN). For DSU break (DSU_BREAK), and the break-now (BN) bit, to have effect the
Break-on-IU-watchpoint (BW) bit must be set in the DSU control register. This bit is set when
DSU_BREAK is active after reset and should also be set by debug monitor software when initializing
the DSU. When the debug mode is entered, the following actions are taken:
•
PC and nPC are saved in temporary registers (accessible by the debug unit)
•
an output signal (DSU_ACT) is asserted to indicate the debug state
•
the timer units are (optionally) stopped to freeze the LEON timers and watchdog
The instruction that caused the processor to enter debug mode is not executed, and the processor state
is kept unmodified. Execution is resumed by clearing the BN bit in the DSU control register or by deasserting DSU_EN. The timer unit will be re-enabled and execution will continue from the saved PC
and nPC. Debug mode can also be entered after the processor has entered error mode, for instance
when an application has terminated and halted the processor. The error mode can be reset and the processor restarted at any address.
When a processor is in the debug mode, an access to ASI diagnostic area is forwarded to the IU which
performs access with ASI equal to value in the DSU ASI register and address consisting of 20 LSB
bits of the original address.
13.3
AHB Trace Buffer
The AHB trace buffer consists of a circular buffer that stores AHB data transfers. The address, data
and various control signals of the AHB bus are stored and can be read out for later analysis. The trace
buffer is 224 bits wide. The information stored is indicated in the table below:
Table 116.AHB Trace buffer data allocation
Bits
Name
Definition
223:160
Load/Store data
AHB HRDATA/HWDATA(127:64)
159:129
Load/Store data
AHB HRDATA/HWDATA(63:32)
127
AHB breakpoint hit
Set to ‘1’ if a DSU AHB breakpoint hit occurred.
126
-
Not used
125:96
Time tag
DSU time tag counter
95
-
Not used
94:80
RESERVED
RESERVED
79
Hwrite
AHB HWRITE
78:77
Htrans
AHB HTRANS
76:74
Hsize
AHB HSIZE
73:71
Hburst
AHB HBURST
70:67
Hmaster
AHB HMASTER
66
Hmastlock
AHB HMASTLOCK
65:64
Hresp
AHB HRESP
63:32
Load/Store data
AHB HRDATA/HWDATA(31:0)
31:0
Load/Store address
AHB HADDR
In addition to the AHB signals, the low part of the DSU time tag counter is also stored in the trace.
The trace buffer is enabled by setting the enable bit (EN) in the trace control register. Each AHB
transfer is then stored in the buffer in a circular manner. The address to which the next transfer is written is held in the trace buffer index register, and is automatically incremented after each transfer. Tracing is stopped when the EN bit is reset, or when a AHB breakpoint is hit. Tracing is temporarily
GR740-UM-DS, Jun 2016, Version 1.4
121
www.cobham.com/gaisler
GR740
suspended when the processor enters debug mode, unless the trace force bit (TF) in the trace control
register is set. If the trace force bit is set, the trace buffer is activated as long as the enable bit is set.
The force bit is reset if an AHB breakpoint is hit and can also be cleared by software. Note that neither
the trace buffer memory nor the breakpoint registers (see below) can be read/written by software
when the trace buffer is enabled.
The DSU has an internal time tag counter and this counter is frozen when the processor enters debug
mode. When AHB tracing is performed in debug mode (using the trace force bit) it may be desirable
to also enable the time tag counter. This can be done using the timer enable bit (TE). Note that the
time tag is also used for the instruction trace buffer and the timer enable bit should only be set when
using the DSU as an AHB trace buffer only, and not when performing profiling or software debugging. The timer enable bit is reset on the same events as the trace force bit.
The AHB trace buffer is enabled after reset when the DSU_EN signal is HIGH and the BREAK signal
is low.
13.3.1 AHB trace buffer filters
The DSU has filters that can be applied to the AHB trace buffer, breakpoints and watchpoints. These
filters are controlled via the AHB trace buffer filter control and AHB trace buffer filter mask registers.
The fields in these registers allows masking access characteristics such as master, slave, read, write
and address range so that accesses that correspond to the specified mask are not written into the trace
buffer. Address range masking is done using the second AHB breakpoint register set. The values of
the LD and ST fields of this register has no effect on filtering.
13.3.2 AHB statistics
The DSU collects statistics from the traced AHB bus and assert signals that are connected to the
LEON4 statistics unit (L4STAT). The statistics outputs can be filtered by the AHB trace buffer filters,
this is controlled by the Performance counter Filter bit (PF) in the AHB trace buffer filter control register. The DSU can collect data for the events listed in table 117 below.
Table 117.AHB events
Event
Description
Note
idle
HTRANS=IDLE
Active when HTRANS IDLE is driven on the AHB slave inputs and
slave has asserted HREADY.
busy
HTRANS=BUSY
Active when HTRANS BUSY is driven on the AHB slave inputs and
slave has asserted HREADY.
nseq
HTRANS=NONSEQ
Active when HTRANS NONSEQ is driven on the AHB slave inputs
and slave has asserted HREADY.
seq
HTRANS=SEQ
Active when HTRANS SEQUENTIAL is driven on the AHB slave
inputs and slave has asserted HREADY.
read
Read access
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY and the HWRITE input is low.
write
Write access
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY and the HWRITE input is high.
hsize[5:0]
Transfer size
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY.
ws
Wait state
Active when HREADY input to AHB slaves is low and AMBA
response is OKAY.
retry
RETRY response
Active when master receives RETRY response
split
SPLIT response
Active when master receives SPLIT response
GR740-UM-DS, Jun 2016, Version 1.4
122
www.cobham.com/gaisler
GR740
Table 117.AHB events
Event
Description
Note
spdel
SPLIT delay
Active during the time a master waits to be granted access to the bus
after reception of a SPLIT response. The core will only keep track of
one master at a time. This means that when a SPLIT response is
detected, the core will save the master index. This event will then be
active until the same master is re-allowed into bus arbitration and is
granted access to the bus. This also means that the delay measured
will include the time for re-arbitration, delays from other ongoing
transfers and delays resulting from other masters being granted
access to the bus before the SPLIT:ed master is granted again after
receiving SPLIT complete.
If another master receives a SPLIT response while this event is
active, the SPLIT delay for the second master will not be measured.
locked
13.4
Locked access
Active while the HMASTLOCK signal is asserted on the AHB slave
inputs.
Instruction trace buffer
The instruction trace buffer consists of a circular buffer that stores executed instructions. The instruction trace buffer is located in the processor, and read out via the DSU. The trace buffer is 128 bits
wide, the information stored is indicated in the table below:
Table 118.Instruction trace buffer data allocation
Bits
Name
Definition
126
Multi-cycle instruction
Set to ‘1’ on the second instance of a multi-cycle instruction
125:96
Time tag
The value of the DSU time tag counter
95:64
Result or Store address/data
Instruction result, Store address or Store data
63:34
Program counter
Program counter (2 lsb bits removed since they are always zero)
33
Instruction trap
Set to ‘1’ if traced instruction trapped
32
Processor error mode
Set to ‘1’ if the traced instruction caused processor error mode
31:0
Opcode
Instruction opcode
During tracing, one instruction is stored per line in the trace buffer with the exception of for example
atomic load/store instructions, which are entered twice (one for the load and one for the store operation). Bits [63:32] in the buffer correspond to the store address and the loaded data for load instructions. Bit 126 is set for the second entry.
When the processor enters debug mode, tracing is suspended. The trace buffer and the trace buffer
control register can be read and written while the processor is in the debug mode. During the instruction tracing (processor in normal mode), the trace buffer cannot be written and trace buffer control
register 0 can not be written. The traced instructions can optionally be filtered on instruction types.
GR740-UM-DS, Jun 2016, Version 1.4
123
www.cobham.com/gaisler
GR740
Which instructions are traced is defined in the instruction trace buffer control register [31:28], as
defined in the table below:
Table 119.Trace filter operation
Trace filter
Instructions traced
0x0
All instructions
0x1
SPARC Format 2 instructions
0x2
Control-flow changes. All Call, branch and trap instructions including branch targets
0x4
SPARC Format 1 instructions (CALL)
0x8
SPARC Format 3 instructions except LOAD or STORE
0xC
SPARC Format 3 LOAD or STORE instructions
0xD
SPARC Format 3 LOAD or STORE instructions to alternate space
0xE
SPARC Format 3 LOAD or STORE instructions to alternate space 0x80 - 0xFF
Load and stores to ASI 0x80 - 0xFF do not cause operations on the on-chip bus and can be used to
implement software trace points.
The instruction trace buffer is enabled after reset when the DSU_EN signal is HIGH and the BREAK
signal is low.
13.5
DSU memory map
The DSU memory map can be seen in table 120 below. In a multiprocessor systems, the register map
is duplicated and address bits 27 - 24 are used to index the processor.
Table 120.DSU memory map
Address offset
Register
0x000000
DSU control register
0x000008
Time tag counter
0x000020
Break and Single Step register
0x000024
Debug Mode Mask register
0x000040
AHB trace buffer control register
0x000044
AHB trace buffer index register
0x000048
AHB trace buffer filter control register
0x00004c
AHB trace buffer filter mask register
0x000050
AHB breakpoint address 1
0x000054
AHB mask register 1
0x000058
AHB breakpoint address 2
0x00005c
AHB mask register 2
0x000070
Instruction count register
0x000080
AHB watchpoint control register
0x000090 - 0x00009C
AHB watchpoint 1 data registers
0x0000A0 - 0x0000AC
AHB watchpoint 1 mask registers
0x0000B0 - 0x0000BC
AHB watchpoint 2 data registers
0x0000C0 - 0x0000CC
AHB watchpoint 2 mask registers
0x100000 - 0x10FFFF
Instruction trace buffer (..0: Trace bits 127 - 96, ..4: Trace bits 95 - 64,
..8: Trace bits 63 - 32, ..C : Trace bits 31 - 0)
0x110000
Instruction Trace buffer control register 0
0x110004
Instruction Trace buffer control register 1
0x200000 - 0x210000
AHB trace buffer (..0: Trace bits 127 - 96, ..4: Trace bits 95 - 64,
..8: Trace bits 63 - 32, ..C : Trace bits 31 - 0)
GR740-UM-DS, Jun 2016, Version 1.4
124
www.cobham.com/gaisler
GR740
Table 120.DSU memory map
Address offset
Register
0x300000 - 0x3007FC
IU register file.
The addresses of the IU registers depends on how many register windows has been
implemented:
%on: 0x300000 + (((psr.cwp * 64) + 32 + n*4) mod (NWINDOWS*64))
%ln: 0x300000 + (((psr.cwp * 64) + 64 + n*4) mod (NWINDOWS*64))
%in: 0x300000 + (((psr.cwp * 64) + 96 + n*4) mod (NWINDOWS*64))
%gn: 0x300000 + (NWINDOWS*64) + n*4
%fn: 0x301000 + n*4
0x300800 - 0x300FFC
IU register file check bits (LEON4FT only)
0x301000 - 0x30107C
FPU register file
0x400000
Y register
0x400004
PSR register
0x400008
WIM register
0x40000C
TBR register
0x400010
PC register
0x400014
NPC register
0x400018
FSR register
0x40001C
CPSR register
0x400020
DSU trap register
0x400024
DSU ASI register
0x400040 - 0x40007C
ASR16 - ASR31
0x700000 - 0x7FFFFC
ASI diagnostic access (ASI = value in DSU ASI register, address = address[19:0])
ASI = 0x9 : Local instruction RAM
ASI = 0xB : Local data RAM
ASI = 0xC : Instruction cache tags
ASI = 0xD : Instruction cache data
ASI = 0xE : Data cache tags
ASI = 0xF : Data cache data
ASI = 0x1E : Separate snoop tags
GR740-UM-DS, Jun 2016, Version 1.4
125
www.cobham.com/gaisler
GR740
13.6
DSU registers
13.6.1 DSU control register
Table 121. 0x000000- CTRL - DSU control register
31
12 11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
PW HL PE EB EE DM BZ BX BS BW BE TE
0
0
0
r
r
rw wc
0
*
*
r
r
*
r
*
0
*
*
*
rw rw rw rw rw rw
31: 12
Reserved
11
Power down (PW) - Returns ‘1’ when processor is in power-down mode.
10
Processor halt (HL) - Returns ‘1’ on read when processor is halted. If the processor is in debug
mode, setting this bit will put the processor in halt mode.
9
Processor error mode (PE) - returns ‘1’ on read when processor is in error mode, else ‘0’. If written
with ‘1’, it will clear the error and halt mode.
8
External Break (EB) - Value of the external BREAK signal
7
External Enable (EE) - Value of the external DSU_EN signal
6
Debug mode (DM) - Indicates when the processor has entered debug mode.
5
Break on error traps (BZ) - if set, will force the processor into debug mode on all except the following traps: priviledged_instruction, fpu_disabled, window_overflow, window_underflow, asynchronous_interrupt, ticc_trap.
BZ is reset to the value of the external BREAK signal.
4
Break on trap (BX) - if set, will force the processor into debug mode when any trap occurs.
BX is reset to the value of the external BREAK signal.
3
Break on S/W breakpoint (BS) - if set, debug mode will be forced when an breakpoint instruction (ta
1) is executed.
2
Break on IU watchpoint (BW) - if set, debug mode will be forced on a IU watchpoint (trap 0xb).
BW is reset to the value of the external BREAK signal.
1
Break on error (BE) - if set, will force the processor to debug mode when the processor would have
entered error condition (trap in trap).
BE is reset to the value of the external BREAK signal.
0
Trace enable (TE) - Enables instruction tracing. If set the instructions will be stored in the trace
buffer. Remains set when then processor enters debug or error mode.
TE is reset to ’1’ when external signal BREAK=LOW, otherwise TE is reset to 0.
GR740-UM-DS, Jun 2016, Version 1.4
126
www.cobham.com/gaisler
GR740
13.6.2 Time tag counter register
The trace buffer time tag counter is incremented each clock as long as the processor is running. The
counter is stopped when the processor enters debug mode (unless the timer enable bit in the AHB
trace buffer control register is set), and restarted when execution is resumed. The value of this register
is used as time tag in the instruction and AHB trace buffers. The same time source is used for the processors’ internal up-counters.
Table 122. 0x000008 - DTTC - DSU time tag counter register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
9
8
7
6
5
4
3
2
1
0
TIMETAG
0
rw
31: 0
DSU Time Tag Value (TIMETAG)
13.6.3 DSU Break and Single Step register
This register is used to break or single step the processors. This register controls all processors in a
multi-processor system, and is only accessible in the DSU memory map of processor 0.
Table 123. 0x000020 - BRSS - DSU break and single step register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RESERVED
9
8
7
6
5
4
3
2
1
SS[3:0]
RESERVED
0
0
0
*
r
rw
r
rw
0
BN[3:0]
31: 17
RESERVED
19: 16
Single step (SSx) - if set, the processor x will execute one instruction and return to debug mode. The
bit remains set after the processor goes into the debug mode.
15: 4
RESERVED
3:0
Break now (BNx) -Force processor x into debug mode if the Break on watchpoint (BW) bit in the
processors DSU control register is set. If cleared, the processor x will resume execution.
The reset value of this field is taken from the external BREAK signal.
13.6.4 DSU Debug Mode Mask Register
When one of the processors in a multiprocessor LEON4 system enters the debug mode the value of
the DSU Debug Mode Mask register determines if the other processors are forced in the debug mode.
This register controls all processors in a multi-processor system, and is only accessible in the DSU
memory map of processor 0.
Table 124. 0x000024 - DBGM - DSU debug mode mask register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RESERVED
9
8
7
6
5
4
3
2
1
DM[3:0]
RESERVED
0
0
0
0
r
rw
r
rw
0
ED[3:0]
31: 17
RESERVED
19: 16
Debug mode mask (DMx) - If set, the corresponding processor will not be able to force running processors into debug mode even if it enters debug mode.
15: 4
RESERVED
3:0
Enter debug mode (EDx) - Force processor x into debug mode if any of processors in a multiprocessor system enters the debug mode. If 0, the processor x will not enter the debug mode.
GR740-UM-DS, Jun 2016, Version 1.4
127
www.cobham.com/gaisler
GR740
13.6.5 DSU trap register
The DSU trap register is a read-only register that indicates which SPARC trap type that caused the
processor to enter debug mode. When debug mode is force by setting the BN bit in the DSU control
register, the trap type will be 0xb (hardware watchpoint trap).
Table 125. 0x400020 - DTR - DSU trap register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
9
8
7
6
5
4
3
2
1
0
RESERVED
EM
TRAPTYPE
RESERVED
0
NR
NR
0
r
r
r
r
31: 13
RESERVED
12
Error mode (EM) - Set if the trap would have cause the processor to enter error mode.
11: 4
Trap type (TRAPTYPE) - 8-bit SPARC trap type
3:0
Read as 0x0
13.6.6 DSU ASI register
The DSU can perform diagnostic accesses to different ASI areas. The value in the ASI diagnostic
access register is used as ASI while the address is supplied from the DSU memory area when performing an access at offset 0x700000.
Table 126. 0x400024 - DASI- DSU ASI diagnostic access register
31
8
0
ASI
0
NR
r
rw
31: 8
RESERVED
7: 0
ASI (ASI) - ASI to be used on diagnostic ASI access
GR740-UM-DS, Jun 2016, Version 1.4
7
RESERVED
128
www.cobham.com/gaisler
GR740
13.6.7 AHB Trace buffer control register
The AHB trace buffer is controlled by the AHB trace buffer control register:
Table 127. 0x000040 - ATBC - AHB trace buffer control register
31
24 23
16 15
RESERVED
9
8
7
6
5
DCNT
RESERVED
DF SF TE TF
0
0
0
0
r
rw
r
rw rw rw rw
0
0
0
4
3
BW
2
1
0
BR DM EN
0x10
0
r
0
*
rw rw rw
31: 24
RESERVED
23: 16
Trace buffer delay counter (DCNT) - Specifies the number of lines that should be written in the trace
buffer befiore entering debug mode after a AHB break/watchpoint has been hit.
15: 9
RESERVED
8
Enable Debug Mode Timer Freeze (DF) - The time tag counter keeps counting in debug mode when
at least one of the processors has the internal timer enabled. If this bit is set to ‘1’ then the time tag
counter is frozen when the processors have entered debug mode.
7
Sample Force (SF) - If this bit is written to ‘1’ it will have the same effect on the AHB trace buffer as
if HREADY was asserted on the bus at the same time as a sequential or non-sequential transfer is
made. This means that setting this bit to ‘1’ will cause the values in the trace buffer’s sample registers to be written into the trace buffer, and new values will be sampled into the registers. This bit will
automatically be cleared after one clock cycle.
Writing to the trace buffer still requires that the trace buffer is enabled (EN bit set to ‘1’) and that the
CPU is not in debug mode or that tracing is forced (TF bit set to ‘1’). This functionality is primarily
of interest if the Processor AHB bus appears to have frozen.
6
Timer enable (TE) - Activates time tag counter also in debug mode. Note that this activates the same
timer source as used for the processor up-counters described in section 6.10.3.
5
Trace force (TF) - Activates trace buffer also in debug mode.
4: 3
Bus width (BW) - This value corresponds to log2(Supported bus width / 32). Value is 2.
2
Break (BR) - If set, the processor will be put in debug mode when AHB trace buffer stops due to
AHB breakpoint hit.
1
Delay counter mode (DM) - Indicates that the trace buffer is in delay counter mode.
0
Trace enable (EN) - Enables the trace buffer.
The reset value of this field is 1 when the external signal BREAK is low, otherwise 0.
13.6.8 AHB trace buffer index register
The AHB trace buffer index register contains the address of the next trace line to be written.
Table 128. 0x000044 - ATBI - AHB trace buffer index register
31
12 11
4
3
0
RESERVED
INDEX
RESERVED
0
NR
0
r
rw
r
31: 12
RESERVED
11: 4
Trace buffer index counter (INDEX) - Address of next trace line to be written.
3: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
129
www.cobham.com/gaisler
GR740
13.6.9 AHB trace buffer filter control register
Table 129. 0x000048 - ATBFC - AHB trace buffer filter control register
31
14 13 12 11 10
RESERVED
WPF
R
0
0
r
rw
9
8
7
4
3
2
1
0
BPF
RESERVED
PF AF FR FW
0
0
0
0
r
rw
r
rw rw rw rw
0
0
0
31: 14
RESERVED
13: 12
AHB watchpoint filtering (WPF) - Bit 13 of this field applies to AHB watchpoint 2 and bit 12
applies to AHB watchpoint 1. If the WPF bit for a watchpoint is set to ‘1’ then the watchpoint will
not trigger unless the access also passes through the filter. This functionality can be used to, for
example, set a AHB watchpoint that only triggers if a specified master performs an access to a specified slave.
11: 10
RESERVED
9: 8
AHB breakpoint filtering (BPF) - Bit 9 of this field applies to AHB breakpoint 2 and bit 8 applies to
AHB breakpoint 1. If the BPF bit for a breakpoint is set to ‘1’ then the breakpoint will not trigger
unless the access also passes through the filter. This functionality can be used to, for instance, set a
AHB breakpoint that only triggers if a specified master performs an access to a specified slave. Note
that if a AHB breakpoint is coupled with an AHB watchpoint then the setting of the corresponding
bit in this field has no effect.
7: 4
RESERVED
3
Performance counter Filter (PF) - If this bit is set to ‘1’, the cores performance counter (statistical)
outputs will be filtered using the same filter settings as used for the trace buffer. If a filter inhibits a
write to the trace buffer, setting this bit to ‘1’ will cause the same filter setting to inhibit the pulse on
the statistical output.
2
Address Filter (AF) - If this bit is set to ‘1’, only the address range defined by AHB trace buffer
breakpoint 2’s address and mask will be included in the trace buffer.
1
Filter Reads (FR) - If this bit is set to ‘1’, read accesses will not be included in the trace buffer.
0
Filter Writes (FW) - If this bit is set to ‘1’, write accesses will not be included in the trace buffer.
13.6.10 AHB trace buffer filter mask register
Table 130. 0x00004C - ATBFM - AHB trace buffer filter mask register
31
16 15
SMASK[15:0]
0
MMASK[15:0]
0
0
rw
rw
31: 16
Slave Mask (SMASK) - If SMASK[n] is set to ‘1’, the trace buffer will not save accesses performed
to slave n. Note that this field has more bits than there are slaves connected to the Processor AHB
bus.
15: 0
Master Mask (MMASK) - If MMASK[n] is set to ‘1’, the trace buffer will not save accesses performed by master n. Note that this field has more bits than there are masters connected to the Processor AHB bus.
13.6.11 AHB trace buffer breakpoint registers
The DSU contains two breakpoint registers for matching AHB addresses. A breakpoint hit is used to
freeze the trace buffer by automatically clearing the enable bit. Freezing can be delayed by programming the DCNT field in the trace buffer control register to a non-zero value. In this case, the DCNT
GR740-UM-DS, Jun 2016, Version 1.4
130
www.cobham.com/gaisler
GR740
value will be decremented for each additional trace until it reaches zero, after which the trace buffer is
frozen. A mask register is associated with each breakpoint, allowing breaking on a block of addresses.
Only address bits with the corresponding mask bit set to ‘1’ are compared during breakpoint detection. To break on AHB load or store accesses, the LD and/or ST bits should be set.
Table 131. 0x000050, 0x000058 - ATBBA - AHB trace buffer break address registers
31
2
0
RES
NR
0
rw
r
31: 2
Break point address (BADDR) - Bits 31:2 of breakpoint address
1
RESERVED
0
1
BADDR[31:2]
Table 132. 0x000054, 0x00005C - ATBBM - AHB trace buffer break mask registers
31
2
BMASK[31:2]
0
NR
0
rw
rw rw
31: 2
Breakpoint mask (BMASK) - See description above tables.
1
Load (LD) - Break on data load address
0
Store (ST) - Break on data store address
GR740-UM-DS, Jun 2016, Version 1.4
1
LD ST
131
0
www.cobham.com/gaisler
GR740
13.6.12 Instruction count register
The DSU contains an instruction count register to allow profiling of application, or generation of
debug mode after a certain clocks or instructions. The instruction count register consists of a 29-bit
down-counter, which is decremented on either each clock (IC=0) or on each executed instruction
(IC=1). In profiling mode (PE=1), the counter will set to all ones after an underflow without generating a processor break. In this mode, the counter can be periodically polled and statistics can be formed
on CPI (clocks per instructions). In non-profiling mode (PE=0), the processor will be put in debug
mode when the counter underflows. This allows a debug tool such as GRMON to execute a defined
number of instructions, or for a defined number of clocks.
Table 133. 0x000070 - ICNT - Instruction trace count register
31 30 29 28
0
CE IC PE
0
0
ICOUNT[28:0]
0
NR
rw rw rw
rw
31
Counter Enable (CE) - Counter enable
30
Instruction Count (IC) - Instruction (1) or clock (0) counting
29
Profiling Enable (PE) - Profiling enable
28: 0
Instruction count (ICOUNT) - Instruction count
GR740-UM-DS, Jun 2016, Version 1.4
132
www.cobham.com/gaisler
GR740
13.6.13 AHB watchpoint control register
The DSU has two AHB watchpoints that can be used to freeze the AHB tracebuffer, or put the processor in debug mode, when a specified data pattern occurs on the AMBA bus. These watchpoints can
also be coupled with the two AHB breakpoints so that a watchpoint will not trigger unless the AHB
breakpoint is triggered. This also means that when a watchpoint is coupled with an AHB breakpoint,
the breakpoint will not cause an AHB tracebuffer freeze, or put the processor(s), in debug mode
unless also the watchpoint is triggered.
The bus data lines are taken through a register stage before being compared with the watchpoint registers in the DSU. Data watchpoints have one extra cycle of latency compared to a
AHB breakpoint due to this pipelining.
Table 134. 0x000080 - AHBWPC - AHB watchpoint control register
31
7
RESERVED
6
5
4
3
IN CP EN R
0
2
1
0
IN CP EN
0
0
0
0
0
r
rw rw rw
r
rw rw rw
0
0
31: 7
RESERVED
6
Invert (IN) - Invert AHB watchpoint 2. If this bit is set the watchpoint will trigger if data on the AHB
bus does NOT match the specified data pattern (typically only usable if the watchpoint has been coupled with an address by setting the CP field).
5
Couple (CP) - Couple AHB watchpoint 2 with AHB breakpoint 1
4
Enable (EN) - Enable AHB watchpoint 2
3
RESERVED
2
Invert (IN) - Invert AHB watchpoint 1. If this bit is set the watchpoint will trigger if data on the AHB
bus does NOT match the specified data pattern (typically only usable if the watchpoint has been coupled with an address by setting the CP field).
1
Couple (CP) - Couple AHB watchpoint 1 with AHB breakpoint 1
0
Enable (EN) - Enable AHB watchpoint 1
GR740-UM-DS, Jun 2016, Version 1.4
133
www.cobham.com/gaisler
GR740
13.6.14 AHB watchpoint data and mask registers
The AHB watchpoint data and mask registers specify the data pattern for an AHB watchpoint. A
watchpoint hit is used to freeze the trace buffer by automatically clearing the enable bit. A watchpoint
hit can also be used to force the processors into debug mode.
A mask register is associated with each data register. Only data bits with the corresponding mask bit
set to ‘1’ are compared during watchpoint detection.
Table 135. 0x000090 to 0x00009C, 0x0000B0 to 0x0000BC- AHBWPD0-7 - AHB watchpoint data registers
31
0
DATA[127-n*32 : 96-n*32]
NR
rw
31: 0
AHB watchpoint data (DATA) - Specifies the data pattern of one word for an AHB watchpoint. The
lower part of the register address specifies with part of the bus that the register value will be compared against: Offset 0x0 specifies the data value for AHB bus bits 127:96, 0x4 for bits 95:64, 0x8
for 63:32 and offset 0xC for bits 31:0.
Table 136. 0x0000A0 to 0x0000AC, 0x0000C0 to 0x0000CC- AHBWPM0-7 - AHB watchpoint mask registers
31
0
MASK[127-n*32 : 96-n*32]
NR
rw
31: 0
AHB watchpoint mask (MASK) - Specifies the mask to select bits for comparison out of one word
for an AHB watchpoint. The lower part of the register address specifies with part of the bus that the
register value will be compared against: Offset 0x0 specifies the data value for AHB bus bits 127:96,
0x4 for bits 95:64, 0x8 for 63:32 and offset 0xC for bits 31:0.
GR740-UM-DS, Jun 2016, Version 1.4
134
www.cobham.com/gaisler
GR740
13.6.15 Instruction trace buffer control register 0
The instruction trace control register contains filter configuration and a pointer that indicates the next
line of the instruction trace buffer to be written.
Table 137. 0x110000 - ITBC0 - Instruction trace buffer control register 0
31
28 27
9
8
0
TFILT
RESERVED
ITPOINTER
0
0
NR
rw
r
rw
31: 28
Trace filter configuration (TFILT) - See table 119.
27: 9
RESERVED
15: 0
Instruction trace buffer pointer (ITPOINTER) - Indicates the next line of the instruction trace buffer
to be written
13.6.16 Instruction trace buffer control register 1
The instruction trace control register 1 contains settings used for trace buffer overflow detection. This
register can be written while the processor is running.
Table 138. 0x110004 - ITBC1 - Instruction trace buffer control register 1
31
28 27 26
RESERVED
W
O
24 23 22
TLIM
0
T
O
V
RESERVED
0
0
0
0
0
r
rw
rw
rw
r
31: 28
RESERVED
27
Watchpoint on overflow (WO) - If this bit is set, and Break on iu watchpoint (BW) is enabled in the
DSU control register, then a watchpoint will be inserted when a trace overflow is detected (TOV
field in this register gets set).
26: 24
Trace Limit (TLIM) - TLIM is compared with the top bits of ITBC0.ITPOINTER to generate the
value in the TOV field below.
23
Trace Overflow (TOV) - Gets set to ‘1’ when the DSU detects that TLIM equals the top three bits of
ITPOINTER.
22: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
135
www.cobham.com/gaisler
GR740
14
JTAG Debug Link with AHB Master Interface
14.1
Overview
The JTAG debug interface provides access to the Debug AHB bus through JTAG. The JTAG debug
interface implements a simple protocol which translates JTAG instructions to AHB transfers. Through
this link, a read or write transfer can be generated to any address on the AHB bus.
TDI
TCK
TMS
JTAG TAP
Controller
TDO
JTAG Communication
Interface
AHB master interface
Debug AHB bus
Figure 12. JTAG Debug link block diagram
The JTAG debug interface will, together with all other cores on the Debug AHB bus, be gated off
when the Debug AHB bus is disabled via the external DSU_EN signal.
14.2
Operation
14.2.1 Transmission protocol
The JTAG Debug link decodes two JTAG instructions and implements two JTAG data registers: the
command/address register and data register. A read access is initiated by shifting in a command consisting of read/write bit, AHB access size and AHB address into the command/address register. The
AHB read access is performed and data is ready to be shifted out of the data register. Write access is
performed by shifting in command, AHB size and AHB address into the command/data register followed by shifting in write data into the data register. Sequential transfers can be performed by shifting
in command and address for the transfer start address and shifting in SEQ bit in data register for following accesses. The SEQ bit will increment the AHB address for the subsequent access. Sequential
transfers should not cross a 1 kB boundary. Sequential transfers are always word based.
Table 139. JTAG debug link Command/Address register
34 33 32 31
W
0
SIZE
34
AHB ADDRESS
Write (W) - ‘0’ - read transfer, ‘1’ - write transfer
33 32
AHB transfer size - “00” - byte, “01” - half-word, “10” - word, “11”- reserved
31 30
AHB address
Table 140. JTAG debug link Data register
32
31
0
SEQ
AHB DATA
32
Sequential transfer (SEQ) - If ‘1’ is shifted in this bit position when read data is shifted out or write
data shifted in, the subsequent transfer will be to next word address. When read out from the device,
this bit is ‘1’ if the AHB access has completed and ‘0’ otherwise.
31 30
AHB Data - AHB write/read data. For byte and half-word transfers data is aligned according to bigendian order where data with address offset 0 data is placed in MSB bits.
GR740-UM-DS, Jun 2016, Version 1.4
136
www.cobham.com/gaisler
GR740
The core will signal AHB access completion by setting bit 32 of the data register. A debug host can
look at bit 32 of the received data to determine if the access was successful. If bit 32 is ‘1’ the access
completed and the data is valid. If bit 32 is ‘0’, the AHB access was not finished when the host started
to read data. In this case the host can repeat the read of the data register until bit 32 is set to ‘1’, signaling that the data is valid and that the AMBA AHB access has completed.
It should be noted that while bit 32 returns ‘0’, new data will not be shifted into the data register. The
debug host should therefore inspect bit 32 when shifting in data for a sequential AHB access to see if
the previous command has completed. If bit 32 is ‘0’, the read data is not valid and the command just
shifted in has been dropped by the core.
14.3
Registers
The core does not implement any registers mapped in the AMBA AHB or APB address space.
GR740-UM-DS, Jun 2016, Version 1.4
137
www.cobham.com/gaisler
GR740
15
SpaceWire codec with AHB host Interface and RMAP target
15.1
Overview
The SpaceWire core provides an interface between the AHB bus and a SpaceWire network. It implements the SpaceWire standard [SPW] with the protocol identification extension [SPWID]. The
Remote Memory Access Protocol (RMAP) target implements [RMAP].
The SpaceWire interface is configured through a set of registers accessed through an APB interface.
Data is transferred through DMA channels using an AHB master interface.
The GRSPW2 SpaceWire core is located on the Debug AHB bus and has an RMAP target that is
enabled after system reset. The core APB interface is also available on the Debug AHB bus but cannot
be accessed by the processors since the bridge connecting the Debug AHB bus to the Processor AHB
bus is uni-directional. The core on the Debug AHB bus thus provides a SpaceWire debug link that can
be used to access all parts of the system. The system’s main SpaceWire links are provided through the
SpaceWire router, see section 17.
The SpaceWire debug interface will, together with all other cores on the Debug AHB bus, be gated off
when the Debug AHB bus is disabled via the external DSU_EN signal.
TXCLK
D(1:0)
TRANSMITTER
S(1:0)
LINKINTERFACE
FSM
SEND
FSM
TRANSMITTER
FSM
RMAP
TRANSMITTER
TRANSMITTER
DMA ENGINE
AHB
MASTER INTERFACE
RECEIVER
DMA ENGINE
D0
RMAP
RECEIVER
RECEIVER
AHB FIFO
N-CHAR
FIFO
RECEIVER DATA
PARALLELIZATION
REGISTERS
APB
INTERFACE
D
PHY
S0
RECEIVER0
DV
Figure 13. Block diagram
15.2
Operation
15.2.1 Overview
The main sub-blocks of the core are the link interface, the RMAP target and the AMBA interface. A
block diagram of the internal structure can be found in figure 13.
The link interface consists of the receiver, transmitter and the link interface FSM. They handle communication on the SpaceWire network. The PHY block provides a common interface for the receiver
to the four different data recovery schemes and is external to this core. The AMBA interface consists
of the DMA engines, the AHB master interface and the APB interface. The link interface provides
FIFO interfaces to the DMA engines. These FIFOs are used to transfer N-Chars between the AMBA
and SpaceWire domains during reception and transmission.
The RMAP target handles incoming packets which are determined to be RMAP commands instead of
the receiver DMA engine. The RMAP command is decoded and if it is valid, the operation is per-
GR740-UM-DS, Jun 2016, Version 1.4
138
www.cobham.com/gaisler
GR740
formed on the AHB bus. If a reply was requested it is automatically transmitted back to the source by
the RMAP transmitter.
15.2.2 Protocol support
The core only accepts packets with a valid destination address in the first received byte. Packets with
address mismatch will be silently discarded (except in promiscuous mode which is covered in section
15.5.10).
The second byte is sometimes interpreted as a protocol ID and described hereafter. The RMAP protocol (ID=0x1) is the only protocol handled separately in hardware while other packets are stored to a
DMA channel. If the RMAP target is present and enabled all RMAP commands will be processed,
executed and replied automatically in hardware. Otherwise RMAP commands are stored to a DMA
channel in the same way as other packets. RMAP replies are always stored to a DMA channel. More
information on the RMAP protocol support is found in section 15.7. When the RMAP target is not
present or disabled, there is no need to include a protocol ID in the packets and the data can start
immediately after the address.
All packets arriving with the extended protocol ID (0x00) are stored to a DMA channel. This means
that the hardware RMAP target will not work if the incoming RMAP packets use the extended protocol ID. Note also that packets with the reserved extended protocol identifier (ID = 0x000000) are not
ignored by the core. It is up to the client receiving the packets to ignore them.
When transmitting packets, the address and protocol-ID fields must be included in the buffers from
where data is fetched. They are not automatically added by the core.
Figure 14 shows the packet types accepted by the core. The core also allows reception and transmission with extended protocol identifiers but without support for RMAP CRC calculations and the
RMAP target.
Addr ProtID D0
D1
D2
D3
..
Dn-2 Dn-1 EOP
D0
D2
D3
D4
..
Dm-2 Dm-1 EOP
Addr
D1
Figure 14. The SpaceWire packet types supported by the core.
15.3
Link interface
The link interface handles the communication on the SpaceWire network and consists of a transmitter,
receiver, a FSM and FIFO interfaces. An overview of the architecture is found in figure 13.
15.3.1 Link interface FSM
The FSM controls the link interface (a more detailed description is found in the SpaceWire standard).
The low-level protocol handling (the signal and character level of the SpaceWire standard) is handled
by the transmitter and receiver while the FSM handles the exchange level.
The link interface FSM is controlled through the control register. The link can be disabled through the
link disable bit, which depending on the current state, either prevents the link interface from reaching
the started state or forces it to the error-reset state. When the link is not disabled, the link interface
FSM is allowed to enter the started state when either the link start bit is set or when a NULL character
has been received and the autostart bit is set.
The current state of the link interface determines which type of characters are allowed to be transmitted which together with the requests made from the host interfaces determine what character will be
sent.
GR740-UM-DS, Jun 2016, Version 1.4
139
www.cobham.com/gaisler
GR740
Time-codes are sent when the FSM is in the run-state and a request is made through the time-interface
(described in section 15.4).
When the link interface is in the connecting- or run-state it is allowed to send FCTs. FCTs are sent
automatically by the link interface when possible. This is done based on the maximum value of 56 for
the outstanding credit counter and the currently free space in the receiver N-Char FIFO. FCTs are sent
as long as the outstanding counter is less than or equal to 48 and there are at least 8 more empty FIFO
entries than the counter value.
N-Chars are sent in the run-state when they are available from the transmitter FIFO and there are
credits available. NULLs are sent when no other character transmission is requested or the FSM is in
a state where no other transmissions are allowed.
The credit counter (incoming credits) is automatically increased when FCTs are received and
decreased when N-Chars are transmitted. Received N-Chars are stored to the receiver N-Char FIFO
for further handling by the DMA interface. Received Time-codes are handled by the time-interface.
15.3.2 Transmitter
The state of the FSM, credit counters, requests from the time-interface and requests from the DMAinterface are used to decide the next character to be transmitted. The type of character and the character itself (for N-Chars and Time-codes) to be transmitted are presented to the low-level transmitter
which is located in a separate clock-domain.
This is done because one usually wants to run the SpaceWire link on a different frequency than the
host system clock. The core has a separate clock input which is used to generate the transmitter clock.
Since the transmitter often runs on high frequency clocks (> 100 MHz) as much logic as possible has
been placed in the system clock domain to minimize power consumption and timing issues.
The transmitter logic in the host clock domain decides what character to send next and sets the proper
control signal and presents any needed character to the low-level transmitter as shown in figure 15.
The transmitter sends the requested characters and generates parity and control bits as needed. If no
requests are made from the host domain, NULLs are sent as long as the transmitter is enabled. Most of
the signal and character levels of the SpaceWire standard is handled in the transmitter. External LVDS
drivers are needed for the data and strobe signals.
D
S
Send Time-code
Send FCT
Send NChar
Time-code[7:0]
NChar[8:0]
Transmitter
Transmitter Clock Domain
Host Clock Domain
Figure 15. Schematic of the link interface transmitter.
A transmission FSM reads N-Chars for transmission from the transmitter FIFO. It is given packet
lengths from the DMA interface and appends EOPs/EEPs and RMAP CRC values if requested. When
it is finished with a packet the DMA interface is notified and a new packet length value is given.
15.3.3 Receiver
The receiver detects connections from other nodes and receives characters as a bit stream recovered
from the data and strobe signals by the PHY module which presents it as a data and data-valid signal.
Both the receiver and PHY are located in a separate clock domain which runs on a clock generated by
the PHY.
GR740-UM-DS, Jun 2016, Version 1.4
140
www.cobham.com/gaisler
GR740
The receiver is activated as soon as the link interface leaves the error reset state. Then after a NULL is
received it can start receiving any characters. It detects parity, escape and credit errors which causes
the link interface to enter the error reset state. Disconnections are handled in the link interface part in
the tx clock domain because no receiver clock is available when disconnected.
Received Characters are flagged to the host domain and the data is presented in parallel form. The
interface to the host domain is shown in figure 16. L-Chars are the handled automatically by the host
domain link interface part while all N-Chars are stored in the receiver FIFO for further handling. If
two or more consecutive EOPs/EEPs are received all but the first one are discarded.
D
Receiver
DV
Got Time-code
Got FCT
Got EOP
Got EEP
Got NChar
Time-code[7:0]
NChar[7:0]
Receiver Clock Domain
Host Clock Domain
Figure 16. Schematic of the link interface receiver.
15.4
Time-Code distribution
Time-codes are control codes that consists of two control flags (bits 7:6) and a time value (bits 5:0),
and they are used to distribute time over the SpaceWire network. The current time value (value of latest received or transmitted time-code), and control flags, can be read from the Time-code register
(TC).
15.4.1 Receiving time-codes
When a control code is received, and either the control flags (bits 7:6) have value “00”, or control flag
filtering is disabled (CTRL.TF bit set to 0), then the received control code is considered to be a TimeCode. If Time-Code reception is enabled (CTRL.TR bit set to 1) then the received time value is stored
in the TC.TIMECNT field. If the received time value equals TC.TIMECNT+1 (modulo 64), then the
Time-Code is considered valid.
When a valid Time-Code is received, in addition to the time value being updated, the received control
flags are stored to the TC.TCTRL field. Also, when a valid Time-Code is received, the STS.TO bit is
set to 1, and an AMBA interrupt is generated if the CTRL.IE bit and CTRL.TQ bit are both set to 1.
15.4.2 Transmitting time-codes
Time-Codes can be transmitted through the AMBA APB registers. In order to send a Time-code,
Time-Code transmission must be enabled by setting the CTRL.TT bit to 1. To transmit a time-code
through the register interface the CTRL.TI bit should be written to 1. When the bit is written the current time value (TC.TIMECNT field) is incremented, and a Time-Code consisting of the new time
value together with the current control flags (TC.TCTRL field) is sent. The CTRL.TI bit will stay
high until the Time-Code has been transmitted. If Time-Code transmission is disabled, writing the
CTRL.TI bit has no effect.
Note that the link interface must be in run-state in order to be able to send a Time-Code.
15.5
Receiver DMA channels
The receiver DMA engine handles reception of data from the SpaceWire network to different DMA
channels.
GR740-UM-DS, Jun 2016, Version 1.4
141
www.cobham.com/gaisler
GR740
15.5.1 Address comparison and channel selection
Packets are received to different channels based on the address and whether a channel is enabled or
not. When the receiver N-Char FIFO contains one or more characters, N-Chars are read by the
receiver DMA engine. The first character is interpreted as the logical address and is compared with
the addresses of each channel starting from 0. The packet will be stored to the first channel with an
matching address. The complete packet including address and protocol ID but excluding EOP/EEP is
stored to the memory address pointed to by the descriptors (explained later in this section) of the
channel.
Each SpaceWire address register has a corresponding mask register. Only bits at an index containing a
zero in the corresponding mask register are compared. This way a DMA channel can accept a range of
addresses. There is a default address register which is used for address checking in all implemented
DMA channels that do not have separate addressing enabled and for RMAP commands in the RMAP
target. With separate addressing enabled the DMA channels’ own address/mask register pair is used
instead.
If an RMAP command is received it is only handled by the target if the default address register
(including mask) matches the received address. Otherwise the packet will be stored to a DMA channel
if one or more of them has a matching address. If the address does not match neither the default
address nor one of the DMA channels’ separate register, the packet is still handled by the RMAP target if enabled since it has to return the invalid address error code. The packet is only discarded (up to
and including the next EOP/EEP) if an address match cannot be found and the RMAP target is disabled.
Packets, other than RMAP commands, that do not match neither the default address register nor the
DMA channels’ address register will be discarded. Figure 17 shows a flowchart of packet reception.
At least 2 non EOP/EEP N-Chars needs to be received for a packet to be stored to the DMA channel
unless the promiscuous mode is enabled in which case 1 N-Char is enough. If it is an RMAP packet
with hardware RMAP enabled 3 N-Chars are needed since the command byte determines where the
packet is processed. Packets smaller than these sizes are discarded.
15.5.2 Basic functionality of a channel
Reception is based on descriptors located in a consecutive area in memory that hold pointers to buffers where packets should be stored. When a packet arrives at the core the channel which should
receive it is first determined as described in the previous section. A descriptor is then read from the
channels’ descriptor area and the packet is stored to the memory area pointed to by the descriptor.
Lastly, status is stored to the same descriptor and increments the descriptor pointer to the next one.
The following sections will describe DMA channel reception in more detail.
15.5.3 Setting up the core for reception
A few registers need to be initialized before reception to a channel can take place. First the link interface need to be put in the run state before any data can be sent. The DMA channel has a maximum
length register which sets the maximum packet size in bytes that can be received to this channel.
Larger packets are truncated and the excessive part is spilled. If this happens an indication will be
given in the status field of the descriptor. The minimum value for the receiver maximum length field
is 4 and the value can only be incremented in steps of four bytes up to the maximum value 33554428.
If the maximum length is set to zero the receiver will not function correctly.
Either the default address register or the channel specific address register (the accompanying mask
register must also be set) needs to be set to hold the address used by the channel. A control bit in the
DMA channel control register determines whether the channel should use default address and mask
registers for address comparison or the channel’s own registers. Using the default register the same
address range is accepted as for other channels with default addressing and the RMAP target while the
separate address provides the channel its own range. If all channels use the default registers they will
GR740-UM-DS, Jun 2016, Version 1.4
142
www.cobham.com/gaisler
GR740
Start Reception
Receive
2 bytes
rmap enabled
and
pid =1 and
defaddr*!defmask =
rxaddr*!defmask
No
Yes
Receive
1 byte
No
Set DMA channel
number to 0
RMAP command
Increment
channel number
Yes
No
Yes
No
Channel enabled
Last DMA channel
Yes
No
Separate addressing
RMAP enabled
No
Yes
No
dma(n).addr*
!dma(n).mask=
rxaddr*!dma(n).mask
defaddr*!defmask =
rxaddr*!defmask
No
Yes
Yes
Process RMAP
command
Store packet to
DMA channel
Discard packet
Figure 17. Flow chart of packet reception.
GR740-UM-DS, Jun 2016, Version 1.4
143
www.cobham.com/gaisler
GR740
accept the same address range and the enabled channel with the lowest number will receive the
packet.
Finally, the descriptor table and control register must be initialized. This will be described in the two
following sections.
15.5.4 Setting up the descriptor table address
The core reads descriptors from an area in memory pointed to by the receiver descriptor table address
register. The register consists of a base address and a descriptor selector. The base address points to
the beginning of the area and must start on a 1024 bytes aligned address. It is also limited to be 1024
bytes in size which means the maximum number of descriptors is 128 since the descriptor size is 8
bytes.
The descriptor selector points to individual descriptors and is increased by 1 when a descriptor has
been used. When the selector reaches the upper limit of the area it wraps to the beginning automatically. It can also be set to wrap at a specific descriptor before the upper limit by setting the wrap bit in
the descriptor. The idea is that the selector should be initialized to 0 (start of the descriptor area) but it
can also be written with another 8 bytes aligned value to start somewhere in the middle of the area. It
will still wrap to the beginning of the area.
If one wants to use a new descriptor table, the receiver enable bit in the corresponding DMA channel
control/status register has to be cleared first. When the RX active bit in the same register is cleared it
is safe to update the descriptor table register. When this is finished and descriptors are enabled the
receiver enable bit can be set again.
15.5.5 Enabling descriptors
As mentioned earlier one or more descriptors must be enabled before reception can take place. Each
descriptor is 8 byte in size and the layout can be found in the tables below. The descriptors should be
written to the memory area pointed to by the receiver descriptor table address register. When new
descriptors are added they must always be placed after the previous one written to the area. Otherwise
they will not be noticed.
A descriptor is enabled by setting the address pointer to point at a location where data can be stored
and then setting the enable bit. The WR bit can be set to cause the selector to be set to zero when
reception has finished to this descriptor. IE should be set if an interrupt is wanted when the reception
has finished. The DMA control register interrupt enable bit must also be set for an interrupt to be generated.
Table 141. GRSPW receive descriptor word 0 (address offset 0x0)
31 30 29 28 27 26 25 24
0
TR DC HC EP IE WR EN
PACKETLENGTH
31
Truncated (TR) - Packet was truncated due to maximum length violation.
30
Data CRC (DC) - 1 if a CRC error was detected for the data and 0 otherwise.
29
Header CRC (HC) - 1 if a CRC error was detected for the header and 0 otherwise.
28
EEP termination (EP) - This packet ended with an Error End of Packet character.
27
Interrupt enable (IE) - If set, an interrupt will be generated when a packet has been received if the
receive interrupt enable bit in the DMA channel control register is set.
26
Wrap (WR) - If set, the next descriptor used by the GRSPW will be the first one in the descriptor
table (at the base address). Otherwise the descriptor pointer will be increased with 0x8 to use the
descriptor at the next higher memory location. The descriptor table is limited to 1 KiB in size and the
pointer will be automatically wrap back to the base address when it reaches the 1 KiB boundary.
25
Enable (EN) - Set to one to activate this descriptor. This means that the descriptor contains valid control values and the memory area pointed to by the packet address field can be used to store a packet.
24: 0
Packet length (PACKETLENGTH) - The number of bytes received to this buffer. Only valid after
EN has been set to 0 by the GRSPW.
GR740-UM-DS, Jun 2016, Version 1.4
144
www.cobham.com/gaisler
GR740
Table 142. GRSPW receive descriptor word 1 (address offset 0x4)
31
0
PACKETADDRESS
31: 0
Packet address (PACKETADDRESS) - The address pointing at the buffer which will be used to store
the received packet.
15.5.6 Setting up the DMA control register
The final step to receive packets is to set the control register in the following steps: The receiver must
be enabled by setting the rxen bit in the DMA control register. This can be done anytime and before
this bit is set nothing will happen. The rxdescav bit in the DMA control register is then set to indicate
that there are new active descriptors. This must always be done after the descriptors have been
enabled or the core might not notice the new descriptors. More descriptors can be activated when
reception has already started by enabling the descriptors and writing the rxdescav bit. When these bits
are set reception will start immediately when data is arriving.
15.5.7 The effect to the control bits during reception
When the receiver is disabled all packets going to the DMA-channel are discarded if the packet’s
address does not fall into the range of another DMA channel. If the receiver is enabled and the address
falls into the accepted address range, the next state is entered where the rxdescav bit is checked. This
bit indicates whether there are active descriptors or not and should be set by the external application
using the DMA channel each time descriptors are enabled as mentioned above. If the rxdescav bit is
‘0’ and the nospill bit is ‘0’ the packets will be discarded. If nospill is ’1’ the core waits until rxdescav
is set and the characters are kept in the N-Char fifo during this time. If the fifo becomes full further Nchar transmissions are inhibited by stopping the transmission of FCTs.
When rxdescav is set the next descriptor is read and if enabled the packet is received to the buffer. If
the read descriptor is not enabled, rxdescav is set to ‘0’ and the packet is spilled depending on the
value of nospill.
The receiver can be disabled at any time and will stop packets from being received to this channel. If
a packet is currently received when the receiver is disabled the reception will still be finished. The
rxdescav bit can also be cleared at any time. It will not affect any ongoing receptions but no more
descriptors will be read until it is set again. Rxdescav is also cleared by the core when it reads a disabled descriptor.
15.5.8 Status bits
When the reception of a packet is finished the enable bit in the current descriptor is set to zero. When
enable is zero, the status bits are also valid and the number of received bytes is indicated in the length
field. The DMA control register contains a status bit which is set each time a packet has been
received. The controller can also be made to generate an interrupt for this event.
The RMAP CRC calculation is always active for all received packets and all bytes except the EOP/
EEP are included. The packet is always assumed to be an RMAP packet and the length of the header
is determined by checking byte 3 which should be the command field. The calculated CRC value is
then checked when the header has been received (according to the calculated number of bytes) and if
it is non-zero the HC bit is set indicating a header CRC error.
The CRC value is not set to zero after the header has been received, instead the calculation continues
in the same way until the complete packet has been received. Then if the CRC value is non-zero the
DC bit is set indicating a data CRC error. This means that the controller can indicate a data CRC error
even if the data field was correct when the header CRC was incorrect. However, the data should not
be used when the header is corrupt and therefore the DC bit is unimportant in this case. When the
header is not corrupted the CRC value will always be zero when the calculation continues with the
data field and the behaviour will be as if the CRC calculation was restarted
GR740-UM-DS, Jun 2016, Version 1.4
145
www.cobham.com/gaisler
GR740
If the received packet is not of RMAP type the header CRC error indication bit cannot be used. It is
still possible to use the DC bit if the complete packet is covered by a CRC calculated using the RMAP
CRC definition. This is because the core does not restart the calculation after the header has been
received but instead calculates a complete CRC over the packet. Thus any packet format with one
CRC at the end of the packet calculated according to RMAP standard can be checked using the DC
bit.
If the packet is neither of RMAP type nor of the type above with RMAP CRC at the end, then both the
HC and DC bits should be ignored.
15.5.9 Error handling
If a packet reception needs to be aborted because of congestion on the network, the suggested solution
is to set link disable to ‘1’. Unfortunately, this will also cause the packet currently being transmitted to
be truncated but this is the only safe solution since packet reception is a passive operation depending
on the transmitter at the other end. A channel reset bit could be provided but is not a satisfactory solution since the untransmitted characters would still be in the transmitter node. The next character
(somewhere in the middle of the packet) would be interpreted as the node address which would probably cause the packet to be discarded but not with 100% certainty. Usually this action is performed
when a reception has stuck because of the transmitter not providing more data. The channel reset
would not resolve this congestion.
If an AHB error occurs during reception the current packet is spilled up to and including the next
EEP/EOP and then the currently active channel is disabled and the receiver enters the idle state. A bit
in the channels control/status register is set to indicate this condition.
15.5.10 Promiscuous mode
The core supports a promiscuous mode where all the data received is stored to the first DMA channel
enabled regardless of the node address and possible early EOPs/EEPs. This means that all non-EOP/
EEP N-Chars received will be stored to the DMA channel. The rxmaxlength register is still checked
and packets exceeding this size will be truncated.
RMAP commands will still be handled by the RMAP hardware target when promiscuous mode is
enabled, if the RMAP enable bit in the core’s Control register is set. If the RMAP enable bit is cleared,
RMAP commands will also be stored to a DMA channel.
15.6
Transmitter DMA channels
The transmitter DMA engine handles transmission of data from the DMA channels to the SpaceWire
network. Each receive channel has a corresponding transmit channel which means there can be up to
4 transmit channels. It is however only necessary to use a separate transmit channel for each receive
channel if there are also separate entities controlling the transmissions. The use of a single channel
with multiple controlling entities would cause them to corrupt each other’s transmissions. A single
channel is more efficient and should be used when possible.
Multiple transmit channels with pending transmissions are arbitrated in a round-robin fashion.
15.6.1 Basic functionality of a channel
A transmit DMA channel reads data from the AHB bus and stores them in the transmitter FIFO for
transmission on the SpaceWire network. Transmission is based on the same type of descriptors as for
the receiver and the descriptor table has the same alignment and size restrictions. When there are new
descriptors enabled the core reads them and transfer the amount data indicated.
GR740-UM-DS, Jun 2016, Version 1.4
146
www.cobham.com/gaisler
GR740
15.6.2 Setting up the core for transmission
Four steps need to be performed before transmissions can be done with the core. First the link interface must be enabled and started by writing the appropriate value to the ctrl register. Then the address
to the descriptor table needs to be written to the transmitter descriptor table address register and one or
more descriptors must also be enabled in the table. Finally, the txen bit in the DMA control register is
written with a one which triggers the transmission. These steps will be covered in more detail in the
next sections.
15.6.3 Enabling descriptors
The descriptor table address register works in the same way as the receiver’s corresponding register
which was covered in section 15.5. The maximum size is 1024 bytes as for the receiver but since the
descriptor size is 16 bytes the number of descriptors is 64.
To transmit packets one or more descriptors have to be initialized in memory which is done in the following way: The number of bytes to be transmitted and a pointer to the data has to be set. There are
two different length and address fields in the transmit descriptors because there are separate pointers
for header and data. If a length field is zero the corresponding part of a packet is skipped and if both
are zero no packet is sent. The maximum header length is 255 bytes and the maximum data length is
16 MiB - 1. When the pointer and length fields have been set the enable bit should be set to enable the
descriptor. This must always be done last. The other control bits must also be set before enabling the
descriptor.
The transmit descriptors are 16 bytes in size so the maximum number in a single table is 64. The different fields of the descriptor together with the memory offsets are shown in the tables below.
The HC bit should be set if RMAP CRC should be calculated and inserted for the header field and
correspondingly the DC bit should be set for the data field. The header CRC will be calculated from
the data fetched from the header pointer and the data CRC is generated from data fetched from the
data pointer. The CRCs are appended after the corresponding fields. The NON-CRC bytes field is set
to the number of bytes in the beginning of the header field that should not be included in the CRC calculation.
The CRCs are sent even if the corresponding length is zero, but when both lengths are zero no packet
is sent not even an EOP.
15.6.4 Starting transmissions
When the descriptors have been initialized, the transmit enable bit in the DMA control register has to
be set to tell the core to start transmitting. New descriptors can be activated in the table on the fly
(while transmission is active). Each time a set of descriptors is added the transmit enable bit in the
corresponding DMA channel control/status register should be set. This has to be done because each
time the core encounters a disabled descriptor this register bit is set to 0.
Table 143. GRSPW transmit descriptor word 0 (address offset 0x0)
31
18 17 16 15 14 13 12 11
RESERVED
DC HC LE IE WR EN
8
NONCRCLEN
7
0
HEADERLEN
31: 18
RESERVED
17
Append data CRC (DC) - Append CRC calculated according to the RMAP specification after the
data sent from the data pointer. The CRC covers all the bytes from this pointer. A null CRC will
be sent if the length of the data field is zero.
GR740-UM-DS, Jun 2016, Version 1.4
147
www.cobham.com/gaisler
GR740
16
Table 143. GRSPW transmit descriptor word 0 (address offset 0x0)
Append header CRC (HC) - Append CRC calculated according to the RMAP specification after the
data sent from the header pointer. The CRC covers all bytes from this pointer except a number of
bytes in the beginning specified by the non-crc bytes field. The CRC will not be sent if the header
length field is zero.
15
Link error (LE) - A Link error occurred during the transmission of this packet.
14
Interrupt enable (IE) - If set, an interrupt will be generated when the packet has been transmitted and
the transmitter interrupt enable bit in the DMA control register is set.
13
Wrap (WR) - If set, the descriptor pointer will wrap and the next descriptor read will be the first one
in the table (at the base address). Otherwise the pointer is increased with 0x10 to use the descriptor at
the next higher memory location.
12
Enable (EN) - Enable transmitter descriptor. When all control fields (address, length, wrap and crc)
are set, this bit should be set. While the bit is set the descriptor should not be touched since this
might corrupt the transmission. The GRSPW clears this bit when the transmission has finished.
11: 8
Non-CRC bytes (NONCRCLEN)- Sets the number of bytes in the beginning of the header which
should not be included in the CRC calculation. This is necessary when using path addressing since
one or more bytes in the beginning of the packet might be discarded before the packet reaches its
destination.
7: 0
Header length (HEADERLEN) - Header Length in bytes. If set to zero, the header is skipped.
Table 144. GRSPW transmit descriptor word 1 (address offset 0x4)
31
0
HEADERADDRESS
31: 0
Header address (HEADERADDRESS) - Address from where the packet header is fetched. Does not
need to be word aligned.
Table 145. GRSPW transmit descriptor word 2 (address offset 0x8)
31
24 23
0
RESERVED
DATALEN
31: 24
RESERVED
23: 0
Data length (DATALEN) - Length in bytes of data part of packet. If set to zero, no data will be sent.
If both data- and header-lengths are set to zero no packet will be sent.
Table 146. GRSPW transmit descriptor word 3(address offset 0xC)
31
0
DATAADDRESS
31: 0
Data address (DATAADDRESS) - Address from where data is read. Does not need to be word
aligned.
15.6.5 The transmission process
When the transmitter enable bit in the DMA channel control/status register is set the core starts reading descriptors immediately. The number of bytes indicated are read and transmitted. When a transmission has finished, status will be written to the first field of the descriptor and a packet sent bit is set
in the DMA control register. If an interrupt was requested it will also be generated. Then a new
GR740-UM-DS, Jun 2016, Version 1.4
148
www.cobham.com/gaisler
GR740
descriptor is read and if enabled a new transmission starts, otherwise the transmit enable bit is cleared
and nothing will happen until it is enabled again.
15.6.6 The descriptor table address register
The internal pointer which is used to keep the current position in the descriptor table can be read and
written through the APB interface. This pointer is set to zero during reset and is incremented each
time a descriptor is used. It wraps automatically when the 1024 bytes limit for the descriptor table is
reached or it can be set to wrap earlier by setting a bit in the current descriptor.
The descriptor table register can be updated with a new table anytime when no transmission is active.
No transmission is active if the transmit enable bit is zero and the complete table has been sent or if
the table is aborted (explained below). If the table is aborted one has to wait until the transmit enable
bit is zero before updating the table pointer.
15.6.7 Error handling
Abort Tx
The DMA control register contains a bit called Abort TX which if set causes the current transmission
to be aborted, the packet is truncated and an EEP is inserted. This is only useful if the packet needs to
be aborted because of congestion on the SpaceWire network. If the congestion is on the AHB bus this
will not help (This should not be a problem since AHB slaves should have a maximum of 16 waitstates). The aborted packet will have its LE bit set in the descriptor. The transmit enable register bit is
also cleared and no new transmissions will be done until the transmitter is enabled again.
AHB error
When an AHB error is encountered during transmission the currently active DMA channel is disabled
and the transmitter goes to the idle mode. A bit in the DMA channel’s control/status register is set to
indicate this error condition and, if enabled, an interrupt will also be generated. Further error handling
depends on what state the transmitter DMA engine was in when the AHB error occurred. If the
descriptor was being read the packet transmission had not been started yet and no more actions need
to be taken.
If the AHB error occurs during packet transmission the packet is truncated and an EEP is inserted.
Lastly, if it occurs when status is written to the descriptor the packet has been successfully transmitted
but the descriptor is not written and will continue to be enabled (this also means that no error bits are
set in the descriptor for AHB errors).
The client using the channel has to correct the AHB error condition and enable the channel again. No
more AHB transfers are done again from the same unit (receiver or transmitter) which was active
during the AHB error until the error state is cleared and the unit is enabled again.
Link error
When a link error occurs during the transmission the remaining part of the packet is discarded up to
and including the next EOP/EEP. When this is done status is immediately written (with the LE bit set)
and the descriptor pointer is incremented. The link will be disconnected when the link error occurs but
the grspw will automatically try to connect again provided that the link-start bit is asserted and the
link-disabled bit is deasserted. If the LE bit in the DMA channel’s control register is not set the transmitter DMA engine will wait for the link to enter run-state and start a new transmission immediately
when possible if packets are pending. Otherwise the transmitter will be disabled when a link error
occurs during the transmission of the current packet and no more packets will be transmitted until it is
enabled again immediately when possible if packets are pending.
GR740-UM-DS, Jun 2016, Version 1.4
149
www.cobham.com/gaisler
GR740
15.7
RMAP
The Remote Memory Access Protocol (RMAP) is used to implement access to resources in the node
via the SpaceWire Link. Some common operations are reading and writing to memory, registers and
FIFOs. This section describes the basics of the RMAP protocol and the target implementation.
15.7.1 Fundamentals of the protocol
RMAP is a protocol which is designed to provide remote access via a SpaceWire network to memory
mapped resources on a SpaceWire node. It has been assigned protocol ID 0x01. It provides three operations write, read and read-modify-write. These operations are posted operations which means that a
source does not wait for an acknowledge or reply. It also implies that any number of operations can be
outstanding at any time and that no timeout mechanism is implemented in the protocol. Time-outs
must be implemented in the user application which sends the commands. Data payloads of up to 16
Mb - 1 is supported in the protocol. A destination can be requested to send replies and to verify data
before executing an operation. A complete description of the protocol is found in the RMAP standard
[RMAP].
15.7.2 Implementation
The core includes a target for RMAP commands which processes all incoming packets with protocol
ID = 0x01, type field (bit 7 and 6 of the 3rd byte in the packet) equal to 01b and an address falling in
the range set by the default address and mask register. When such a packet is detected it is not stored
to the DMA channel, instead it is passed to the RMAP receiver.
The core implements all three commands defined in the standard with some restrictions. Support is
only provided for 32-bit big-endian systems. This means that the first byte received is the msb in a
word. The target will not receive RMAP packets using the extended protocol ID which are always
dumped to the DMA channel.
The RMAP receiver processes commands. If they are correct and accepted the operation is performed
on the AHB bus and a reply is formatted. If an acknowledge is requested the RMAP transmitter automatically send the reply. RMAP transmissions have priority over DMA channel transmissions.
There is a user accessible destination key register which is compared to destination key field in
incoming packets. If there is a mismatch and a reply has been requested the error code in the reply is
set to 3. Replies are sent if and only if the ack field is set to ‘1’.
GR740-UM-DS, Jun 2016, Version 1.4
150
www.cobham.com/gaisler
GR740
When a failure occurs during a bus access the error code is set to 1 (General Error). There is predetermined order in which error-codes are set in the case of multiple errors in the core. It is shown in table
147.
Table 147.The order of error detection in case of multiple errors in the GRSPW. The error detected first has number 1.
Detection Order
Error Code
Error
1
12
Invalid destination logical address
2
2
Unused RMAP packet type or command code
3
3
Invalid destination key
4
9
Verify buffer overrun
5
11
RMW data length error
6
10
Authorization failure
7*
1
General Error (AHB errors during non-verified writes)
8
5/7
Early EOP / EEP (if early)
9
4
Invalid Data CRC
10
1
General Error (AHB errors during verified writes or RMW)
11
7
EEP
12
6
Cargo Too Large
*The AHB error is not guaranteed to be detected before Early EOP/EEP or Invalid Data CRC. For very long accesses
the AHB error detection might be delayed causing the other two errors to appear first.
Read accesses are performed on the fly, that is they are not stored in a temporary buffer before transmitting. This means that the error code 1 will never be seen in a read reply since the header has
already been sent when the data is read. If the AHB error occurs the packet will be truncated and
ended with an EEP.
Errors up to and including Invalid Data CRC (number 8) are checked before verified commands. The
other errors do not prevent verified operations from being performed.
The details of the support for the different commands are now presented. All defined commands
which are received but have an option set which is not supported in this specific implementation will
not be executed and a possible reply is sent with error code 10.
15.7.3 Write commands
The write commands are divided into two subcategories when examining their capabilities: verified
writes and non-verified writes. Verified writes have a length restriction of 4 bytes and the address
must be aligned to the size. That is 1 byte writes can be done to any address, 2 bytes must be halfword
aligned, 3 bytes are not allowed and 4 bytes writes must be word aligned. Since there will always be
only one AHB operation performed for each RMAP verified write command the incrementing
address bit can be set to any value.
Non-verified writes have no restrictions when the incrementing bit is set to 1. If it is set to 0 the number of bytes must be a multiple of 4 and the address word aligned. There is no guarantee how many
words will be written when early EOP/EEP is detected for non-verified writes.
15.7.4 Read commands
Read commands are performed on the fly when the reply is sent. Thus if an AHB error occurs the
packet will be truncated and ended with an EEP. There are no restrictions for incrementing reads but
non-incrementing reads have the same alignment restrictions as non-verified writes. Note that the
“Authorization failure” error code will be sent in the reply if a violation was detected even if the
length field was zero. Also note that no data is sent in the reply if an error was detected i.e. if the status
field is non-zero.
GR740-UM-DS, Jun 2016, Version 1.4
151
www.cobham.com/gaisler
GR740
15.7.5 RMW commands
All read-modify-write sizes are supported except 6 which would have caused 3 B being read and written on the bus. The RMW bus accesses have the same restrictions as the verified writes. As in the verified write case, the incrementing bit can be set to any value since only one AHB bus operation will be
performed for each RMW command. Cargo too large is detected after the bus accesses so this error
will not prevent the operation from being performed. No data is sent in a reply if an error is detected
i.e. the status field is non-zero.
15.7.6 Control
The RMAP target mostly runs in the background without any external intervention, but there are a
few control possibilities.
There is an enable bit in the control register of the core which can be used to completely disable the
RMAP target. When it is set to ‘0’ no RMAP packets will be handled in hardware, instead they are all
stored to the DMA channel.
There is a possibility that RMAP commands will not be performed in the order they arrive. This can
happen if a read arrives before one or more writes. Since the target stores replies in a buffer with more
than one entry several commands can be processed even if no replies are sent. Data for read replies is
read when the reply is sent and thus writes coming after the read might have been performed already
if there was congestion in the transmitter. To avoid this the RMAP buffer disable bit can be set to
force the target to only use one buffer which prevents this situation.
The last control option for the target is the possibility to set the destination key which is found in a
separate register.
GR740-UM-DS, Jun 2016, Version 1.4
152
www.cobham.com/gaisler
GR740
Table 148.GRSPW hardware RMAP handling of different packet type and command fields.
Bit 7
Bit 6
Bit 5
Reserved
Verify
data
Command Write / before
/ Response Read
write
0
0
-
-
-
-
Response
Stored to DMA-channel.
0
1
0
0
0
0
Not used
Does nothing. No reply is sent.
0
1
0
0
0
1
Not used
Does nothing. No reply is sent.
0
1
0
0
1
0
Read single
address
Executed normally. Address has
to be word aligned and data size
a multiple of four. Reply is sent.
If alignment restrictions are violated error code is set to 10.
0
1
0
0
1
1
Read incrementing
address.
Executed normally. No restrictions. Reply is sent.
0
1
0
1
0
0
Not used
Does nothing. No reply is sent.
0
1
0
1
0
1
Not used
Does nothing. No reply is sent.
0
1
0
1
1
0
Not used
Does nothing. Reply is sent with
error code 2.
0
1
0
1
1
1
Read-Modify-Write
incrementing address
Executed normally. If length is
not one of the allowed rmw values nothing is done and error
code is set to 11. If the length
was correct, alignment restrictions are checked next. 1 byte
can be rmw to any address. 2
bytes must be halfword aligned.
3 bytes are not allowed. 4 bytes
must be word aligned. If these
restrictions are violated nothing
is done and error code is set to
10. If an AHB error occurs error
code is set to 1. Reply is sent.
0
1
1
0
0
0
Write, single-address,
do not verify
before writing, no
acknowledge
Executed normally. Address has
to be word aligned and data size
a multiple of four. If alignment is
violated nothing is done. No
reply is sent.
0
1
1
0
0
1
Write, incrementing
address, do
not verify
before writing, no
acknowledge
Executed normally. No restrictions. No reply is sent.
0
1
1
0
1
0
Write, single-address,
do not verify
before writing, send
acknowledge
Executed normally. Address has
to be word aligned and data size
a multiple of four. If alignment is
violated nothing is done and
error code is set to 10. If an AHB
error occurs error code is set to 1.
Reply is sent.
GR740-UM-DS, Jun 2016, Version 1.4
Bit 4
Bit 3
Bit 2
Command
Action
Acknow- Increment
ledge
Address
153
www.cobham.com/gaisler
GR740
Table 148.GRSPW hardware RMAP handling of different packet type and command fields.
15.8
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Command
Action
Reserved
Verify
data
Command Write / before
/ Response Read
write
Acknow- Increment
ledge
Address
0
1
1
0
1
1
Write, incrementing
address, do
not verify
before writing, send
acknowledge
Executed normally. No restrictions. If AHB error occurs error
code is set to 1. Reply is sent.
0
1
1
1
0
0
Write, single
address, verify before
writing, no
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done. Same alignment restrictions apply as for rmw. No reply
is sent.
0
1
1
1
0
1
Write, incrementing
address, verify before
writing, no
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done. Same alignment restrictions apply as for rmw. If they
are violated nothing is done. No
reply is sent.
0
1
1
1
1
0
Write, single
address, verify before
writing, send
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done and error code is set to 9.
Same alignment restrictions
apply as for rmw. If they are violated nothing is done and error
code is set to 10. If an AHB error
occurs error code is set to 1.
Reply is sent.
0
1
1
1
1
1
Write, incrementing
address, verify before
writing, send
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done and error code is set to 9.
Same alignment restrictions
apply as for rmw. If they are violated nothing is done and error
code is set to 10. If an AHB error
occurs error code is set to 1.
Reply is sent.
1
0
-
-
-
-
Unused
Stored to DMA-channel.
1
1
-
-
-
-
Unused
Stored to DMA-channel.
AMBA interface
The AMBA interface consists of an APB interface, an AHB master interface and DMA FIFOs. The
APB interface provides access to the user registers. The DMA engines have 32-bit wide FIFOs to the
AHB master interface which are used when reading and writing to the bus.
The transmitter DMA engine reads data from the bus in bursts which are half the FIFO size in length.
A burst is always started when the FIFO is half-empty or if it can hold the last data for the packet. The
burst containing the last data might have shorter length if the packet is not an even number of bursts in
size.
GR740-UM-DS, Jun 2016, Version 1.4
154
www.cobham.com/gaisler
GR740
The receiver DMA works in the same way except that it checks if the FIFO is half-full and then performs a burst write to the bus which is half the fifo size in length. The last burst might be shorter. Byte
accesses are used for non word-aligned buffers and/or packet lengths that are not a multiple of four
bytes. There might be 1 to 3 single byte writes when writing the beginning and end of the received
packets.
15.8.1 APB slave interface
As mentioned above, the APB interface provides access to the user registers which are 32-bits in
width. The accesses to this interface are required to be aligned word accesses. The result is undefined
if this restriction is violated.
15.8.2 AHB master interface
The core contains a single master interface which is used by both the transmitter and receiver DMA
engines. The arbitration algorithm between the channels is done so that if the current owner requests
the interface again it will always acquire it. This will not lead to starvation problems since the DMA
engines always deassert their requests between accesses.
The AHB accesses can be of size byte, halfword and word (HSIZE = 0x000, 0x001, 0x010). Byte and
halfword accesses are always NONSEQ.
The burst length will be half the AHB FIFO size except for the last transfer for a packet which might
be smaller. Shorter accesses are also done during descriptor reads and status writes.
The AHB master also supports non-incrementing accesses where the address will be constant for several consecutive accesses. HTRANS will always be NONSEQ in this case while for incrementing
accesses it is set to SEQ after the first access. This feature is included to support non-incrementing
reads and writes for RMAP.
If the core does not need the bus after a burst has finished there will be one wasted cycle (HTRANS =
IDLE).
BUSY transfer types are never requested and the core provides full support for ERROR, RETRY and
SPLIT responses.
15.9
Registers
The core is programmed through registers mapped into APB address space.
Table 149.GRSPW registers
APB address offset
Register acronym
Register name
0x00
SPW2.CTRL
Control
0x04
SPW2.STS
Status
0x08
SPW2.DEFADDR
Node address
0x0C
SPW2.CLKDIV
Clock divisor
0x10
SPW2.DKEY
Destination key
0x14
SPW2.TC
Time
0x18 - 0x1C
-
RESERVED
0x20
SPW2.DMACTRL
DMA control/status, channel 1
0x24
SPW2.DMAMAXLEN
DMA RX maximum length, channel 1
0x28
SPW2.DMATXDESC
DMA transmit descriptor table address, channel 1
0x2C
SPW2.DMARXDESC
DMA receive descriptor table address, channel 1
0x30
SPW2.DMAADDR
DMA address, channel 1
GR740-UM-DS, Jun 2016, Version 1.4
155
www.cobham.com/gaisler
GR740
Table 150. 0x00 - SPW2.CTRL - Control
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RA RX RC
NCH
PO
RESERVED
1
1
0
0x1
0
0x00
r
r
r
r
r
r
RD RE
0
1
rw rw
RES
0x0
r
9
8
7
6
5
4
TL TF TR TT LI TQ R RS PM TI
0
0
0
0
0
0
0
3
2
1
0
IE AS LS LD
0
0
0
rw rw rw rw rw rw
r
rw rw rw rw rw rw rw
0
1
0
0
31
RMAP available (RA) - Set to one if the RMAP target is available. Only readable.
30
RX unaligned access (RX) - Set to one if unaligned writes are available for the receiver. Only readable.
29
RMAP CRC available (RC) - Set to one if RMAP CRC is enabled in the core. Only readable.
28: 27
Number of DMA channels (NCH) - The number of available DMA channels minus one (Number of channels = NCH+1).
26
Number of ports (PO) - The number of available SpaceWire ports minus one.
25: 18
RESERVED
17
RMAP buffer disable (RD) - If set only one RMAP buffer is used. This ensures that all RMAP commands
will be executed consecutively.
16
RMAP Enable (RE) - Enable RMAP target.
15: 14
RESERVED
13
Transmitter enable lock control (TL) - Enables / disables the transmitter enable lock functionality
described by the DMACTRL.TL bit. 0 = Disabled, 1 = Enabled.
12
Time-code control flag filter (TF) - When set to 1, a received time-code must have its control flag
bits set to “00” to be considered valid. When set to 0, all control flag bits are allowed.
11
Time Rx Enable (TR) - Enable time-code receptions.
10
Time Tx Enable (TT) - Enable time-code transmissions.
9
Link error IRQ (LI) - Generate interrupt when a link error occurs.
8
Tick-out IRQ (TQ) - Generate interrupt when a valid time-code is received.
7
RESERVED
6
Reset (RS) - Make complete reset of the SpaceWire node. Self clearing.
5
Promiscuous Mode (PM) - Enable Promiscuous mode.
4
Tick In (TI) - The host can generate a tick by writing a one to this field. This will increment the timer counter
and the new value is transmitted after the current character is transferred.
3
Interrupt Enable (IE) - If set, an interrupt is generated when one of bit 8 to 10 is set and its corresponding
event occurs.
2
Autostart (AS) - Automatically start the link when a NULL has been received.
1
Link Start (LS) - Start the link, i.e. allow a transition from ready to started state.
0
Link Disable (LD) - Disable the SpaceWire codec.
Table 151. 0x04 - SPW2.STS - Status
31 30
28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
R
NRXD
NTXD
LS
RESERVED
0x0
0
0
0x0
0x000
r
r
r
r
r
9
8
7
EE IA
0
0
wc wc
6
5
RES
0x0
r
4
3
2
1
0
PE DE ER CE TO
0
0
0
0
0
wc wc wc wc wc
31: 28
RESERVED
27: 26
Number of receive descriptors (NRXD) - Shows the size of the DMA receive descriptor table. Constant value of
0, indicating 128 descriptors.
25: 24
Number of transmit descriptors (NTXD) - Shows the size of the DMA transmit descriptor table. Constant value
of 0, indicating 64 descriptors.
8
Early EOP/EEP (EE) - Set to one when a packet is received with an EOP after the first byte for a non-RMAP
packet and after the second byte for an RMAP packet. Cleared when written with a one.
GR740-UM-DS, Jun 2016, Version 1.4
156
www.cobham.com/gaisler
GR740
7
Table 151. 0x04 - SPW2.STS - Status
Invalid Address (IA) - Set to one when a packet is received with an invalid destination address field, i.e it does
not match the nodeaddr register. Cleared when written with a one.
6: 5
RESERVED
4
Parity Error (PE) - A parity error has occurred. Cleared when written with a one.
3
Disconnect Error (DE) - A disconnection error has occurred. Cleared when written with a one.
2
Escape Error (ER) - An escape error has occurred. Cleared when written with a one.
1
Credit Error (CE) - A credit has occurred. Cleared when written with a one.
0
Tick Out (TO) - A new time count value was received and is stored in the time counter field. Cleared when written with a one.
Table 152. 0x08 - SPW2.DEFADDR - Default address
31
16 15
8
7
0
RESERVED
DEFMASK
DEFADDR
0x0000
0x00
0xFE
r
rw
rw
31: 8
RESERVED
15: 8
Default mask (DEFMASK) - Default mask used for node identification on the SpaceWire network. This
field is used for masking the address before comparison. Both the received address and the DEFADDR field
are anded with the inverse of DEFMASK before the address check.
7: 0
Default address (DEFADDR) - Default address used for node identification on the SpaceWire network.
Table 153. 0x0C - SPW2.CLKDIV - Clock divisor
31
16 15
8
7
0
RESERVED
CLKDIVSTART
CLKDIVRUN
0x0000
0x27
x027
r
rw
rw
31: 16
RESERVED
15: 8
Clock divisor startup (CLKDIVSTART) - Clock divisor value used for the clock-divider during startup
(link-interface is in other states than run). The actual divisor value is Clock Divisor register + 1.
7: 0
Clock divisor run (CLKDIVRUN) - Clock divisor value used for the clock-divider when the link-interface is
in the run-state. The actual divisor value is Clock Divisor register + 1.
Table 154. 0x10 - SPW2.DKEY - Destination key
31
8
7
0
RESERVED
DESTKEY
0x000000
0x00
r
r
31: 8
RESERVED
7: 0
Destination key (DESTKEY) - RMAP destination key.
Table 155. 0x14 - SPW2.TC - Time-code
31
GR740-UM-DS, Jun 2016, Version 1.4
8
7
6
5
0
RESERVED
TCTRL
TIMECNT
0x000000
0x0
0x00
157
www.cobham.com/gaisler
GR740
Table 155. 0x14 - SPW2.TC - Time-code
r
rw
rw
31: 8
RESERVED
7: 6
Time control flags (TCTRL) - The current value of the time control flags. Sent with time-code resulting
from a tick-in. Received control flags are also stored in this register.
5: 0
Time counter (TIMECNT) - The current value of the system time counter. It is incremented for each tick-in
and the incremented value is transmitted. The register can also be written directly but the written value will
not be transmitted. Received time-counter values are also stored in this register.
Table 156. 0x20 - SPW2.DMACTRL - DMA control/status, channel 1
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RESERVED
0x00
r
EP TR
0
0
wc wc
RES
0x0
r
4
3
2
RP TP TL LE SP SA EN NS RD RX AT RA TA PR PS AI
RI
TI RE TE
0
0
0
0
0
0
0
0
9
8
0
7
0
6
0
5
0 NR NR NR
1
0
0
0
0
0
wc wc wc rw rw rw rw rw rw
r
rw wc wc wc wc rw rw rw rw rw
0
31: 24
RESERVED
23
EEP termination (EP) - Set to 1 when a received packet for the corresponding DMA channel ended with an
Error End of Packet (EEP) character.
22
Truncated (TR) - Set to 1 when a received packet for the corresponding DMA channel is truncated due to a
maximum length violation.
21: 20
RESERVED
19
Receive packet IRQ (RP) - This bit is set to 1 when an AMBA interrupt was generated due to the fact that a
packet was received for the corresponding DMA channel.
18
Transmit packet IRQ (TP) - This bit is set to 1 when an AMBA interrupt was generated due to the fact that a
packet was transmitted for the corresponding DMA channel.
17
Transmitter enable lock (TL) - This bit is set to 1 if the CTRL.TL bit is set, and the transmitter for the
corresponding DMA channel is disabled due to a link error (controlled by the DMACTRL.LE bit).
While this bit is set, it is not possible to re-enable the transmitter (e.g. not possible to set the TE bit to
1).
16
Link error disable (LE) - Disable transmitter when a link error occurs. No more packets will be transmitted until the transmitter is enabled again.
15
Strip pid (SP) - Remove the pid byte (second byte) of each packet. The address byte (first byte) will also be
removed when this bit is set independent of the SA bit.
14
Strip addr (SA) - Remove the addr byte (first byte) of each packet.
13
Enable addr (EN) - Enable separate node address for this channel.
12
No spill (NS) - If cleared, packets will be discarded when a packet is arriving and there are no active descriptors. If set, the GRSPW will wait for a descriptor to be activated.
11
Rx descriptors available (RD) - Set to one, to indicate to the GRSPW that there are enabled descriptors in the
descriptor table. Cleared by the GRSPW when it encounters a disabled descriptor:
10
RX active (RX) - Is set to ‘1’ if a reception to the DMA channel is currently active otherwise it is ‘0’.
9
Abort TX (AT) - Set to one to abort the currently transmitting packet and disable transmissions. If no transmission is active the only effect is to disable transmissions. Self clearing.
8
RX AHB error (RA) - An error response was detected on the AHB bus while this receive DMA channel was
accessing the bus.
7
TX AHB error (TA) - An error response was detected on the AHB bus while this transmit DMA channel was
accessing the bus.
6
Packet received (PR) - This bit is set each time a packet has been received. never cleared by the SW-node.
5
Packet sent (PS) - This bit is set each time a packet has been sent. Never cleared by the SW-node.
4
AHB error interrupt (AI) - If set, an interrupt will be generated each time an AHB error occurs when this
DMA channel is accessing the bus.
3
Receive interrupt (RI) - If set, an interrupt will be generated each time a packet has been received. This happens both if the packet is terminated by an EEP or EOP.
GR740-UM-DS, Jun 2016, Version 1.4
158
www.cobham.com/gaisler
GR740
2
Table 156. 0x20 - SPW2.DMACTRL - DMA control/status, channel 1
Transmit interrupt (TI) - If set, an interrupt will be generated each time a packet is transmitted. The interrupt
is generated regardless of whether the transmission was successful or not.
1
Receiver enable (RE) - Set to one when packets are allowed to be received to this channel.
0
Transmitter enable (TE) - Write a one to this bit each time new descriptors are activated in the table. Writing
a one will cause the SW-node to read a new descriptor and try to transmit the packet it points to. This bit is
automatically cleared when the SW-node encounters a descriptor which is disabled.
Table 157. 0x24 - SPW2.DMAMAXLEN - DMA RX maximum length, channel 1
31
25 24
2
1
0
RESERVED
RXMAXLEN
RES
0x00
N/R
0x0
r
rw
r
31: 25
RESERVED
24: 2
RX maximum length (RXMAXLEN) - Receiver packet maximum length in 32-bit words.
1: 0
RESERVED
Table 158. 0x28 - SPW2.DMATXDESC - DMA transmitter descriptor table address, channel 1
31
10
9
4
3
0
DESCBASEADDR
DESCSEL
RESERVED
N/R
0x00
0x0
rw
rw
r
31: 10
Descriptor table base address (DESCBASEADDR) - Sets the base address of the descriptor table
9: 4
Descriptor selector (DESCSEL) - Offset into the descriptor table. Shows which descriptor is currently used
by the GRSPW. For each new descriptor read, the selector will increase with 16 and eventually wrap to zero
again.
3: 0
RESERVED
Table 159. 0x2C - SPW2.DMARXDESC - DMA receiver descriptor table address, channel 1
31
10
9
3
2
0
DESCBASEADDR
DESCSEL
RESERVED
N/R
0x00
0x0
rw
rw
r
31: 10
Descriptor table base address (DESCBASEADDR) - Sets the base address of the descriptor table.
9: 3
Descriptor selector (DESCSEL) - Offset into the descriptor table. Shows which descriptor is currently used.
For each new descriptor read, the selector will increase with 8 and eventually wrap to zero again.
2: 0
RESERVED
Table 160. 0x30 - SPW2.DMAADDR - DMA address, channel 1
31
31: 8
16 15
8
7
0
RESERVED
MASK
ADDR
0x0000
N/R
N/R
r
rw
rw
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
159
www.cobham.com/gaisler
GR740
15: 8
7: 0
Table 160. 0x30 - SPW2.DMAADDR - DMA address, channel 1
Mask (MASK) - Mask used for node identification on the SpaceWire network. This field is used for masking
the address before comparison. Both the received address and the ADDR field are anded with the inverse of
MASK before the address check.
Address (ADDR) - Address used for node identification on the SpaceWire network for the corresponding
dma channel when the EN bit in the DMA control register is set.
GR740-UM-DS, Jun 2016, Version 1.4
160
www.cobham.com/gaisler
GR740
16
AHB Trace buffer tracing Master I/O AHB bus
16.1
Overview
The trace buffer consists of a circular buffer that stores AMBA AHB data transfers performed on the
Master I/O AHB bus. The address, data and various control signals of the AHB bus are stored and can
be read out, via the core’s interface attached to the Debug AHB bus, for later analysis. Note that the
LEON4 Debug Support Unit (DSU4) also includes an AHB trace buffer, tracing the Processor AHB
bus.
The trace buffer will, together with all other cores on the Debug AHB bus, be gated off when the
Debug AHB bus is disabled via the external DSU_EN signal.
The trace buffer is 128 bits wide, the information stored is indicated in the table below:
Table 161.AHB Trace buffer data allocation
Bits
Name
Definition
127:96
Time tag
The value of the time tag counter
95
AHB breakpoint hit
Set to ‘1’ if a AHB breakpoint hit occurred.
94:80
Hirq
AHB HIRQ[15:1]
79
Hwrite
AHB HWRITE
78:77
Htrans
AHB HTRANS
76:74
Hsize
AHB HSIZE
73:71
Hburst
AHB HBURST
70:67
Hmaster
AHB HMASTER
66
Hmastlock
AHB HMASTLOCK
65:64
Hresp
AHB HRESP
63:32
Load/Store data
AHB HRDATA or HWDATA
31:0
Load/Store address
AHB HADDR
In addition to the AHB signals, a 32-bit counter is also stored in the trace as time tag. The time tag
value is taken from the debug support unit and the debug support unit timer must be enabled for this
value to increment. The same timer source is also activated when at least one of the processors has the
up-counter, described in section 6.10.3, enabled.
16.2
Operation
The 1 KiB trace buffer is enabled by setting the enable bit (EN) in the trace control register. Each
AMBA AHB transfer is then stored in the buffer in a circular manner. The address to which the next
transfer is written is held in the trace buffer index register, and is automatically incremented after each
transfer. Tracing is stopped when the EN bit is reset, or when a AHB breakpoint is hit. An interrupt is
generated when a breakpoint is hit.
16.2.1 AHB statistics
The core will generate events that can be monitored with the LEON statistics unit (L4STAT). The statistical outputs can be filtered by the AHB trace buffer filters, this is controlled by the Performance
counter Filter bit (PF) in the AHB trace buffer control register. The core can collect data for the events
listed in table 162 below
GR740-UM-DS, Jun 2016, Version 1.4
161
www.cobham.com/gaisler
GR740
Table 162.AHB events
16.3
Event
Description
Note
idle
HTRANS=IDLE
Active when HTRANS IDLE is driven on the AHB slave inputs and
slave has asserted HREADY.
busy
HTRANS=BUSY
Active when HTRANS BUSY is driven on the AHB slave inputs and
slave has asserted HREADY.
nseq
HTRANS=NONSEQ
Active when HTRANS NONSEQ is driven on the AHB slave inputs
and slave has asserted HREADY.
seq
HTRANS=SEQ
Active when HTRANS SEQUENTIAL is driven on the AHB slave
inputs and slave has asserted HREADY.
read
Read access
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY and the HWRITE input is low.
write
Write access
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY and the HWRITE input is high.
hsize
Transfer size
Active when HTRANS is SEQUENTIAL or NON-SEQUENTIAL,
slave has asserted HREADY.
ws
Wait state
Active when HREADY input to AHB slaves is low and AMBA
response is OKAY.
retry
RETRY response
Active when master receives RETRY response
split
SPLIT response
Active when master receives SPLIT response
Registers
16.3.1 Register address map
The trace buffer occupies 128 KiB of address space in the Debug bus AHB I/O area. The following
register address are decoded:
Table 163.Trace buffer address space
Address
Register
0x000000
Trace buffer control register
0x000004
Trace buffer index register
0x000008
Time tag value
0x00000C
Trace buffer master/slave filter register
0x000010
AHB break address 1
0x000014
AHB mask 1
0x000018
AHB break address 2
0x00001C
AHB mask 2
0x010000 - 0x020000
Trace buffer
..0
Trace bits 127 - 96
...4
Trace bits 95 - 64
...8
Trace bits 63 - 32
...C
Trace bits 31 - 0
GR740-UM-DS, Jun 2016, Version 1.4
162
www.cobham.com/gaisler
GR740
16.3.2 Trace buffer control register
The trace buffer is controlled by the trace buffer control register:
Table 164.0x000000 - CTRL - Trace buffer control register
31
23 22
16 15 14
RESERVED
12 11
DCNT
RESERVED
0
0
r
rw
9
8
7
6
PF
BW
0
0
0b00
r
rw
r
5
4
3
2
1
0
RF AF FR FW DM EN
0
0
*
rw rw rw rw
0
0
0
r
rw
31: 23
RESERVED
22: 16
Trace buffer delay counter (DCNT) - Specifies the number of lines that should be written in the trace
buffer befiore entering debug mode after a AHB break/watchpoint has been hit.
15: 9
RESERVED
8
Performance counter Filter (PF) - If this bit is set to ‘1’, the cores performance counter (statistical)
outputs will be filtered using the same filter settings as used for the trace buffer. If a filter inhibits a
write to the trace buffer, setting this bit to ‘1’ will cause the same filter setting to inhibit the pulse on
the statistical output, which is connected to the LEON4 statisics unit. The filter settings are controlled by fields AF, FT and FW (bits 4:2) below.
7: 6
Bus width (BW) - Read-only register with value "00" indicating a bus width of 32 bits.
5
Retry filter (RF) - If this bit is set to ‘1’, AHB retry responses will not be included in the trace buffer.
4
Address Filter (AF) - If this bit is set to ‘1’, only the address range defined by AHB trace buffer
breakpoint 2’s address and mask will be included in the trace buffer.
3
Filter Reads (FR) - If this bit is set to ‘1’, read accesses will not be included in the trace buffer. This
bit can only be set of the core has been implemented with support for filtering.
2
Filter Writes (FW) - If this bit is set to ‘1’, write accesses will not be included in the trace buffer.
This bit can only be set of the core has been implemented with support for filtering.
1
Delay counter mode (DM) - Indicates that the trace buffer is in delay counter mode.
0
Trace enable (EN) - Enables the trace buffer
This field has reset value 1 if the BREAK signal is LOW and has reset value 0 otherwise.
16.3.3 Trace buffer index register
The trace buffer index register indicates the address of the next 128-bit line to be written.
Table 165.0x000004 - INDEX - Trace buffer index register
31
11 10
4
3
0
RESERVED
INDEX
RESERVED
0
NR
0
r
rw
r
31: 11
RESERVED
10: 4
Trace buffer index counter (INDEX) - Indicates the address of the next 128-bit line to be written.
3: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
163
www.cobham.com/gaisler
GR740
16.3.4 Trace buffer time tag register
The time tag value displayed in this register is stored in the trace. The time tag value is taken from the
debug support unit and the debug support unit timer must be enabled for this value to increment. The
same timer source is also activated when at least one of the processors has the up-counter, described in
section 6.10.3, enabled.
Table 166.0x000008 - TIMETAG - Trace buffer time tag register
31
0
TIMETAG
0
t
31: 0
Time tag value (TIMETAG) - See description above table
16.3.5 Trace buffer master/slave filter register
The master/slave filter register allows filtering out specified master and slaves from the trace. This
register can only be assigned if the trace buffer has been implemented with support for filtering.
Table 167.0x00000C - MSFILT - Trace buffer master/slave filter register
31
16 15
SMASK[15:0]
0
MMASK[15:0]
0
0
rw
rw
31: 16
Slave Mask (SMASK) - If SMASK[n] is set to ‘1’, the trace buffer will not save accesses performed
to slave n.
15: 0
Master Mask (MMASK) - If MMASK[n] is set to ‘1’, the trace buffer will not save accesses performed by master n.
GR740-UM-DS, Jun 2016, Version 1.4
164
www.cobham.com/gaisler
GR740
16.3.6 Trace buffer breakpoint registers
The trace buffer contains two breakpoint registers for matching AHB addresses. A breakpoint hit is
used to freeze the trace buffer by clearing the enable bit. Freezing can be delayed by programming the
DCNT field in the trace buffer control register to a non-zero value. In this case, the DCNT value will
be decremented for each additional trace until it reaches zero and after two additional entries, the trace
buffer is frozen. A mask register is associated with each breakpoint, allowing breaking on a block of
addresses. Only address bits with the corresponding mask bit set to ‘1’ are compared during breakpoint detection. To break on AHB load or store accesses, the LD and/or ST bits should be set.
Table 168. 0x000010, 0x000018 - TBBA - Trace buffer break address registers
31
2
0
RES
NR
0
rw
r
31: 2
Break point address (BADDR) - Bits 31:2 of breakpoint address
1
RESERVED
0
1
BADDR[31:2]
Table 169. 0x000014, 0x00001C - TBBM - Trace buffer break mask registers
31
2
BMASK[31:2]
0
NR
0
rw
rw rw
31: 2
Breakpoint mask (BMASK) - See description above tables.
1
Load (LD) - Break on data load address
0
Store (ST) - Break on data store address
GR740-UM-DS, Jun 2016, Version 1.4
1
LD ST
165
0
www.cobham.com/gaisler
GR740
17
SpaceWire router
17.1
Overview
The SpaceWire router core implements a SpaceWire routing switch as defined in [SPW]. It provides
an Remote Memory Access Protocol (RMAP) target according to [RMAP] for configuration at port 0
used for accessing internal configuration and status registers. In addition to this there are two different
port types: SpaceWire ports and AMBA ports. The router implements a total of eight SpaceWire ports
and four AMBA ports.
Among the features supported by the router are: group adaptive routing, packet distribution, system
time-distribution, distributed interrupts, port timers to recover from deadlock situations, SpaceWire-D
[SPWD] packet truncation based time-slot violations, and SpaceWire Plug-and-Play [SPWPNP].
Figure 18. Block diagram
17.2
Operation
The router ports are interconnected using a non-blocking switch matrix which can connect any input
port to any output port. Access to each output port is arbitrated using a round-robin arbitration
scheme. A single routing-table is used for the whole router. Access to the table is also arbitrated using
a round-robin scheme.
The ports consist of configuration port 0 and two different types of external ports: SpaceWire links
and AMBA interfaces. All ports have the same interface to the switch matrix and behave in the same
manner. The difference in behavior is on the external side of the port. The SpaceWire ports provide
standard SpaceWire link interfaces using on-chip LVDS. The AMBA ports transfer characters from
and to an AHB bus using DMA. The different port types are described in further detail in sections
17.3, 17.4 and 17.5.
GR740-UM-DS, Jun 2016, Version 1.4
166
www.cobham.com/gaisler
GR740
17.2.1 Port numbering
The router’s ports are numbered as follows: configuration port has port number 0, the SpaceWire
ports have port numbers 1-8, and the AMBA ports have port numbers 9-12.
17.2.2 Routing table
A single routing table is provided. The access to the routing table is arbitrated using a round-robin
arbiter, with each port being of equal priority. The operation is pipelined and one lookup can be done
each cycle. This way the maximum latency is equal to the number of ports in the router minus one.
The impact on throughput should be negligible provided that packets do not arrive at the same time.
The probability for this is higher when the traffic only consists of very small packets sent continuously (the average size being about the same as the number of ports). This should be a very uncommon case. Latency is still bounded and probably negligible in comparison to other latencies in most
systems.
The routing table is configured with either RMAP or AMBA AHB accesses to the configuration port.
Configuration of the routing table does not introduce any extra latency for packets, since the configuration accesses have lower priority than packet traffic. The routing table is split into two parts, one
which controls the port mapping for the address (RTR.RTPMAP registers), and one which controls
properties for the address, such as priority and header deletion (RTR.RTACTRL registers).
17.2.3 Port mapping
For both physical and logical addresses it is possible to configure which port(s) the incoming packet
should be routed to. This is done by programming the corresponding RTR.RTPMAP register. The
RTR.RTPMAP registers also controls whether or not group adaptive routing or packet distribution
should be used for the incoming packet. The RTR.RTPMAP registers are not initialized after reset /
power-up. For physical addresses this has the effect that the incoming packet is routed to the port that
matches the address in the packet, without any group adaptive routing or packet distribution. For logical addresses, an uninitialized RTR.RTPMAP register (or if the RTR.RTPMAP.PE field has been
written with all zeroes) has the effect that the incoming packet is spilled. See table 199 in section
17.5.3 for more details.
17.2.4 Address control
For both physical and logical addresses it is possible to configure the priority, and to enable the spillif-not-ready feature (explained in section 17.2.10). For logical addresses it is also possible to enable /
disable the address, and enable / disable header deletion. Physical addresses are always enabled, and
always have header deletion enabled, as specified by [SPW]. This configuration for an address is done
by programming the corresponding RTR.RTACTRL register. Logical addresses are disabled after
reset / power-up. An incoming packet with a disabled logical address is spilled. See table 200 in section 17.5.3 for details.
17.2.5 Output port arbitration
Each output port is arbitrated individually based on the address of the incoming packet, using two priority levels, with round-robin at each level. Each physical address and logical address can be configured in the routing table (RTR.RTACTRL register) to be either high or low priority. Priority
assignments can have large impact on the delays for packets, because packets can be large and the
speed of the data consumer and link itself may not be known. This should therefore be considered
when assigning priorities.
GR740-UM-DS, Jun 2016, Version 1.4
167
www.cobham.com/gaisler
GR740
17.2.6 Group adaptive routing
Group adaptive routing can be used to allow incoming packets to be sent on different output ports
depending on which of the output ports that are currently ready. It can be enabled for both physical
and logical addresses, and is configured by programming the corresponding RTR.RTPMAP register.
When a packet arrives, and group adaptive routing is enabled for the packet’s address, the router looks
at the group of ports selected by the corresponding RTR.RTPMAP register and transmits the packet
on the port with the lowest index that is currently ready. Ready in this context means that the port’s
link interface is in run-state and currently not sending any other packet. If none of the selected output
ports are ready, the incoming packet will either be spilled or transmitted on the first port that becomes
ready. The action taken depends on the setting of the input port’s data character timer (see section
17.2.15), the spill-if-not-ready feature for the address (see section 17.2.10), and the link-start-onrequest feature for the output ports (see section 17.2.13). See table 199 in section 17.5.3 for details on
how to enable and configure group adaptive routing.
17.2.7 Packet distribution
Packet distribution can be used to implement multicast and broadcast addresses, and can be enabled
for both physical and logical addresses. Packet distribution is enabled and configured by programming the corresponding RTR.RTPMAP register.
When a packet arrives, and packet distribution is enabled for the packet’s address, the router looks at
the group of ports selected by the corresponding RTR.RTPMAP register. If all of the selected ports are
ready, the packet is transmitted on all the ports. Ready in this context means that the port’s link interface is in run-state and currently not sending any other packet. If one or more of the selected ports are
not ready, the incoming packet will either be spilled or transmitted once all ports are ready. The action
taken depends on the setting of the input port’s data character timer (section 17.2.15), the spill-if-notready feature for the address (section 17.2.10), and the link-start-on-request feature for the output
ports (section 17.2.13). See table 199 in section 17.5.3 for details on how to enable and configure
packet distribution.
17.2.8 Port disable
A port can be disabled for data traffic by setting the corresponding RTR.PCTRL.DI bit. Incoming
packets on a disabled port are silently spilled, and no packets are routed to a disabled port. A disabled
port will not be included in any group used for group adaptive routing or packet distribution, even if
the corresponding bit in that address’ RTR.RTPMAP.PE field is set. When routing packets that are
incoming on other ports, the router will simply behave as if the disabled port did not exist.
Note that the link interface for a SpaceWire port or the AMBA interface of an AMBA port is not
affected by the RTR.PCTRL.DI bit, only the port’s interface to the router switch matrix is affected.
Also, the RTR.PCTRL.DI bit only affects routing of data, thus the transmission and reception of timecodes and interrupt codes are not affected.
17.2.9 Static routing
The router supports a feature called static routing, which can be enabled individually per port. When
enabled, all incoming packets on the port are routed based on the physical address specified in the
port’s RTR.PCTRL2.SC field, and the setting of the corresponding RTR.PCTRL2.SC bit, instead of
the address in the packets. Header deletion is not used for the incoming packets when static routing is
enabled, which means that the first byte of the packets are always sent to the output port as well. Static
routing to port 0 is not allowed, and will generate an invalid address error if attempted.
Note that when static routing is enabled for a port, it is not possible to access the configuration port
from the same port.
GR740-UM-DS, Jun 2016, Version 1.4
168
www.cobham.com/gaisler
GR740
17.2.10 Spill-if-not-ready
The spill-if-not-ready feature can be enabled individually for each physical and logical address by
configuring the corresponding RTR.RTACTRL.SR bit. When enabled, an incoming packet is spilled
if the selected output port is not ready. For a SpaceWire port to be ready it means that the port’s link
interface must be in run-state. Ready in this sense only means that the port’s SpaceWire link interface
must be in run-state. The configuration port, and the AMBA ports, are always considered ready, since
they do not have any SpaceWire link interfaces. If group adaptive routing is used for the incoming
packet then the packet is only spilled if none of the ports in the group are ready. If packet distribution
is used for the incoming packet then the packet is spilled unless all the selected output ports are ready.
The spill-if-not-ready feature has priority over the incoming port’s data character timer (section
17.2.15) and the output port’s link-start-on-request feature (section 17.2.13). This means that if the
spill-if-not-ready feature is enabled, the packet is spilled before the timer starts, and the link-start-onrequest feature will never be activated.
17.2.11 Self addressing
Self addressing occurs when a selected output port for a packet is the same port as the input port.
Whether or not this is allowed is controlled by the RTR.RTRCFG.SA bit. If self addressing is not
allowed, the incoming packet is spilled and an invalid address error occurs. When group adaptive
routing is used, and self addressing is not allowed, the input port is still allowed to be in the group of
ports configured for the packet. The packet is not spilled until the router actually selects the input port
as output port. If the router selects one of the other ports in the group, the packet is not spilled. When
packet distribution is used, and self addressing is not allowed, the input port is not allowed to be in the
group of ports configured for the packet, since the packet should be sent to all ports in the group.
17.2.12 Invalid address error
An invalid address error occurs under the conditions listed below.
•
When an incoming packet’s address corresponds to a non-existing port (physical addresses 1331).
•
When an incoming packet’s address is a logical address that is not enabled (RTR.RTACTRL.EN
= 0).
•
When an incoming packet’s address is a logical address for which the corresponding RTR.RTPMAP register is not initialized, or the corresponding RTR.RTPMAP.PE field set to all zeroes.
•
When only one output port is selected for an incoming packet, and that port is disabled
(RTR.PCTRL.DI = 1).
•
When self addressing occurs, and the router is configured not to allow self addressing
(RTR.RTRCFG.SA = 0).
•
When a packet is routed with the static routing feature, and the physical address programmed in
RTR.PCTRL2.SD is 0 (static routing to port 0 is not allowed).
For all the invalid address cases above, the incoming packet is spilled, and the RTR.PSTS.IA bit corresponding to the input port will be set to 1.
17.2.13 Link-start-on-request
The link-start-on-request feature gives the possibility to automatically start a SpaceWire port’s link
interface when a packet is routed to the port (i.e, port is selected as output port). Each port can have
the feature individually enabled by setting the corresponding RTR.PCTRL.LR bit to 1.
If a packet arrives, and the link interface of the selected output port is not in run-state, and the port has
the link-start-on-request feature enabled, the router will try to start the link interface under the following conditions:
GR740-UM-DS, Jun 2016, Version 1.4
169
www.cobham.com/gaisler
GR740
1. The link interface is not already trying to start (RTR.PCTRL.LS = 0).
2. The link is not disabled (RTR.PCTRL.LD = 0).
3. The spill-if-not-ready feature is not enabled for the packet being routed.
The link will continue to be started until either the RTR.PCTRL.LD bit is set to 1, or until the link is
disabled through the auto-disconnect feature, described in section 17.2.14.
The link-start-on-request feature is only available for the SpaceWire ports, since the configuration
port and AMBA ports does not have a SpaceWire link interface.
17.2.14 Auto-disconnect
The auto-disconnect feature gives the possibility to automatically disable the link interface of a
SpaceWire port if the port has been inactive for a long enough period of time. Each port can have the
feature individually enabled by setting their corresponding RTR.PCTRL.AD bit to 1. The amount of
time the port needs to be inactive for is decided by the settings of the global prescaler register
(RTR.PRESCALER), and the port’s individual timer register (RTR.PTIMER). This time period is the
same as the timeout period used by the port’s data character timer when recovering from deadlock situations (see section 17.2.15). If the auto-disconnect feature is enabled, then a SpaceWire port will
automatically disable its link interface under the following conditions:
1. The link interface entered run-state because it was started by the link-start-on-request feature,
described in section 17.2.13.
2. The packet that caused the link interface to start has finished (either sent or spilled).
3. Nothing has been transmitted or received on the port for the duration of the time period specified by
the RTR.PRESCALER register together with the corresponding RTR.PTIMER register.
4. The port’s corresponding RTR.PCTRL.LS bit has not been set to 1.
The auto-disconnect feature is only available for the SpaceWire ports, since the configuration port and
AMBA ports does not have a SpaceWire link interface.
17.2.15 Port data character timers
Each port has an individual data character timer, which can be used to timeout an ongoing data transfer in order to recover from a deadlock situation. There are two different timeouts defined: overrun
timeout, and underrun timeout. An overrun timeout is when the input port has data available, but the
output port(s) can not accept data fast enough. An underrun timeout is when the output port(s) can
accept more data, but the input port can not provide data fast enough.
The timeout period for a port is set in its corresponding RTR.PTIMER register, and the timer is
enabled through the corresponding RTR.PCTRL.TR bit. Timeouts due to overrun and underrun can
also be individually enabled / disabled through the corresponding RTR.PCTRL2.OR and
RTR.PCTRL2.UR bits.
It is always the input port’s data character timer that is used for timing data transfers, and the timer is
started after a wormhole has been established between the input port and output port(s). When the
timer is started, it counts down on every tick from the global prescaler (RTR.PRESCALER register).
If a data character is transmitted from the input port to the output port(s), then the timer is restarted. If
the timer expires, the ongoing packet is spilled, and an EEP is written to the transmit FIFO of the output port(s).
The range of the timeout period depends on the system clock frequency, and is calculated with the following formula:
<timeout period> = (<clock period> x (RTR.PRESCALER+1)) x RTR.PTIMER
Sub-sections 17.2.15.1 through 17.2.15.4 clarifies the behaviour of the timers for different scenarios
that can occur when a packet arrives.
GR740-UM-DS, Jun 2016, Version 1.4
170
www.cobham.com/gaisler
GR740
17.2.15.1 Timer disabled
If the data character timer for an input port is disabled, the incoming packet will wait indefinitely for
the output port(s) to be ready, unless the spill-if-not-ready feature is enabled for the packet’s address
(see section 17.2.10).
17.2.15.2 Timer enabled, but output port(s) not in run state
If the spill-if-not-ready feature (see section 17.2.10) is disabled for the incoming packet’s address, the
input port’s data character timer is started when the packet arrives, and if the output port’s SpaceWire
link interface has not entered run-state when the timer expires, the packet is spilled. Note that the
AMBA ports are always considered to be in run-state, since they do not have any SpaceWire link
interfaces. When group adaptive routing is used for the incoming packet, it is enough for one of the
possible output ports to enter run-state before the timer expires for the packet to be transmitted. If
packet distribution is used, all the output ports must enter run-state before the timer expires, otherwise
the packet is spilled.
If the link-start-on-request feature is enabled for an output port, that port will try to enter run-state
when the packet arrives. However, the input port’s timer is unaffected of this, and will still only wait
for its configured timeout period, before spilling the packet.
A timeout due to the output port not being in run-state is classified as a overrun timeout, which means
that the RTR.PCFG2.OR bit for the input port must be set in order for the packet to be spilled. If the
RTR.PCFG2.OR bit is not set, the packet will wait indefinitely (unless spill-if-not-ready is enabled).
17.2.15.3 Timer enabled, output port(s) in run-state but busy with other transmission
The input port’s data character timer will not start, and the incoming packet will wait indefinitely until
the output port either becomes free or leaves run-state. When group adaptive routing is used for the
incoming packet, it is enough for one of the possible output ports to be in run-state to prevent the
timer from starting. If packet distribution is used, all the output ports must be in run-state to prevent
the timer from starting.
Note that the AMBA ports are always considered to be in run-state, since they do not have any SpaceWire link interfaces.
17.2.15.4 Timer functionality when accessing the configuration port
The timer functionality is basically the same for the configuration port as for the other ports. When the
command is being received, the configuration port is the output port of the data transfer, and when the
reply is being sent, the configuration port is the input port of the data transfer. The differences
between the configuration port and the other ports are:
•
The configuration port can always accept data fast enough, which means that an overrun timeout
will never occur when a command is being received.
•
The configuration port can always send data fast enough, which means that an underrun timeout
will never occur when a reply is being sent.
17.2.16 Packet length truncation
Packet length truncation monitors the length of an incoming packet, and increases a counter for each
received data character. If the counter reaches a value larger than the input port’s RTR.MAXPLEN
register, and truncation is enabled for the input port (RTR.PCTRL.PL = 1), the rest of the packet is
spilled, and an EEP is written to the FIFO of the output port(s). Each port has its own RTR.MAXPLEN register and counter in order to allow different maximum lengths for different ports.
Packet length truncation can also be enabled for port 0. In that case, it is the length of the RMAP /
SpaceWire Plug-and-Play reply packet that is monitored.
GR740-UM-DS, Jun 2016, Version 1.4
171
www.cobham.com/gaisler
GR740
17.2.17 System time-distribution
The router supports system time distribution through time-codes, as defined in [SPW]. It contains a
global time-counter register (RTR.TC) where the latest received time-code can be read. Both the
SpaceWire ports and the AMBA ports support time-code transmission and reception. All incoming
time-codes update the RTR.TC register. If the incoming time-code has a time value which is plus one
(modulo 64) compared to the old RTR.TC value, the time-code is forwarded to all the other ports. The
time-code is never sent out onto the port on which it arrived. More details about the sending and
receiving of time-codes through the AMBA ports are given in section 17.4.3.
Time-codes can be globally enabled / disabled through the RTR.TC register, as well as individually
enabled / disabled per port through respective RTR.PCTRL.TE bit. When time-codes are disabled for
a port, all incoming time-codes on that port are discarded, and no time-codes will be forwarded to the
port.
The router can be configured to either filter out all incoming time-codes that does not have the two
control flags (bit 7:6) set to “00”, or to discard the control flags and allow them to have any value.
This configuration is done through the RTR.RTRCFG.TF bit. The value of the control flags for the last
received time-code can also be read from the RTR.TC register. Note that if interrupt distribution is
globally enabled (RTR.RTRCTRL.IE = 1), only control flags “00” are considered as time-codes, no
matter the value of the RTR.RTRCFG.TF bit.
17.2.18 SpaceWire distributed interrupt support
The router supports SpaceWire distributed interrupts. There are three different type of distributed
interrupt codes defined: interrupt code, interrupt acknowledgement code, and extended interrupt code.
All distributed interrupt codes are control codes that have the control flags (bits 7:6) set to “10”. Bit 5
of the control code specifies if the code is an interrupt code (bit 5 = ‘0’), or an interrupt acknowledgement code or extended interrupt code (bit 5 = ‘1’). The interrupt codes and extended interrupt codes
are generated by the source of the interrupt event, while the interrupt acknowledgement code is sent
by the interrupt handler for the corresponding interrupt number.
Note that since interrupt acknowledgement codes and extended interrupt codes are identical, they can
not co-exist in the same network. The router can operate in two different modes, one with 32 interrupts with acknowledgements, and one with 64 interrupts without acknowledgements. The mode is
selected through the RTR.RTRCFG.EE bit. When configured for the mode with 32 interrupts, all
codes with bit 5 set to one are handled as interrupt acknowledgement codes. When configured for the
mode with 64 interrupts, all codes with bit 5 set to one are handled as extended interrupt codes.
The router has two 32-bit ISR register (RTR.ISR0 and RTR.ISR1) where each bit corresponds to one
interrupt number. A bit in the ISR registers is set to 1 when an interrupt code, or extended interrupt
codes, with the corresponding interrupt number is received. A bit in the ISR registers is set to 0 when
an interrupt acknowledgement code with the corresponding interrupt number is received. This way
the ISR registers reflects the status for all interrupt numbers. Each interrupt number also have its own
timer which is used to clear the ISR register bit if an interrupt acknowledgement code is not received
before the timer expires, as well as an optional timer which is used to control how fast a bit in the
RTR.ISR register is allowed to toggle. See section 17.2.18.2 for more details on the ISR timers. Note
that it is possible to clear the bits in the ISR registers, however these registers should normally only be
used for diagnostics and FDIR.
17.2.18.1 Receiving and transmitting distributed interrupt codes
When a distributed interrupt code is received on a port, or the auxiliary time code / distributed interrupt code interface, the following requirements must be fulfilled in order for the code to be distributed:
1. Interrupt distribution is globally enabled (RTR.RTRCTRL.IE = 1), and enabled for the port that
received the code (corresponding RTR.PCTRL.IC = 1).
GR740-UM-DS, Jun 2016, Version 1.4
172
www.cobham.com/gaisler
GR740
2. If the received code is an interrupt code the RTR.PCTRL2.IR bit for the port must be set to 1. If the
received code is an interrupt acknowledgement code or extended interrupt code the RTR.PCTRL2.AR
bit for the port must be set to 1.
3. If the code is an interrupt code or extended interrupt code, the interrupt number’s corresponding bit
in RTR.ISR0 or RTR.ISR1 must be 0. If the code is an interrupt acknowledgement code, the corresponding bit in RTR.ISR0 must be 1.
4. No previous interrupt code / interrupt acknowledgement code with the same interrupt number is
waiting to be distributed.
5. The ISR change timers (see section 17.2.18.2) are either globally disabled (RTR.RTRCFG.IC = 0)
or the interrupt number’s corresponding ISR change timer has expired.
If one of the requirements above is not fulfilled, then the received code is discarded. If all of the
requirements above are fulfilled, then the received code is placed in a queue. The queue is then serviced in one of the four following ways, depending on the settings of the RTR.RTRCFG.IS and
RTR.RTRCFG.IP bits:
1. All interrupt codes have priority over all interrupt acknowledgement codes / extended interrupt
codes (RTR.RTRCFG.IP = 0), and the interrupt numbers are serviced through a round-robin scheme
(RTR.RTRCFG.IS = 0). This is the default service scheme after reset / power-up.
2. All interrupt codes have priority over all interrupt acknowledgement codes / extended interrupt
codes (RTR.RTRCFG.IP = 0), and the interrupt numbers are serviced with priority to lower interrupt
numbers (RTR.RTRCFG.IS = 1).
3. All interrupt acknowledgement codes / extended interrupt codes have priority over all interrupt
codes (RTR.RTRCFG.IP = 1), and the interrupt numbers are serviced through a round-robin scheme
(RTR.RTRCFG.IS = 0).
4. All interrupt acknowledgement codes / extended interrupt codes have priority over all interrupt
codes (RTR.RTRCFG.IP = 1), and the interrupt numbers are serviced with priority to lower interrupt
numbers (RTR.RTRCFG.IS = 1).
When a distributed interrupt code has been selected from the queue, it is forwarded to all ports (except
the port it was received on) that has interrupt distribution enabled (RTR.PCTRL.IC = 1), and that has
enabled transmission of interrupt codes or interrupt acknowledgement codes / extended interrupt
codes (RTR.PCTRL2.IT and RTR.PCTRL2.AT respectively).
17.2.18.2 Interrupt distribution timers
Each interrupt number has two corresponding timers, called the ISR timer, and ISR change timer:
The ISR timer is started and reloaded with the value from the RTR.ISRTIMER register each time a
received interrupt code sets the corresponding RTR.ISR0 or RTR.ISR1 bit to 1. If an interrupt
acknowledgement code is received, the corresponding ISR timer is stopped. If the ISR timer expires
before an interrupt acknowledgement code is received, the corresponding bit in the RTR.ISR0 or
RTR.ISR1 register is cleared. The use of ISR timers is always enabled. In the mode with interrupt
acknowledgement codes, the purpose of the timers is to recover from situations where an interrupt
acknowledgement code is lost. In the mode with 64 interrupt codes, the purpose of the ISR timers is to
limit the rate of which interrupt codes are forwarded. It is important to configure the reload value for
the ISR timer correctly. In the mode with interrupt acknowledgement codes, the reload value shall not
be less than the worst propagation delay for the interrupt code, plus the maximum delay in the interrupt handler, plus the worst propagation delay for the interrupt acknowledgement code. In the mode
with 64 interrupt codes, the reload value shall not be less than the worst propagation delay for the
interrupt code.
The ISR change timers are timers that optionally can be used to control the minimum delay between
two consecutive changes to the same RTR.ISR bit. The purpose of the timers is to protect against
unexpected occurrences of interrupt codes / interrupt acknowledgement codes that could occur, for
GR740-UM-DS, Jun 2016, Version 1.4
173
www.cobham.com/gaisler
GR740
example, due to a network malfunction or a babbling idiot. If the use of ISR change timers is enabled
(RTR.RTRCFG.IC = 1), then the ISR change timer for an RTR.ISR bit is started and reloaded with the
value from the RTR.ISRCTIMER register each time a received interrupt code / interrupt acknowledgement code makes the RTR.ISR bit change value. Until the timer has expired, the corresponding
RTR.ISR bit is not allowed to change value, and any received interrupt code / interrupt acknowledgement code with that interrupt number is discarded. In the mode with 64 interrupt codes, the ISR
change timers should be disabled.
17.2.18.3 interrupt code generation
In addition to distributing interrupt codes received on the ports, the router can also generate an interrupt code when an internal error event occurs. In addition to the errors described in 17.2.20 the router
can also be configured to send an interrupt code when a SpaceWire port’s link interface enters runstate.
Everything in sections 17.2.18.1 and 17.2.18.2 also applies when the interrupt code is generated by
the router. The only difference is that an interrupt code generated by the router will not be discarded if
it is not allowed to be distributed. Instead, the interrupt code will be distributed later, as soon as it is
allowed. The only time an interrupt code generated by the router is not distributed is if the bits in
RTR.PIP are cleared by software before the interrupt code is allowed to be sent.
The interrupt code generation is controlled through the RTR.ICODEGEN. RTR.IMASK,
RTR.IPMASK, and RTR.PIP registers, and in addition to the enable / disable bit (RTR.ICODEGEN.EN), the following features are available:
•
Each port has a corresponding bit in the RTR.IPMASK register in order to control whether or not
an error on that specific port should generate an interrupt.
•
The different error types have there own mask bit in the RTR.IMASK register in order to control
whether or not that specific error should generate an interrupt.
•
The interrupt number to use for a generated interrupt code is programmable through the
RTR.ICODEGEN.IN field.
•
The generated interrupt code can be configured to be either level type, or edge type. Level type
means that a new interrupt code will be sent as long as any bit in the RTR.PIP register is set.
Edge type means that a new interrupt code will only be sent when a new error event occurs, i.e an
RTR.PIP bit toggles from 0 to 1. The type is selected by the RTR.ICODEGEN.IT bit.
•
A timer can be enabled through the RTR.ICODEGEN.TE bit. This timer controls the minimum
time between a received interrupt acknowledgement code and the distribution of a new generated
interrupt code. The timer is started and reloaded with the value from the RTR.AITIMER register
when an interrupt acknowledgement code is received, if the router was the source of the corresponding interrupt code. Until the timer has expired, a new generated interrupt code will not be
distributed. The reload value must not be less than the worst propagation delay for the interrupt
acknowledgement code.
•
Through the RTR.ICODEGEN.AH and RTR.ICODEGEN.UA bits the router can be configured
to, upon receiving an interrupt acknowledgement code, or the when the ISR timer expires, automatically clear the RTR.PIP bits that were set when the interrupt code was generated.
17.2.19 Auxiliary time-code / distributed interrupt code interface
The router provides an auxiliary time-code / distributed interrupt code interface that is connected
internally to the SpaceWire TDP core. The rules that determine whether a distributed interrupt code
received on the auxiliary interface should be forwarded to the ports are the same as when a distributed
interrupt code is received from any other port. However, for time-codes, the value received on the
auxiliary interface is always forwarded, and the router’s internal time-count value is updated. Note
that time-codes and distributed interrupt codes received through the auxiliary interface is not forwarded back out onto the interface itself. Just as for the router ports, the auxiliary interface has enable
GR740-UM-DS, Jun 2016, Version 1.4
174
www.cobham.com/gaisler
GR740
/ disable bits for time-codes (RTR.RTRCFG.AT) and interrupt codes (RTR.RTRCFG.AI), which need
to be set high in order for respective code to be transmitted / received on the interface.
See section 36 for more information regarding the SpaceWire Time Distribution Protocol.
17.2.20 Error detection and reporting
The router can detect and report the following errors:
•
SpaceWire link errors: parity error, disconnect error, escape error, credit error (see [SPW] for
definition of these errors).
•
Packet spill due to: timeout (see section 17.2.15), packet length truncation (see section 17.2.16),
time code / distributed interrupt code truncation (see section 17.2.21.1), and spill-if-not-ready
feature (see section 17.2.10)
•
Invalid address error (see section 17.2.12)
•
Memory errors in the memory used for the routing table, RMAP command buffers, and port
transmit and receive FIFO (see section 17.2.22)
•
RMAP errors (see section 17.5.1)
•
SpaceWire Plug-and-Play errors (see section 17.5.4)
Each error type has corresponding status bits in respective RTR.PSTS register (RTR.PSTSCFG for
the configuration port). Common for all the status bits is that they are set when the error is detected,
and stay set until they are cleared manually.
The router can also be configured to distribute an interrupt code when an error is detected. See section
17.2.18 for more details.
17.2.21 SpaceWire-D support
17.2.21.1 Time-code and distributed interrupt code truncation
The router supports truncation of packets when it receives valid time-codes or distributed interrupt
codes. A time-code is considered valid when the value equals the internal time count plus one (modulo 64). A distributed interrupt code is considered valid if the corresponding ISR bit is flipped due to
reception of the code. The feature can be enabled individually for each port by setting the corresponding RTR.PCTRL.TS bit to 1. A filter, allowing only certain time-codes and distributed interrupt codes
to spill packets, can also be configured individually for each port (see RTR.PCTRL2 register).
If a packet transfer is ongoing when a valid time-code or distributed interrupt code is received, and the
code matches the filter in RTR.PCTRL2, the rest of the packet is spilled, and an EEP is written to the
FIFO of the output port(s).
Time-code and distributed interrupt code truncation can also be enabled for port 0. In that case, it is
the RMAP / SpaceWire Plug-and-Play reply packet that is spilled.
17.2.22 On-chip memories
When an uncorrectable error is detected in the memories used for the routing table and port mapping
when a packet is being routed, the packet will be discarded.
Uncorrectable errors in the porst’ FIFO memories are not handled since they only affect the contents
of the routed packet not the operation of the router itself. These type of errors should be caught by
CRC checks if used in the packet.
The RTR.PSTS.ME bit for the ports is only usable for detecting errors and statistics since there is no
need to correct the error manually since the packet has already been routed when it is detected. The
RTR.PSTSCFG.ME indication for the routing table and port setup registers can be used for starting a
GR740-UM-DS, Jun 2016, Version 1.4
175
www.cobham.com/gaisler
GR740
scrubbing operation if detected. There is also an option of having automatic scrubbing (see section
17.2.22.1).
For the FIFOs used when sending data from an AMBA port there is an option to spill the ongoing
packet when an error occur. This is controlled by the RTR.AMBACTRL.ME bit.
17.2.22.1 Autoscrub
With autoscrubbing the routing table and port setup registers will be periodically read and rewritten.
This is done to prevent buildup of SEUs to cause an uncorrectable error in the memories. It will run in
the background and has no impact on routing table lookup for traffic but can delay configuration
accesses with two cycles.
The scrubber starts at address 0 and simultaneously writes one location in the port setup memory and
the routing table memory. It then waits for a timeout period until it writes the next word. Eventually
the last location is reached and the process starts over from address 0.
The period between each word refresh is approximately 226 core clock cycles. The scrubber uses a
free slot when data traffic does not need to perform a table lookup to read and write the memories
which causes a small undeterminism in the period.
17.3
SpaceWire ports
When a port is configured as a SpaceWire link it consists of a SpaceWire codec with FIFO interfaces.
The SpaceWire encoder-decoder implements an encoder-decoder compliant to the SpaceWire standard [SPW]. It provides a generic host-interface consisting of control signals, status signals, timecode interface and 9-bit wide data buses connecting to a pair of FIFOs.
Transmitter outputs are Single Data Rate (SDR). The receiver recovers data using DDR sampling.
TXCLK
D(1:0)
TRANSMITTER
S(1:0)
TRANSMITTER
FIFO
CONTROL SIGNALS
LINK INTERFACE
FSM
D0
STATUS SIGNALS
D
PHY
S0
RECEIVER0
RECEIVER
FIFO
DV
Figure 19. Block diagram
17.3.1 Codec overview
A block diagram of the internal structure of the core can be found in figure 19. It consists of the
receiver, transmitter and the link interface FSM. They handle communication on the SpaceWire network. The PHY block contains the data recovery logic, this implementation uses sampling.
Time-codes are transmitted through a signal interface as specified in the SpaceWire standard [SPW].
GR740-UM-DS, Jun 2016, Version 1.4
176
www.cobham.com/gaisler
GR740
17.3.1.1 Link-interface FSM
The link-interface FSM controls the link interface (a more detailed description is found in the [SPW]).
The low-level protocol handling (the signal and character level of the SpaceWire standard [SPW]) is
handled by the transmitter and receiver while the FSM handles the exchange level.
The link-interface FSM is controlled through the control signals. The link can be disabled through the
link disabled signal, which depending on the current state, either prevents the link-interface from
reaching the started state or forces it to the error-reset state. When the link is not disabled, the link
interface FSM is allowed to enter the started state when either the link start signal is asserted or when
a NULL character has been received and the autostart signal is asserted.
The current state of the link-interface determines which type of characters are allowed to be transmitted which together with the requests made from the host interface determine what character will be
sent.
Time-codes are sent when the FSM is in the run-state and a request is made through an internal timeinterface.
When the link-interface is in the connecting- or run-state it is allowed to send FCTs. FCTs are sent
automatically by the link-interface when possible. This is done based on the maximum value of 56 for
the outstanding credit counter and the currently free space in the receiver FIFO. FCTs are sent as long
as the outstanding counter is less than or equal to 48 and there are at least 8 more empty FIFO entries
than the counter value.
N-Chars are sent in the run-state when they are available from the transmitter FIFO and there are
credits available. NULLs are sent when no other character transmission is requested or the FSM is in
a state where no other transmissions are allowed.
The credit counter (incoming credits) is automatically increased when FCTs are received and
decreased when N-Chars are transmitted. Received N-Chars are stored to the receiver N-Char FIFO
while received Time-codes are handled by the time-interface.
17.3.1.2 Transmitter
The state of the FSM, credit counters, requests from the time-interface and requests from the transmitter FIFO are used to decide the next character to be transmitted. The type of character and the character itself (for N-Chars and Time-codes) to be transmitted are presented to the low-level transmitter
which is located in a separate clock-domain.
The transmitter logic in the host clock domain decides what character to send next and sets the proper
control signal and presents any needed character to the low-level transmitter as shown in figure 20.
The transmitter sends the requested characters and generates parity and control bits as needed. If no
requests are made from the host domain, NULLs are sent as long as the transmitter is enabled. Most of
the signal and character levels of the SpaceWire standard [SPW] is handled in the transmitter.
D
S
Send Time-code
Send FCT
Send NChar
Time-code[7:0]
NChar[8:0]
Transmitter
Transmitter Clock Domain
Host Clock Domain
Figure 20. Schematic of the link interface transmitter.
GR740-UM-DS, Jun 2016, Version 1.4
177
www.cobham.com/gaisler
GR740
17.3.1.3 Receiver
The receiver detects connections from other nodes and receives characters as a bit stream recovered
from the data and strobe signals by the PHY module which presents it as a data and data-valid signal.
Both the receiver and PHY are located in a separate clock domain which runs on a clock generated by
the PHY.
The receiver is activated as soon as the link-interface leaves the error reset state. Then after a NULL is
received it can start receiving any characters. It detects parity, escape and credit errors which causes
the link interface to enter the error-reset state. Disconnections are handled in the link-interface part in
the tx clock domain because no receiver clock is available when disconnected.
Received characters are flagged to the host domain and the data is presented in parallel form. The
interface to the host domain is shown in figure 21. L-Chars are the handled automatically by the host
domain link-interface part while all N-Chars are stored in the receiver FIFO for further handling. If
two or more consecutive EOPs/EEPs are received all but the first are discarded.
D
Receiver
DV
Got Time-code
Got FCT
Got EOP
Got EEP
Got NChar
Time-code[7:0]
NChar[7:0]
Receiver Clock Domain
Host Clock Domain
Figure 21. Schematic of the link interface receiver.
17.3.2 Setting link-rate
The initialization divisor register (RTR.IDIV) determines the link-rate during initialization (all states
up to and including the connecting-state) for all SpaceWire ports. The register is also used to calculate
the link interface FSM timeouts for all SpaceWire ports (6.4 us and 12.8 us, as defined in the SpaceWire standard [SPW]). The RTR.IDIV register should always be set so that a 10 Mbit/s link-rate is
achieved during initialization. In that case the timeout values will also be calculated correctly.
To achieve a 10 Mbit/s link-rate, the RTR.IDIV register should be set according to the following formula:
RTR.IDIV = (<frequency in MHz of internal SpaceWire clock> / 10) - 1
The link-rate in run-state can be controlled individually per SpaceWire port with the run-state divisor
located in each port’s control register (RTR.PCTRL.RD field). The link-rate in run-state is calculated
according to the following formula:
<link-rate in Mbits/s> = <frequency in MHz of internal SpaceWire clock> / (RTR.PCTRL.RD+1)
The value in RTR.PCTRL.RD only affects the link-rate in run-state, and does not affect the 6.4 us or
12.8 us timeouts values.
17.4
AMBA ports
The AMBA ports consists of what is basically a Cobham Gaisler GRSPW2 core with the SpaceWire
codec removed. The same drivers that are provided for the GRSPW2 core can be used for each
AMBA port on the router. Only an additional driver is needed which handles the setup of all the registers on the configuration port.
GR740-UM-DS, Jun 2016, Version 1.4
178
www.cobham.com/gaisler
GR740
17.4.1 Overview
The router’s AMBA ports are configured through a set of registers accessed through an APB interface. Data is transferred through one to four DMA channels using an AHB master interface.
TO ROUTER
SWITCH MATRIX
SEND
FSM
TRANSMITTER
FSM
RMAP
TRANSMITTER
TRANSMITTER
DMA ENGINE
AHB
MASTER INTERFACE
RECEIVER
DMA ENGINE
FROM ROUTER
SWITCH MATRIX
RMAP
RECEIVER
RECEIVER
AHB FIFO
N-CHAR
FIFO
RECEIVER DATA
PARALLELIZATION
REGISTERS
APB
INTERFACE
Figure 22. Block diagram of the Router DMA port
17.4.2 Operation
The main sub-blocks of the router AHB interfaces are the DMA engines, the RMAP target and the
AMBA interface. A block diagram of the internal structure can be found in figure 22.
The AMBA interface is divided into the AHB master interface and the APB interface. The DMA
engines have FIFO interfaces to the router switch matrix. These FIFOs are used to transfer N-Chars
between the AMBA bus and the other ports in the router.
The RMAP target handles incoming packets which are determined to be RMAP commands instead of
the receiver DMA engine. The RMAP command is decoded and if it is valid, the operation is performed on the AHB bus. If a reply was requested it is automatically transmitted back to the source by
the RMAP transmitter.
The core is controlled by writing to a set of user registers through the APB interface and a set of signals. The different sub-modules are discussed in further detail in later sections.
17.4.2.1 Protocol support
The AMBA port only accepts packets with a valid destination address in the first received byte. Packets with address mismatch will be silently discarded (except in promiscuous mode which is covered in
section 17.4.4.10).
The second byte is sometimes interpreted as a protocol ID a described hereafter. The RMAP protocol
(ID=0x1) is the only protocol handled separately in hardware while other packets are stored to a DMA
channel. If the RMAP target is present and enabled all RMAP commands will be processed, executed
and replied automatically in hardware. Otherwise RMAP commands are stored to a DMA channel in
the same way as other packets. RMAP replies are always stored to a DMA channel. More information
on the RMAP protocol support is found in section 17.4.6 (note that this RMAP target is different from
the one in the configuration port). When the RMAP target is not present or disabled, there is no need
to include a protocol ID in the packets and the data can start immediately after the address.
All packets arriving with the extended protocol ID (0x00) are stored to a DMA channel. This means
that the hardware RMAP target will not work if the incoming RMAP packets use the extended protocol ID. Note also that packets with the reserved extended protocol identifier (ID = 0x000000) are not
ignored by the AMBA port. It is up to the client receiving the packets to ignore them.
GR740-UM-DS, Jun 2016, Version 1.4
179
www.cobham.com/gaisler
GR740
When transmitting packets, the address and protocol-ID fields must be included in the buffers from
where data is fetched. They are not automatically added by the AMBA port DMA engine.
Figure 23 shows the packet types accepted by the port. The port also allows reception and transmission with extended protocol identifiers but without support for RMAP CRC calculations and the
RMAP target.
Addr ProtID D0
D1
D2
D3
..
Dn-2 Dn-1 EOP
D0
D2
D3
D4
..
Dm-2 Dm-1 EOP
Addr
D1
Figure 23. The SpaceWire packet types supported by the port.
17.4.3 Time-Code / distributed interrupt interface
17.4.3.1 Sending and receiving Time-Codes
To transmit a Time-Code through the register interface of an AMBA port the RTR.AMBACTRL.TI
bit should be written to 1. When the bit is written the AMBA port’s current time value
(RTR.AMBATC.TIMECNT field) is incremented, and a Time-Code consisting of the new time value
together with the current control flags (RTR.AMBATC.TCTRL field) is sent. The RTR.AMBACTRL.TI bit will stay high until the Time-Code has been transmitted. Note that whether or not a TimeCode is forwarded to any other port after it has been sent by an AMBA port depends on the settings
explained in 17.2.17.
A Time-Code that is received by an AMBA port will be stored in the port’s RTR.AMBATC register.
There is also a possibility to generate AMBA interrupts, and tick out pulses that are router to the
GPTIMER units for time latching (see section 5.9). This is controlled through the RTR.AMBACTRL
and RTR.AMBATC registers. See section 17.4.8 for details about these registers.
17.4.3.2 Sending and receiving distributed interrupts
To transmit a distributed interrupt code through the register interface of an AMBA port, the
RTR.AMBAINTCTRL.II bit in should be written to 1. When the bit is written the value of the
RTR.AMBAINTCTRL.TXINT field determine which distributed interrupt code that will be sent.
Note that whether or not a distributed interrupt code is forwarded to any other port after it has been
sent by an AMBA port depends on the settings of the router and the state of that specific interrupt in
the network. See 17.2.18 for details.
A distributed interrupt can also be configured to be generated automatically by the AMBA port upon
certain events. This is controlled through the RTR.AMBAINTCTRL and RTR.AMBADMACTRL
registers. See section 17.4.8 for details about these registers and features.
Distributed interrupt codes that are received by an AMBA port can be programmed to generate
AMBA interrupts as well as tick out pulses that are routed to the GPTIMER units for time latching
(see section 5.9). See section 17.4.8 for details about these features.
17.4.4 Receiver DMA channels
The receiver DMA engine handles reception of data from the SpaceWire network to different DMA
channels.
17.4.4.1 Address comparison and channel selection
Packets are received to different channels based on the address and whether a channel is enabled or
not. When the receiver N-Char FIFO contains one or more characters, N-Chars are read by the
receiver DMA engine. The first character is interpreted as the logical address and is compared with
GR740-UM-DS, Jun 2016, Version 1.4
180
www.cobham.com/gaisler
GR740
the addresses of each channel starting from 0. The packet will be stored to the first channel with an
matching address. The complete packet including address and protocol ID but excluding EOP/EEP is
stored to the memory address pointed to by the descriptors (explained later in this section) of the
channel.
Each SpaceWire address register has a corresponding mask register. Only bits at an index containing a
zero in the corresponding mask register are compared. This way a DMA channel can accept a range of
addresses. There is a default address register which is used for address checking in all implemented
DMA channels that do not have separate addressing enabled and for RMAP commands in the RMAP
target. With separate addressing enabled the DMA channels’ own address/mask register pair is used
instead.
If an RMAP command is received it is only handled by the target if the default address register
(including mask) matches the received address. Otherwise the packet will be stored to a DMA channel
if one or more of them has a matching address. If the address does not match neither the default
address nor one of the DMA channels’ separate register, the packet is still handled by the RMAP target if enabled since it has to return the invalid address error code. The packet is only discarded (up to
and including the next EOP/EEP) if an address match cannot be found and the RMAP target is disabled.
Packets, other than RMAP commands, that do not match neither the default address register nor the
DMA channels’ address register will be discarded. Figure 24 shows a flowchart of packet reception.
At least 2 non EOP/EEP N-Chars needs to be received for a packet to be stored to the DMA channel
unless the promiscuous mode is enabled in which case 1 N-Char is enough. If it is an RMAP packet
with hardware RMAP enabled 3 N-Chars are needed since the command byte determines where the
packet is processed. Packets smaller than these sizes are discarded.
17.4.4.2 Basic functionality of a channel
Reception is based on descriptors located in a consecutive area in memory that hold pointers to buffers where packets should be stored. When a packet arrives at the port the channel which should
receive it is first determined as described in the previous section. A descriptor is then read from the
channels’ descriptor area and the packet is stored to the memory area pointed to by the descriptor.
Lastly, status is stored to the same descriptor and increments the descriptor pointer to the next one.
The following sections will describe DMA channel reception in more detail.
17.4.4.3 Setting up the port for reception
A few registers need to be initialized before reception to a channel can take place. The DMA channel
has a maximum length register which sets the maximum packet size in bytes that can be received to
this channel. Larger packets are truncated and the excessive part is spilled. If this happens an indication will be given in the status field of the descriptor. The minimum value for the receiver maximum
length field is 4 and the value can only be incremented in steps of four bytes up to the maximum value
33554428. If the maximum length is set to zero the receiver will not function correctly.
Either the default address register or the channel specific address register (the accompanying mask
register must also be set) needs to be set to hold the address used by the channel. A control bit in the
DMA channel control register determines whether the channel should use default address and mask
registers for address comparison or the channel’s own registers. Using the default register the same
address range is accepted as for other channels with default addressing and the RMAP target while the
separate address provides the channel its own range. If all channels use the default registers they will
accept the same address range and the enabled channel with the lowest number will receive the
packet.
Finally, the descriptor table and control register must be initialized. This will be described in the two
following sections.
GR740-UM-DS, Jun 2016, Version 1.4
181
www.cobham.com/gaisler
GR740
Start Reception
Receive
2 bytes
rmap enabled
and
pid =1 and
defaddr*!defmask =
rxaddr*!defmask
No
Yes
Receive
1 byte
No
Set DMA channel
number to 0
RMAP command
Increment
channel number
Yes
No
Yes
No
Last DMA channel
Channel enabled
No
Yes
Separate addressing
RMAP enabled
No
Yes
No
dma(n).addr*
!dma(n).mask=
rxaddr*!dma(n).mask
defaddr*!defmask =
rxaddr*!defmask
No
Yes
Yes
Process RMAP
command
Store packet to
DMA channel
Discard packet
Figure 24. Flow chart of packet reception (promiscuous mode disabled).
17.4.4.4 Setting up the descriptor table address
The port reads descriptors from an area in memory pointed to by the receiver descriptor table address
register. The register consists of a base address and a descriptor selector. The base address points to
GR740-UM-DS, Jun 2016, Version 1.4
182
www.cobham.com/gaisler
GR740
the beginning of the area and must start on a 1024 bytes aligned address. It is also limited to be 1024
bytes in size which means the maximum number of descriptors is 128 since the descriptor size is 8
bytes.
The descriptor selector points to individual descriptors and is increased by 1 when a descriptor has
been used. When the selector reaches the upper limit of the area it wraps to the beginning automatically. It can also be set to wrap at a specific descriptor before the upper limit by setting the wrap bit in
the descriptor. The idea is that the selector should be initialized to 0 (start of the descriptor area) but it
can also be written with another 8 bytes aligned value to start somewhere in the middle of the area. It
will still wrap to the beginning of the area.
If one wants to use a new descriptor table the receiver enable bit in the corresponding AMBA port
DMA contol/status register has to be cleared first. When the RX active bit in the same register is
cleared it is safe to update the descriptor table register. When this is finished and descriptors are
enabled the receiver enable bit can be set again.
17.4.4.5 Enabling descriptors
As mentioned earlier one or more descriptors must be enabled before reception can take place. Each
descriptor is 8 byte in size and the layout can be found in the tables below. The descriptors should be
written to the memory area pointed to by the receiver descriptor table address register. When new
descriptors are added they must always be placed after the previous one written to the area. Otherwise
they will not be noticed.
A descriptor is enabled by setting the address pointer to point at a location where data can be stored
and then setting the enable bit. The WR bit can be set to cause the selector to be set to zero when
reception has finished to this descriptor. IE should be set if an interrupt is wanted when the reception
has finished. The DMA control register interrupt enable bit must also be set for an interrupt to be generated.
Table 170. RXDMA receive descriptor word 0 (address offset 0x0)
31 30 29 28 27 26 25 24
0
TR DC HC EP IE WR EN
PACKETLENGTH
31
Truncated (TR) - Packet was truncated due to maximum length violation.
30
Data CRC (DC) - 1 if a CRC error was detected for the data and 0 otherwise.
29
Header CRC (HC) - 1 if a CRC error was detected for the header and 0 otherwise.
28
EEP termination (EP) - This packet ended with an Error End of Packet character.
27
Interrupt enable (IE) - If set, an interrupt will be generated when a packet has been received if the
receive interrupt enable bit in the DMA channel control register is set.
26
Wrap (WR) - If set, the next descriptor used by the GRSPW will be the first one in the descriptor
table (at the base address). Otherwise the descriptor pointer will be increased with 0x8 to use the
descriptor at the next higher memory location. The descriptor table is limited to 1 KiB in size and the
pointer will be automatically wrap back to the base address when it reaches the 1 KiB boundary.
25
Enable (EN) - Set to one to activate this descriptor. This means that the descriptor contains valid control values and the memory area pointed to by the packet address field can be used to store a packet.
24: 0
Packet length (PACKETLENGTH) - The number of bytes received to this buffer. Only valid after
EN has been set to 0 by the GRSPW.
Table 171. RXDMA receive descriptor word 1 (address offset 0x4)
31
0
PACKETADDRESS
31: 0
Packet address (PACKETADDRESS) - The address pointing at the buffer which will be used to store
the received packet.
GR740-UM-DS, Jun 2016, Version 1.4
183
www.cobham.com/gaisler
GR740
17.4.4.6 Setting up the DMA control register
The final step to receive packets is to set the control register in the following steps: The receiver must
be enabled by setting the rxen bit in the DMA control register (see section 17.4.8). This can be done
anytime and before this bit is set nothing will happen. The rxdescav bit in the DMA control register is
then set to indicate that there are new active descriptors. This must always be done after the descriptors have been enabled or the port might not notice the new descriptors. More descriptors can be activated when reception has already started by enabling the descriptors and writing the rxdescav bit.
When these bits are set reception will start immediately when data is arriving.
17.4.4.7 The effect to the control bits during reception
When the receiver is disabled all packets going to the DMA-channel are discarded if the packet’s
address does not fall into the range of another DMA channel. If the receiver is enabled and the address
falls into the accepted address range, the next state is entered where the rxdescav bit is checked. This
bit indicates whether there are active descriptors or not and should be set by the external application
using the DMA channel each time descriptors are enabled as mentioned above. If the rxdescav bit is
‘0’ and the nospill bit is ‘0’ the packets will be discarded. If nospill is ’1’ the AMBA port waits until
rxdescav is set and the characters are kept in the N-Char fifo during this time. If the fifo becomes full
further N-char transmissions are inhibited by stopping the transmission of FCTs.
When rxdescav is set the next descriptor is read and if enabled the packet is received to the buffer. If
the read descriptor is not enabled, rxdescav is set to ‘0’ and the packet is spilled depending on the
value of nospill.
The receiver can be disabled at any time and will stop packets from being received to this channel. If
a packet is currently received when the receiver is disabled the reception will still be finished. The
rxdescav bit can also be cleared at any time. It will not affect any ongoing receptions but no more
descriptors will be read until it is set again. Rxdescav is also cleared by the port when it reads a disabled descriptor.
17.4.4.8 Status bits
When the reception of a packet is finished the enable bit in the current descriptor is set to zero. When
enable is zero, the status bits are also valid and the number of received bytes is indicated in the length
field. The DMA control register contains a status bit which is set each time a packet has been
received. The port can also be made to generate an interrupt for this event.
The RMAP CRC calculation is always active for all received packets and all bytes except the EOP/
EEP are included. The packet is always assumed to be an RMAP packet and the length of the header
is determined by checking byte 3 which should be the command field. The calculated CRC value is
then checked when the header has been received (according to the calculated number of bytes) and if
it is non-zero the HC bit is set indicating a header CRC error.
The CRC value is not set to zero after the header has been received, instead the calculation continues
in the same way until the complete packet has been received. Then if the CRC value is non-zero the
DC bit is set indicating a data CRC error. This means that the port can indicate a data CRC error even
if the data field was correct when the header CRC was incorrect. However, the data should not be used
when the header is corrupt and therefore the DC bit is unimportant in this case. When the header is not
corrupted the CRC value will always be zero when the calculation continues with the data field and
the behaviour will be as if the CRC calculation was restarted
If the received packet is not of RMAP type the header CRC error indication bit cannot be used. It is
still possible to use the DC bit if the complete packet is covered by a CRC calculated using the RMAP
CRC definition. This is because the port does not restart the calculation after the header has been
received but instead calculates a complete CRC over the packet. Thus any packet format with one
CRC at the end of the packet calculated according to RMAP standard [RMAP] can be checked using
the DC bit.
GR740-UM-DS, Jun 2016, Version 1.4
184
www.cobham.com/gaisler
GR740
If the packet is neither of RMAP type nor of the type above with RMAP CRC at the end, then both the
HC and DC bits should be ignored.
17.4.4.9 Error handling
If an AHB error occurs during reception the current packet is spilled up to and including the next
EEP/EOP and then the currently active channel is disabled and the receiver enters the idle state. A bit
in the channels control/status register is set to indicate this condition.
17.4.4.10 Promiscuous mode
The port supports a promiscuous mode where all the data received is stored to the first DMA channel
enabled regardless of the node address and possible early EOPs/EEPs. This means that all non-eop/
eep N-Chars received will be stored to the DMA channel. The AMBA port DMA RX maximum
length register is still checked and packets exceeding this size will be truncated.
RMAP commands will still be handled by RMAP hardware target when promiscuous mode is
enabled, if the RMAP enable bit in the AMBA port DMA control/status register is set. If the RMAP
enable bit is cleared, RMAP commands will also be stored to a DMA channel.
17.4.5 Transmitter DMA channels
The transmitter DMA engine handles transmission of data from the DMA channels to the SpaceWire
network. Each receive channel has a corresponding transmit channel which means there can be up to
4 transmit channels. It is however only necessary to use a separate transmit channel for each receive
channel if there are also separate entities controlling the transmissions. The use of a single channel
with multiple controlling entities would cause them to corrupt each other’s transmissions. A single
channel is more efficient and should be used when possible.
Multiple transmit channels with pending transmissions are arbitrated in a round-robin fashion.
17.4.5.1 Basic functionality of a channel
A transmit DMA channel reads data from the AHB bus and stores them in the transmitter FIFO for
transmission on the SpaceWire network. Transmission is based on the same type of descriptors as for
the receiver and the descriptor table has the same alignment and size restrictions. When there are new
descriptors enabled the port reads them and transfer the amount data indicated.
17.4.5.2 Setting up the core for transmission
Four steps need to be performed before transmissions can be done with the port. First the link interface must be enabled and started by writing the appropriate value to the ctrl register. Then the address
to the descriptor table needs to be written to the transmitter descriptor table address register and one or
more descriptors must also be enabled in the table. Finally, the txen bit in the DMA control register is
written with a one which triggers the transmission. These steps will be covered in more detail in the
next sections.
17.4.5.3 Enabling descriptors
The descriptor table address register works in the same way as the receiver’s corresponding register
which was covered in section 17.4.4. The maximum size is 1024 bytes as for the receiver but since the
descriptor size is 16 bytes the number of descriptors is 64.
To transmit packets one or more descriptors have to be initialized in memory which is done in the following way: The number of bytes to be transmitted and a pointer to the data has to be set. There are
two different length and address fields in the transmit descriptors because there are separate pointers
for header and data. If a length field is zero the corresponding part of a packet is skipped and if both
are zero no packet is sent. The maximum header length is 255 bytes and the maximum data length is
16 MiB - 1. When the pointer and length fields have been set the enable bit should be set to enable the
GR740-UM-DS, Jun 2016, Version 1.4
185
www.cobham.com/gaisler
GR740
descriptor. This must always be done last. The other control bits must also be set before enabling the
descriptor.
The transmit descriptors are 16 bytes in size so the maximum number in a single table is 64. The different fields of the descriptor together with the memory offsets are shown in the tables below.
The HC bit should be set if RMAP CRC should be calculated and inserted for the header field and
correspondingly the DC bit should be set for the data field. The header CRC will be calculated from
the data fetched from the header pointer and the data CRC is generated from data fetched from the
data pointer. The CRCs are appended after the corresponding fields. The NON-CRC bytes field is set
to the number of bytes in the beginning of the header field that should not be included in the CRC calculation.
The CRCs are sent even if the corresponding length is zero, but when both lengths are zero no packet
is sent not even an EOP.
17.4.5.4 Starting transmissions
When the descriptors have been initialized, the transmit enable bit in the DMA control register has to
be set to tell the port to start transmitting. New descriptors can be activated in the table on the fly
(while transmission is active). Each time a set of descriptors is added the transmit enable bit in the
corresponding AMBA port DMA control/status register should be set. This has to be done because
each time the core encounters a disabled descriptor this register bit is set to 0.
Table 172. TXDMA transmit descriptor word 0 (address offset 0x0)
31
18 17 16 15 14 13 12 11
RESERVED
DC HC RE IE WR EN
8
7
NONCRCLEN
0
HEADERLEN
31: 18
RESERVED
17
Append data CRC (DC) - Append CRC calculated according to the RMAP specification after the
data sent from the data pointer. The CRC covers all the bytes from this pointer. A null CRC will
be sent if the length of the data field is zero.
16
Append header CRC (HC) - Append CRC calculated according to the RMAP specification after the
data sent from the header pointer. The CRC covers all bytes from this pointer except a number of
bytes in the beginning specified by the non-crc bytes field. The CRC will not be sent if the header
length field is zero.
15
RESERVED
14
Interrupt enable (IE) - If set, an interrupt will be generated when the packet has been transmitted and
the transmitter interrupt enable bit in the DMA control register is set.
13
Wrap (WR) - If set, the descriptor pointer will wrap and the next descriptor read will be the first one
in the table (at the base address). Otherwise the pointer is increased with 0x10 to use the descriptor at
the next higher memory location.
12
Enable (EN) - Enable transmitter descriptor. When all control fields (address, length, wrap and crc)
are set, this bit should be set. While the bit is set the descriptor should not be touched since this
might corrupt the transmission. The GRSPW clears this bit when the transmission has finished.
11: 8
Non-CRC bytes (NONCRCLEN)- Sets the number of bytes in the beginning of the header which
should not be included in the CRC calculation. This is necessary when using path addressing since
one or more bytes in the beginning of the packet might be discarded before the packet reaches its
destination.
7: 0
Header length (HEADERLEN) - Header Length in bytes. If set to zero, the header is skipped.
Table 173. TXDMA transmit descriptor word 1 (address offset 0x4)
31
0
HEADERADDRESS
GR740-UM-DS, Jun 2016, Version 1.4
186
www.cobham.com/gaisler
GR740
Table 173. TXDMA transmit descriptor word 1 (address offset 0x4)
31: 0
Header address (HEADERADDRESS) - Address from where the packet header is fetched. Does not
need to be word aligned.
Table 174. TXDMA transmit descriptor word 2 (address offset 0x8)
31
24 23
0
RESERVED
DATALEN
31: 24
RESERVED
23: 0
Data length (DATALEN) - Length of data part of packet. If set to zero, no data will be sent. If both
data- and header-lengths are set to zero no packet will be sent.
Table 175. TXDMA transmit descriptor word 3(address offset 0xC)
31
0
DATAADDRESS
31: 0
Data address (DATAADDRESS) - Address from where data is read. Does not need to be word
aligned.
17.4.5.5 The transmission process
When the transmit enable bit int the AMBA port DMA control/status register is set the port starts
reading descriptors immediately. The number of bytes indicated are read and transmitted. When a
transmission has finished, status will be written to the first field of the descriptor and a packet sent bit
is set in the DMA control register. If an interrupt was requested it will also be generated. Then a new
descriptor is read and if enabled a new transmission starts, otherwise the transmit enable bit is cleared
and nothing will happen until it is enabled again.
17.4.5.6 The descriptor table address register
The internal pointer which is used to keep the current position in the descriptor table can be read and
written through the APB interface. This pointer is set to zero during reset and is incremented each
time a descriptor is used. It wraps automatically when the 1024 bytes limit for the descriptor table is
reached or it can be set to wrap earlier by setting a bit in the current descriptor.
The descriptor table register can be updated with a new table anytime when no transmission is active.
No transmission is active if the transmit enable bit is zero and the complete table has been sent or if
the table is aborted (explained below). If the table is aborted one has to wait until the transmit enable
bit is zero before updating the table pointer.
17.4.5.7 Error handling
17.4.5.7.1Abort Tx
The DMA control register contains a bit called Abort TX which if set causes the current transmission
to be aborted, the packet is truncated and an EEP is inserted. This is only useful if the packet needs to
be aborted because of congestion on the SpaceWire network. If the congestion is on the AHB bus this
will not help (This should not be a problem since AHB slaves should have a maximum of 16 waitstates). The aborted packet will have its LE bit set in the descriptor. The transmit enable register bit is
also cleared and no new transmissions will be done until the transmitter is enabled again.
GR740-UM-DS, Jun 2016, Version 1.4
187
www.cobham.com/gaisler
GR740
17.4.5.7.2AHB error
When an AHB error is encountered during transmission the currently active DMA channel is disabled
and the transmitter goes to the idle mode. A bit in the DMA channel’s control/status register is set to
indicate this error condition and, if enabled, an interrupt will also be generated. Further error handling
depends on what state the transmitter DMA engine was in when the AHB error occurred. If the
descriptor was being read the packet transmission had not been started yet and no more actions need
to be taken.
If the AHB error occurs during packet transmission the packet is truncated and an EEP is inserted.
Lastly, if it occurs when status is written to the descriptor the packet has been successfully transmitted
but the descriptor is not written and will continue to be enabled (this also means that no error bits are
set in the descriptor for AHB errors).
The client using the channel has to correct the AHB error condition and enable the channel again. No
more AHB transfers are done again from the same unit (receiver or transmitter) which was active
during the AHB error until the error state is cleared and the unit is enabled again.
17.4.6 RMAP target
The Remote Memory Access Protocol (RMAP) is used to implement access to resources on the AHB
bus via the SpaceWire Link. Some common operations are reading and writing to memory, registers
and FIFOs. This section describes the target implementation.
17.4.6.1 Fundamentals of the protocol
RMAP is a protocol which is designed to provide remote access via a SpaceWire network to memory
mapped resources on a SpaceWire node. It has been assigned protocol ID 0x01. It provides three operations write, read and read-modify-write. These operations are posted operations which means that a
source does not wait for an acknowledge or reply. It also implies that any number of operations can be
outstanding at any time and that no timeout mechanism is implemented in the protocol. Time-outs
must be implemented in the user application which sends the commands. Data payloads of up to 16
MiB - 1 is supported in the protocol. A destination can be requested to send replies and to verify data
before executing an operation. A complete description of the protocol is found in the RMAP standard
[RMAP].
17.4.6.2 Implementation
The port includes a target for RMAP commands which processes all incoming packets with protocol
ID = 0x01, type field (bit 7 and 6 of the 3rd byte in the packet) equal to 01b and an address falling in
the range set by the default address and mask register. When such a packet is detected it is not stored
to the DMA channel, instead it is passed to the RMAP receiver.
The target implements all three commands defined in [RMAP], with some restrictions. Support is
only provided for 32-bit big-endian systems. This means that the first byte received is the msb in a
word. The target will not receive RMAP packets using the extended protocol ID which are always
dumped to the DMA channel.
The RMAP receiver processes commands. If they are correct and accepted the operation is performed
on the AHB bus and a reply is formatted. If an acknowledge is requested the RMAP transmitter automatically send the reply. RMAP transmissions have priority over DMA channel transmissions.
There is a user accessible destination key register which is compared to destination key field in
incoming packets. If there is a mismatch and a reply has been requested the error code in the reply is
set to 3. Replies are sent if and only if the ack field is set to ‘1’.
GR740-UM-DS, Jun 2016, Version 1.4
188
www.cobham.com/gaisler
GR740
When a failure occurs during a bus access the error code is set to 1 (General Error). There is predetermined order in which error-codes are set in the case of multiple errors in the core. It is shown in table
176.
Table 176.The order of error detection in case of multiple errors. The error detected first has number 1.
Detection Order
Error Code
Error
1
12
Invalid destination logical address
2
2
Unused RMAP packet type or command code
3
3
Invalid destination key
4
9
Verify buffer overrun
5
11
RMW data length error
6
10
Authorization failure
7*
1
General Error (AHB errors during non-verified writes)
8
5/7
Early EOP / EEP (if early)
9
4
Invalid Data CRC
10
1
General Error (AHB errors during verified writes or RMW)
11
7
EEP
12
6
Too Much Data
*The AHB error is not guaranteed to be detected before Early EOP/EEP or Invalid Data CRC. For very long accesses
the AHB error detection might be delayed causing the other two errors to appear first.
Read accesses are performed on the fly, that is they are not stored in a temporary buffer before transmitting. This means that the error code 1 will never be seen in a read reply since the header has
already been sent when the data is read. If the AHB error occurs the packet will be truncated and
ended with an EEP.
Errors up to and including Invalid Data CRC (number 8) are checked before verified commands. The
other errors do not prevent verified operations from being performed.
The details of the support for the different commands are now presented. All defined commands
which are received but have an option set which is not supported in this specific implementation will
not be executed and a possible reply is sent with error code 10.
17.4.6.3 Write commands
The write commands are divided into two subcategories when examining their capabilities: verified
writes and non-verified writes. Verified writes have a length restriction of 4 bytes and the address
must be aligned to the size. That is 1 byte writes can be done to any address, 2 bytes must be halfword
aligned, 3 bytes are not allowed and 4 bytes writes must be word aligned. Since there will always be
only one AHB operation performed for each RMAP verified write command the incrementing
address bit can be set to any value.
Non-verified writes have no restrictions when the incrementing bit is set to 1. If it is set to 0 the number of bytes must be a multiple of 4 and the address word aligned. There is no guarantee how many
words will be written when early EOP/EEP is detected for non-verified writes.
17.4.6.4 Read commands
Read commands are performed on the fly when the reply is sent. Thus if an AHB error occurs the
packet will be truncated and ended with an EEP. There are no restrictions for incrementing reads but
non-incrementing reads have the same alignment restrictions as non-verified writes. Note that the
“Authorization failure” error code will be sent in the reply if a violation was detected even if the
length field was zero. Also note that no data is sent in the reply if an error was detected i.e. if the status
field is non-zero.
GR740-UM-DS, Jun 2016, Version 1.4
189
www.cobham.com/gaisler
GR740
17.4.6.5 RMW commands
All read-modify-write sizes are supported except 6 which would have caused 3 B being read and written on the bus. The RMW bus accesses have the same restrictions as the verified writes. As in the verified write case, the incrementing bit can be set to any value since only one AHB bus operation will be
performed for each RMW command. Cargo too large is detected after the bus accesses so this error
will not prevent the operation from being performed. No data is sent in a reply if an error is detected
i.e. the status field is non-zero.
17.4.6.6 Control
The RMAP target mostly runs in the background without any external intervention, but there are a
few control possibilities.
There is an enable bit in the control register of the core which can be used to completely disable the
RMAP target. When it is set to ‘0’ no RMAP packets will be handled in hardware, instead they are all
stored to the DMA channel.
There is a possibility that RMAP commands will not be performed in the order they arrive. This can
happen if a read arrives before one or more writes. Since the target stores replies in a buffer with more
than one entry several commands can be processed even if no replies are sent. Data for read replies is
read when the reply is sent and thus writes coming after the read might have been performed already
if there was congestion in the transmitter. To avoid this the RMAP buffer disable bit can be set to
force the target to only use one buffer which prevents this situation.
The last control option for the target is the possibility to set the destination key which is found in a
separate register.
Table 177.AMBA port hardware RMAP handling of different packet type and command fields.
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Acknowledge
Increment
Address
Command
Action
Reserved
Command
/ Response
Write /
Read
Verify
data
before
write
0
0
-
-
-
-
Response
Stored to DMA-channel.
0
1
0
0
0
0
Not used
Does nothing. No reply is sent.
0
1
0
0
0
1
Not used
Does nothing. No reply is sent.
0
1
0
0
1
0
Read single
address
Executed normally. Address has
to be word aligned and data size
a multiple of four. Reply is sent.
If alignment restrictions are violated error code is set to 10.
0
1
0
0
1
1
Read incrementing
address.
Executed normally. No restrictions. Reply is sent.
0
1
0
1
0
0
Not used
Does nothing. No reply is sent.
0
1
0
1
0
1
Not used
Does nothing. No reply is sent.
0
1
0
1
1
0
Not used
Does nothing. Reply is sent with
error code 2.
GR740-UM-DS, Jun 2016, Version 1.4
190
www.cobham.com/gaisler
GR740
Table 177.AMBA port hardware RMAP handling of different packet type and command fields.
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Acknowledge
Increment
Address
Command
Action
Reserved
Command
/ Response
Write /
Read
Verify
data
before
write
0
1
0
1
1
1
Read-Modify-Write
incrementing address
Executed normally. If length is
not one of the allowed rmw values nothing is done and error
code is set to 11. If the length
was correct, alignment restrictions are checked next. 1 byte
can be rmw to any address. 2
bytes must be halfword aligned.
3 bytes are not allowed. 4 bytes
must be word aligned. If these
restrictions are violated nothing
is done and error code is set to
10. If an AHB error occurs error
code is set to 1. Reply is sent.
0
1
1
0
0
0
Write, single-address,
do not verify
before writing, no
acknowledge
Executed normally. Address has
to be word aligned and data size
a multiple of four. If alignment is
violated nothing is done. No
reply is sent.
0
1
1
0
0
1
Write, incrementing
address, do
not verify
before writing, no
acknowledge
Executed normally. No restrictions. No reply is sent.
0
1
1
0
1
0
Write, single-address,
do not verify
before writing, send
acknowledge
Executed normally. Address has
to be word aligned and data size
a multiple of four. If alignment is
violated nothing is done and
error code is set to 10. If an AHB
error occurs error code is set to 1.
Reply is sent.
0
1
1
0
1
1
Write, incrementing
address, do
not verify
before writing, send
acknowledge
Executed normally. No restrictions. If AHB error occurs error
code is set to 1. Reply is sent.
0
1
1
1
0
0
Write, single
address, verify before
writing, no
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done. Same alignment restrictions apply as for rmw. No reply
is sent.
GR740-UM-DS, Jun 2016, Version 1.4
191
www.cobham.com/gaisler
GR740
Table 177.AMBA port hardware RMAP handling of different packet type and command fields.
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Acknowledge
Increment
Address
Command
Action
Reserved
Command
/ Response
Write /
Read
Verify
data
before
write
0
1
1
1
0
1
Write, incrementing
address, verify before
writing, no
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done. Same alignment restrictions apply as for rmw. If they
are violated nothing is done. No
reply is sent.
0
1
1
1
1
0
Write, single
address, verify before
writing, send
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done and error code is set to 9.
Same alignment restrictions
apply as for rmw. If they are violated nothing is done and error
code is set to 10. If an AHB error
occurs error code is set to 1.
Reply is sent.
0
1
1
1
1
1
Write, incrementing
address, verify before
writing, send
acknowledge
Executed normally. Length must
be 4 or less. Otherwise nothing is
done and error code is set to 9.
Same alignment restrictions
apply as for rmw. If they are violated nothing is done and error
code is set to 10. If an AHB error
occurs error code is set to 1.
Reply is sent.
1
0
-
-
-
-
Unused
Stored to DMA-channel.
1
1
-
-
-
-
Unused
Stored to DMA-channel.
17.4.7 AMBA interface
The AMBA interface consists of an APB interface, an AHB master interface and DMA FIFOs. The
APB interface provides access to the user registers which are described in section 17.4.8. The DMA
engines have 32-bit wide FIFOs to the AHB master interface which are used when reading and writing to the bus.
The transmitter DMA engine reads data from the bus in bursts which are half the FIFO size in length.
A burst is always started when the FIFO is half-empty or if it can hold the last data for the packet. The
burst containing the last data might have shorter length if the packet is not an even number of bursts in
size.
The receiver DMA works in the same way except that it checks if the FIFO is half-full and then performs a burst write to the bus which is half the fifo size in length. Byte accesses are used for non
word-aligned buffers and/or packet lengths that are not a multiple of four bytes. There might be 1 to 3
single byte writes when writing the beginning and end of the received packets.
17.4.7.1 APB slave interface
As mentioned above, the APB interface provides access to the user registers which are 32-bits in
width. The accesses to this interface are required to be aligned word accesses. The result is undefined
if this restriction is violated.
GR740-UM-DS, Jun 2016, Version 1.4
192
www.cobham.com/gaisler
GR740
17.4.7.2 AHB master interface
The port contains a single master interface which is used by both the transmitter and receiver DMA
engines. The arbitration algorithm between the channels is done so that if the current owner requests
the interface again it will always acquire it. This will not lead to starvation problems since the DMA
engines always deassert their requests between accesses.
The burst length will be half the AHB FIFO size except for the last transfer for a packet which might
be smaller. Shorter accesses are also done during descriptor reads and status writes.
The AHB master also supports non-incrementing accesses where the address will be constant for several consecutive accesses. HTRANS will always be NONSEQ in this case while for incrementing
accesses it is set to SEQ after the first access. This feature is included to support non-incrementing
reads and writes for RMAP.
If the core does not need the bus after a burst has finished there will be one wasted cycle (HTRANS =
IDLE).
BUSY transfer types are never requested and the port provides full support for ERROR, RETRY and
SPLIT responses.
17.4.8 Registers
The port is programmed through registers mapped into APB address space. The addresses in the table
below are offsets from each AMBA port’s base address. The actual AMBA AHB address used to
access the port is determined by adding the offset to respective AMBA port’s base address, specified
in table 7, in section 2.3. An identical set of registers described in this section exist for each AMBA
port. The register layout used is explained in section 1.10
GR740-UM-DS, Jun 2016, Version 1.4
193
www.cobham.com/gaisler
GR740
.
Table 178.AMBA port registers
APB address offset**
Register name
Acronym
0x00
AMBA port Control
RTR.AMBACTRL
0x40
AMBA port Status
RTR.AMBASTS
0x08
AMBA port Default address
RTR.AMBADEFADDR
0x0C
RESERVED
0x10
AMBA port Destination key
RTR.AMBADKEY
0x14
AMBA port Time-code
RTR.AMBATC
0x18
RESERVED
0x20, 0x40, 0x60, 0x80
AMBA port DMA control/status, channels 1 - 4 *
RTR.AMBADMACTRL
0x24, 0x44, 0x64, 0x84
AMBA port DMA RX maximum length, channels 1 - 4 *
RTR.AMBADMAMAXLEN
0x28, 0x48, 0x68, 0x88
AMBA port DMA transmit descriptor table address, channels 1 - 4 *
RTR.AMBADMATXDESC
0x2C, 0x4C, 0x6C, 0x8C
AMBA port DMA receive descriptor table address, channels
1-4*
RTR.AMBADMARXDESC
0x30, 0x50, 0x70, 0x90
AMBA port DMA address, channels 1 - 4 *
RTR.AMBADMAADDR
0x34, 0x54, 0x74, 0x94
RESERVED
0x38, 0x58, 0x78, 0x98
RESERVED
0x3C, 0x5C, 0x7C, 0x9C
RESERVED
0xA0
AMBA port Distributed interrupt control
RTR.AMBAINTCTRL
0xA4
AMBA port Interrupt receive
RTR.AMBAINTRX
0xA8
AMBA port Interrupt acknowledgement / extended interrupt
receive
RTR.AMBAACKRX
0xAC
AMBA port Interrupt timeout, interrupt 0-31
RTR.AMBAINTTO0
0xB0
AMBA port Interrupt timeout, interrupt 32-63
RTR.AMBAINTTO1
0xB4
AMBA port Interrupt mask, interrupt 0-31
RTR.AMBAINTMSK0
0xB8
AMBA port Interrupt mask, interrupt 32-63
RTR.AMBAINTMSK1
0xBC - 0xFF
RESERVED
0x100 - 0xFFF
See note ** below
* One identical register per DMA channel. Register is only described once
** Each AMBA port is allocated a 4 KiB memory area in the GR740 memory map. The router registers are aliased within
this memory range which means that an access to offset 0x104 or 0x204 .. 0xF04 all access the same register. To ensure
software should only access
Table 179. 0x00 - RTR.AMBACTRL - AMBA port Control
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RA RX RC
NCH
RES
DI ME
RESERVED
1
1
1
0x3
0x0
1
0
0x00
r
r
r
r
r
r
rw
r
RD RE
0
1
rw rw
RESERVED
9
8
7
6
5
4
TQ R RS PM TI
0
0
3
2
1
0
IE RESERVED
0x00
0
0
0
r
rw
r
rw rw rw rw
0
0x0
r
31
RMAP available (RA) - Constant value of 1, indicating that the RMAP target is implemented.
30
RX unaligned access (RX) - Constant value of 1, indicating that unaligned writes are available for the receiver.
29
RMAP CRC available (RC) - Constant value of 1, indicating that RMAP CRC is enabled in the core.
28: 27
Number of DMA channels (NCH) - The number of available DMA channels minus one. Constant value of 0x3.
26: 25
RESERVED
24
Distributed interrupt support (DI) - Constant value of 1, indicating that distributed interrupts are supported.
GR740-UM-DS, Jun 2016, Version 1.4
194
www.cobham.com/gaisler
GR740
23
Table 179. 0x00 - RTR.AMBACTRL - AMBA port Control
Memory error truncation enable (ME) - If set to 1, a packet being transmitted will be truncated with an EEP if an
error occur while reading from the AMBA port’s TX FIFO.
22: 18
RESERVED
17
RMAP buffer disable (RD) - If set only one RMAP buffer is used. This ensures that all RMAP commands will
be executed consecutively.
16
RMAP Enable (RE) - Enable RMAP target.
15: 9
RESERVED
8
Tick-out IRQ (TQ) - Generate interrupt when a valid time-code is received if the time-code also matches the
time-code filter specified by the RTR.AMBATIME.TCMSK and RTR.AMBATIME.TCVAL fields.
7
Time-code tick-out enable (TO) - If set to 1, the internal tickout signal is set when a valid time-code is received.
if the time-code also matches the time-code filter specified by the RTR.AMBATIME.TCMSK and RTR.AMBATIME.TCVAL fields. The internal tickout signal is connected to the timer units as described in section 5.9 and in
the LATCHCFG register description in section 25.3.
6
Reset (RS) - Make complete reset of the SpaceWire node. Self clearing.
5
Promiscuous Mode (PM) - Enable Promiscuous mode.
4
Tick In (TI) - The host can generate a tick by writing a one to this field. This will increment the timer counter
and the new value is transmitted after the current character is transferred. A tick can also be generated by asserting the tick_in signal.
3
Interrupt Enable (IE) - If set, an interrupt is generated when bit 8 is set and its corresponding event occurs.
2: 0
RESERVED
Table 180. 0x04 - RTR.AMBASTS - AMBA port Status
31 30
28 27 26 25 24 23
13 12 11
9
R
NIRQ
NRXD
NTXD
RESERVED
0
0x6
0
0
0x000
0
0x0
r
r
r
r
r
wc
r
8
7
ME RESERVED EE IA
0
0
RESERVED
0
wc wc
TO
0x00
0
r
wc
31
RESERVED
30: 28
Number of interrupts (NIRQ) - Shows the number of support distributed interrupt, according to the formula
2NIRQ. Constant value of 0x6 = 64 supported interrupts.
27: 26
Number of receive descriptors (NRXD) - Shows the size of the DMA receive descriptor table. Constant value of
0, indicating 128 descriptors.
25: 24
Number of transmit descriptors (NTXD) - Shows the size of the DMA transmit descriptor table. Constant value
of 0, indicating 64 descriptors.
23: 13
RESERVED
12
Memory error packet truncation (ME) - This bit is set to one when a transmitted packet is truncated with an EEP
due to a memory error in the AMBA ports’ TX FIFO.
11: 9
RESERVED
8
Early EOP/EEP (EE) - Set to one when a packet is received with an EOP after the first byte for a non-RMAP
packet and after the second byte for an RMAP packet.
7
Invalid Address (IA) - Set to one when a packet is received with an invalid destination address field, i.e it does
not match the nodeaddr register.
6: 1
RESERVED
0
Tick Out (TO) - A new time count value was received and is stored in the time counter field.
Table 181. 0x08 - RTR.AMBADEFADDR - AMBA port Default address
31
16 15
8
7
0
RESERVED
DEFMASK
DEFADDR
0x0000
0x00
0xFE
r
rw
rw
GR740-UM-DS, Jun 2016, Version 1.4
195
www.cobham.com/gaisler
GR740
Table 181. 0x08 - RTR.AMBADEFADDR - AMBA port Default address
31: 8
RESERVED
15: 8
Default mask (DEFMASK) - Default mask used for node identification on the SpaceWire network. This field is
used for masking the address before comparison. Both the received address and the DEFADDR field are anded
with the inverse of DEFMASK before the address check.
7: 0
Default address (DEFADDR) - Default address used for node identification on the SpaceWire network. Reset
value: 254.
Table 182. 0x10 - RTR.AMBADKEY - AMBA port Destination key
31
8
7
0
RESERVED
DESTKEY
NA
0x00
r
rw
31: 8
RESERVED
7: 0
Destination key (DESTKEY) - RMAP destination key.
Table 183. 0x14 - RTR.AMBATC - AMBA port Time-code
31
24 23
16 15
8
7
6
5
0
TCMSK
TCVAL
RESERVED
TCTRL
TIMECNT
0x00
0x00
0x00
0x0
0x00
rw
rw
r
rw
rw
31: 24
Time-code filter mask (TCMSK) - If a bit in this field is set to 1 then the corresponding bit in the RTR.AMBATIME.TCVAL field must match the value of the same bit in a received time-code in order for that time-code to
generate an AMBA IRQ or a pulse on the internal tickout signals connected to the general purpose timers.
23: 16
Time-code filter value (TCVAL) - For each bit set to 1 in the RTR.AMBATIME.TCMSK, the corresponding bit
in this field must match the value of the same bit of a received time-code in order to that time-code to generate
and AMBA IRQ, or a pulse on the internal tickout signals connected to the general purpose timers.
15: 8
RESERVED
7: 6
Time control flags (TCTRL) - The current value of the time control flags. Sent with time-code resulting from a
tick-in. Received control flags are also stored in this register.
5: 0
Time counter (TIMECNT) - The current value of the system time counter. It is incremented for each tick-in and
the incremented value is transmitted. The register can also be written directly but the written value will not be
transmitted. Received time-counter values are also stored in this register
Table 184. 0x20,0x40,0x60,0x80 - RTR.AMBADMACTRL - AMBA port DMA control/status
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
INTNUM
31: 27
RESERVED EP TR IE
0x00
0x0
rw
r
0
0
0
IT RP TP
RES
0
0x0
0
0
wc wc rw rw wc wc
r
4
3
2
SP SA EN NS RD RX AT RA TA PR PS AI
RI
TI RE TE
0
0
0
0
9
8
0
7
0
6
0
5
0 NR NR NR
1
0
0
0
0
0
rw rw rw rw rw
r
rw wc wc wc wc rw rw rw rw rw
0
Interrupt number (INTNUM) - The interrupt number used for this DMA channel when sending a distributed
interrupt code that was generated due to any of the events maskable by the RTR.AMBADMACTRL.IE and
RTR.AMBADMACTRL.IT bits.
26: 24
RESERVED
23
EEP termination (EP) - Set to 1 when a received packet for the corresponding DMA channel ended with an
Error End of Packet (EEP) character.
22
Truncated (TR) - Set to 1 when a received packet for the corresponding DMA channel is truncated due to a
maximum length violation.
GR740-UM-DS, Jun 2016, Version 1.4
196
www.cobham.com/gaisler
GR740
Table 184. 0x20,0x40,0x60,0x80 - RTR.AMBADMACTRL - AMBA port DMA control/status
Interrupt transmit enable on EEP (IE) - When set to 1, the distributed interrupt code specified in the
RTR.AMBADMACTRL.INTNUM field is generated when a received packet on this DMA channel ended
with an Error End of Packet (EEP) character.
21
20
Interrupt-code transmit enable on truncation (IT) - When set to 1, the distributed interrupt code specified in
the RTR.AMBADMACTRL.INTNUM field is generated when a received packet on this DMA channel is
truncated due to a maximum length violation.
19
Receive packet IRQ (RP) - This bit is set to 1 when an AMBA interrupt was generated due to the fact that a
packet was received for the corresponding DMA channel.
18
Transmit packet IRQ (TP) - This bit is set to 1 when an AMBA interrupt was generated due to the fact that a
packet was transmitted for the corresponding DMA channel.
17: 16
RESERVED
15
Strip pid (SP) - Remove the pid byte (second byte) of each packet. The address byte (first byte) will also be
removed when this bit is set independent of the SA bit.
14
Strip addr (SA) - Remove the addr byte (first byte) of each packet.
13
Enable addr (EN) - Enable separate node address for this channel.
12
No spill (NS) - If cleared, packets will be discarded when a packet is arriving and there are no active descriptors. If set, the GRSPW will wait for a descriptor to be activated.
11
Rx descriptors available (RD) - Set to one, to indicate to the GRSPW that there are enabled descriptors in the
descriptor table. Cleared by the GRSPW when it encounters a disabled descriptor:
10
RX active (RX) - Is set to ‘1’ if a reception to the DMA channel is currently active otherwise it is ‘0’.
9
Abort TX (AT) - Set to one to abort the currently transmitting packet and disable transmissions. If no transmission is active the only effect is to disable transmissions. Self clearing.
8
RX AHB error (RA) - An error response was detected on the AHB bus while this receive DMA channel was
accessing the bus.
7
TX AHB error (TA) - An error response was detected on the AHB bus while this transmit DMA channel was
accessing the bus.
6
Packet received (PR) - This bit is set each time a packet has been received. never cleared by the SW-node.
5
Packet sent (PS) - This bit is set each time a packet has been sent. Never cleared by the SW-node.
4
AHB error interrupt (AI) - If set, an interrupt will be generated each time an AHB error occurs when this
DMA channel is accessing the bus.
3
Receive interrupt (RI) - If set, an interrupt will be generated each time a packet has been received. This happens both if the packet is terminated by an EEP or EOP.
2
Transmit interrupt (TI) - If set, an interrupt will be generated each time a packet is transmitted. The interrupt
is generated regardless of whether the transmission was successful or not.
1
Receiver enable (RE) - Set to one when packets are allowed to be received to this channel.
0
Transmitter enable (TE) - Write a one to this bit each time new descriptors are activated in the table. Writing
a one will cause the SW-node to read a new descriptor and try to transmit the packet it points to. This bit is
automatically cleared when the SW-node encounters a descriptor which is disabled.
Table 185. 0x24,0x44,0x64,0x84 - RTR.AMBADMAMAXLEN - AMBA port DMA RX maximum length
31
25 24
2
1
0
RESERVED
RXMAXLEN
RES
0x00
N/R
0x0
r
rw
r
31: 25
RESERVED
24: 2
RX maximum length (RXMAXLEN) - Receiver packet maximum length in 32-bit words.
1: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
197
www.cobham.com/gaisler
GR740
Table 186. 0x28,0x48,0x68,0x88 - RTR.AMBADMATXDESC - AMBA port DMA transmit descriptor table address
31
10
9
4
3
0
DESCBASEADDR
DESCSEL
RESERVED
N/R
0x00
0x0
rw
rw
r
31: 10
Descriptor table base address (DESCBASEADDR) - Sets the base address of the descriptor table.
9: 4
Descriptor selector (DESCSEL) - Offset into the descriptor table. Shows which descriptor is currently used
by the GRSPW. For each new descriptor read, the selector will increase with 16 and eventually wrap to zero
again.
3: 0
RESERVED
Table 187. 0x2C,0x4C,0x6C,0x8C - RTR.AMBADMARXDESC - AMBA port DMA receive descriptor table address
31
10
9
3
2
0
DESCBASEADDR
DESCSEL
RESERVED
N/R
0x00
0x0
rw
rw
r
31: 10
Descriptor table base address (DESCBASEADDR) - Sets the base address of the descriptor table. Not reset.
9: 3
Descriptor selector (DESCSEL) - Offset into the descriptor table. Shows which descriptor is currently used
by the GRSPW. For each new descriptor read, the selector will increase with 8 and eventually wrap to zero
again. Reset value: 0.
2: 0
RESERVED
Table 188. 0x30,0x50,0x70,0x90 - RTR.AMBADMAADDR - AMBA port DMA address
31
16 15
8
7
0
RESERVED
MASK
ADDR
0x0000
N/R
N/R
r
rw
rw
31: 8
RESERVED
15: 8
Mask (MASK) - Mask used for node identification on the SpaceWire network. This field is used for masking
the address before comparison. Both the received address and the ADDR field are anded with the inverse of
MASK before the address check.
7: 0
Address (ADDR) - Address used for node identification on the SpaceWire network for the corresponding
dma channel when the EN bit in the DMA control register is set.
Table 189. 0xA0 - RTR.AMBAINTCTRL - AMBA port Distributed interrupt control
31
26 25 24 23 22 21 20 19 18 17 16 15 14 13 12
INTNUM
R EE IA
RES
0x00
0
0
0x0
rw
r
rw rw
0
r
TQ AQ IQ
0
0
0
rw rw rw
RES
0x0
r
AA AT IT
0
1
1
rw rw rw
8
RESERVED
0x00
r
7
6
ID
II
TXINT
0
0
0x00
wc rw*
5
0
rw
31: 26
Interrupt number (INTNUM) - The interrupt-number used when sending an interrupt code that was generated due to any of the events maskable by the RTR.AMBAINCTRL.ER and RTR.AMBAINTCTRL.IA bits.
Note that when RTR.RTRCFG.EE = 0 (interrupt with acknowledgement mode), this field must no be set to a
value greater than 31.
25
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
198
www.cobham.com/gaisler
GR740
24
23
Table 189. 0xA0 - RTR.AMBAINTCTRL - AMBA port Distributed interrupt control
Interrupt transmit on early EOP/EEP (EE) - If set to 1, a distributed interrupt code with the interrupt number
specified in the RTR.AMBAINTCTRL.INTNUM field is sent each time an event occurs such that the
STS.EE bit is set to 1 (even if the bit was already set when the event occurred).
Interrupt transmit on invalid address (IA) - If set to 1, a distributed interrupt code with the interrupt number
specified in the RTR.AMBAINTCTRL.INTNUM field is sent each time an event occurs such that the
STS.IA bit is set to 1 (even if the bit was already set when the event occurred).
22: 21
RESERVED
20
Interrupt timeout IRQ enable (TQ) - When set to 1, an AMBA interrupt is generated when a bit in the
RTR.AMBAINTTO0 or RTR.AMBAINTTO1 registers is set. Note that the RTR.AMBACTRL.IE bit also
must be set for this bit to have any effect.
19
Interrupt acknowledgment / extended interrupt receive IRQ enable (AQ) - When set to 1, an AMBA interrupt is generated when an interrupt acknowledgment code or extended interrupt code is received such that a
bit in the RTR.AMBAACKRX register is set to 1 (even if the bit was already set when the code was
received).
18
Interrupt-code receive IRQ enable (IQ) - When set to 1, an AMBA interrupt is generated when an interrupt
code is received such that a bit in the RTR.AMBAINTRX register is set to 1 (even if the bit was already set
when the code was received).
17: 16
RESERVED
15
Handle all interrupt acknowledgment codes (AA) - Is set to 0, only those received interrupt acknowledgment
codes that match an interrupt code sent by software are handled. If set to 1, all received interrupt acknowledgment codes are handled.
14
Interrupt acknowledgment / extended interrupt tickout enable (AT) - When set to 1, the internal tickout signal from this AMBA port is set when an interrupt acknowledgment code or extended interrupt code is
received such that a bit in the RTR.AMBAACKRX register is set to 1 (even if the bit was already set when
the code was received). The internal tickout signal is connected to the timer units as described in section 5.9
and in the LATCHCFG register description in section 25.3.
13
Interrupt tickout enable (IT) - When set to 1, the internal tickout signal from this AMBA port is set when an
interrupt code is received such that a bit in the RTR.AMBAINTRX register is set to 1 (even if the bit was
already set when the code was received). The internal tickout signal is connected to the timer units as
described in section 5.9 and in the LATCHCFG register description in section 25.3.
12: 8
RESERVED
7
Interrupt discarded (ID) - This bit is set to 1 when a distributed interrupt code that software tried to send by
writing the RTR.AMBAINTCTRL.II bit was discarded by the routers switch matrix. There is a maximum of
ten clock cycle delay between the RTR.AMBAINTCTRL.II bit being written and this bit being set.
6
Interrupt-code tick-in (II) - When this field is written to 1 the distributed interrupt code specified by the
RTR.AMBAINTCTRL.TXINT field will be sent out from the AMBA port to the routers switch matrix. This
bit is automatically cleared and always reads ‘0’. Writing a ‘0’ has no effect.
5: 0
Transmit distributed interrupt code (TXINT) - The distributed interrupt code that the core will send when the
RTR.AMBAINTCTRL.II is written with 1.
Table 190. 0xA4 - RTR.AMBAINTRX - AMBA port Interrupt receive
31
0
RXIRQ
0x00000000
wc
31: 0
Received interrupt code (RXIRQ) - Each bit corresponds to the interrupt number with the same number as
the bit index. The core sets a bit to 1 when it receives an interrupt code for which the corresponding bit in the
RTR.AMBAINTMSK0 register is set to 1.
Table 191. 0xA8 - RTR.AMBAACKRX - AMBA port Interrupt acknowledgment / extended interrupt receive
31
0
RXACK
GR740-UM-DS, Jun 2016, Version 1.4
199
www.cobham.com/gaisler
GR740
Table 191. 0xA8 - RTR.AMBAACKRX - AMBA port Interrupt acknowledgment / extended interrupt receive
0x00000000
wc
31: 0
Received interrupt acknowledgement code / extended interrupt code (RXACK) - When operating in
extended interrupt mode (RTR.RTRCFG.EE = 1) then each bit corresponds to the interrupt number with the
same number as the bit index plus 32, i.e bit 0 corresponds to interrupt number 32, bit 1 to interrupt number
33 etc. The core sets a bit to 1 when it receives an extended interrupt code for which the corresponding bit in
the RTR.AMBAINTMSK1 register is set to 1.
When operating in interrupt with acknowledgement mode (RTR.RTRCFG.EE = 0) then each bit corresponds
to the interrupt number with the same number as the bit index. The core sets a bit to 1 when it receives an
interrupt acknowledgement code for which the corresponding bit in the RTR.AMBAINTMSK0 register is
set, and either if RTR.AMBAINTCTRL.AA is set to 1 or for which the matching interrupt code was sent by
software.
Table 192. 0xAC - RTR.AMBAINTTO0 - AMBA port Interrupt timeout, interrupt 0-31
31
0
INTTO
0x00000000
wc
31: 0
Interrupt code timeout (INTTO) - Each bit corresponds to the interrupt number with the same number as the
bit index. This bit is set to 1 when an interrupt code that was sent by software doesn’t receive an interrupt
acknowledgement code for the duration of a timeout period (specified in the RTR.ISRTIMER register), and
if the corresponding bit in the RTR.AMBAINTMSK0 register is set.
Table 193. 0xAC - RTR.AMBAINTTO1 - AMBA port Interrupt timeout, interrupt 32-63
31
0
INTTO
0x00000000
wc
31: 0
Extended interrupt code timeout (INTTO) - Each bit corresponds to the interrupt number with the same
number as the bit index plus 32, i.e bit 0 corresponds to interrupt number 32, bit 1 to interrupt number 33
etc.. This bit is set to 1 when an extended interrupt code that was sent by software time out, i.e after the duration of a timeout period (specified in the RTR.ISRTIMER register), and if the corresponding bit in the
RTR.AMBAINTMSK1 register is set.
Table 194. 0xB0 - RTR.AMBAINTMSK0 - AMBA port Interrupt mask, interrupt 0-31
31
0
MASK
0x00000000
rw
31: 0
Interrupt mask (MASK) - Each bit corresponds to the interrupt number with the same value as the bit index.
If a bit is set to 0, all received interrupt codes and interrupt acknowledgement codes with the interrupt identifier corresponding to that bit is ignored. If a bit is set to 1, then the matching distributed interrupt code is
handled.
GR740-UM-DS, Jun 2016, Version 1.4
200
www.cobham.com/gaisler
GR740
Table 195. 0xB0 - RTR.AMBAINTMSK1 - AMBA port Interrupt mask, interrupt 32-63
31
0
MASK
0x00000000
rw
31: 0
17.5
Interrupt mask (MASK) - Each bit corresponds to the interrupt number with the same value as the bit index.
plus 32, i.e bit 0 corresponds to interrupt number 32, bit 1 to interrupt number 33 etc. If a bit is set to 0, all
received extended interrupt codes with the interrupt identifier corresponding to that bit is ignored. If a bit is
set to 1, then the matching distributed interrupt code is handled.
Configuration port
The configuration port has port number 0. It consists of an RMAP target, AMBA AHB slave interface, SpaceWire Plug-and-Play interface, and a set of configuration and status registers.
17.5.1 RMAP target
17.5.1.1 Overview
The configuration port’s RMAP target implements the RMAP protocol, as defined in the RMAP standard [RMAP]. Verified writes and reads of 4 B, and read-modify-writes of 4 B (8 B if the mask field
is included in the count) are supported. Replies from the configuration port are always sent to the port
they arrived on, regardless of the values of the RMAP command’s Initiator Logical Address field, and
Reply Address field. The address space of the configuration port is specified in section 17.5.3.
Additional requirements on the RMAP commands imposed by the configuration port’s RMAP target
are:
•
The Target Logical Address field must be 0xFE.
•
The Address fields must contain a 4 B aligned address.
•
The Extended Address field must be 0x00.
•
Key field must be 0x00.
•
For write and read commands the Data Length fields must contain a value of either 0 or 4.
•
For read-modify-write commands the Data Length fields must contain a value of either 0 or 8.
•
For write commands the Verify Data Before Write bit in the Instruction field must be set to 1.
How the RMAP target handles commands that does not meet the above requirement is detailed in sections 17.5.1.2 and 17.5.1.4.
17.5.1.2 RMAP command support
Table 196 lists all possible RMAP commands and shows how the configuration port’s RMAP target
handles them. An RMAP command will always have bits 7:6 of the command’s Instruction field set to
“01”, and those bits are therefore left out of the table. Bits 1:0 of the command’s Instruction field
determines the length of the command’s Reply Address Field, and does not affect the action taken, so
GR740-UM-DS, Jun 2016, Version 1.4
201
www.cobham.com/gaisler
GR740
they have been left out of the table as well. The action taken assumes that no errors were detected in
the RMAP packet. For handling of RMAP packet error, see section 17.5.1.4
Table 196.RMAP command decoding and handling.
Bit 5
Bit 4
Bit 3
Bit 2
Write /
Read
Verify
Data
Before
Write
Reply
Increment
Addr
Function
Action taken
0
0
0
0
Invalid
No operation performed. Error code 0x02 is saved in the
RTR.PCTRLCFG.EC field. No reply is sent.
0
0
0
1
Invalid
No operation performed. Error code 0x02 is saved in the
RTR.PCTRLCFG.EC field. No reply is sent.
0
0
1
0
Read single
address
Read operation performed, if the requirements in section
17.5.1.1 are met.
0
0
1
1
Read incrementing address
Read operation performed, if the requirements in section
17.5.1.1 are met.
0
1
0
0
Invalid
No operation performed. Error code 0x02 is saved in the
RTR.PCTRLCFG.EC field. No reply is sent.
0
1
0
1
Invalid
No operation performed. Error code 0x02 is saved in the
RTR.PCTRLCFG.EC field. No reply is sent.
0
1
1
0
Invalid
No operation performed. Reply is sent with error code 0x02.
Error code is also saved in the RTR.PCTRLCFG.EC field.
0
1
1
1
Read-modifywrite incrementing address
Read-modify-write operation performed if the requirements in
section 17.5.1.1 are met.
1
0
0
0
Write, single
address, don’t
verify before
writing, no reply
No operation performed. Error code 0x0A is saved in the
RTR.PCTRLCFG.EC field. No reply sent.
1
0
0
1
Write, incrementing address,
don’t verify
before writing,
no reply
No operation performed. Error code 0x0A is saved in the
RTR.PCTRLCFG.EC field. No reply sent.
1
0
1
0
Write, single
address, don’t
verify before
write, send reply
No operation performed. Reply is sent with error code 0x0A.
Error code is also saved in the RTR.PCTRLCFG.EC field.
1
0
1
1
Write, incrementing address,
don’t verify
before write,
send reply
No operation performed. Reply is sent with error code 0x0A.
Error code is also saved in the RTR.PCTRLCFG.EC field.
1
1
0
0
Write, single
address, verify
before writing,
no reply
Write operation performed if the requirements in section
17.5.1.1 are met.
GR740-UM-DS, Jun 2016, Version 1.4
202
www.cobham.com/gaisler
GR740
Table 196.RMAP command decoding and handling.
Bit 5
Bit 4
Bit 3
Bit 2
Write /
Read
Verify
Data
Before
Write
Reply
Increment
Addr
1
1
0
1
1
1
1
Function
Action taken
1
Write, incrementing address,
verify before
writing, no reply
Write operation performed if the requirements in section
17.5.1.1 are met.
1
0
Write, single
address, verify
before writing,
send reply
Write operation performed if the requirements in section
17.5.1.1 are met.
1
1
Write, incrementing address,
verify before
writing, send
reply
Write operation performed if the requirements in section
17.5.1.1 are met.
17.5.1.3 Access control
After reset / power-up the configuration port’s address space can be accessed from all the ports. Configuration port accesses can be individually disabled per port by clearing the corresponding
RTR.PCTRL.CE bit. Write commands, and read-modify-write commands to the configuration area
can be globally disabled by writing a 0 to the RTR.CFGWE.WE bit. When a correct RMAP command
is received but not allowed due to one or more of the access control features being enabled, a reply
with Status field set to 0x0A (Authorization failure) is sent (if requested), and the RTR.PSTSCFG.EC
field is updated to reflect the error. If a reply is not requested, the RTR.PSTSCFG.EC field is still set.
In both cases, the operation is not performed.
17.5.1.4 RMAP Error handling
Table 197 shows the order in which errors in an RMAP command are detected. As soon as an error is
detected, the command is discarded. If a reply should be sent, to a command that included an error, the
reply is sent as soon as possible after the error is detected. This means that the reply might be sent out
before the complete incoming RMAP command has been received. Note that since the complete
RMAP command is buffered before it is executed, a command that contains an error is never executed.
Table 197.RMAP target error detection order
Detection
Order
Error type
RMAP
error
code
1
Wrong Protocol Identifier
N/A
The RTR.PSTSCFG.PT bit is set in order to indicate that the error
occurred. No reply is sent.
2
EOP / EEP before completed header
N/A
The RTR.PSTSCFG.EO / RTR.PSTSCFG.EE bit is set in order to indicate that the error occurred. No reply is sent.
3
Header CRC error
N/A
The RTR.PSTSCFG.HC bit is set in order to indicate that the error
occurred. No reply is sent.
4
Unused RMAP packet
type
N/A
If the packet type (bit 7:6 of the packet’s Instruction field) is “10” or
“11” then the bit RTR.PSTSCFG.PT is set. For the value “00” (indicating a reply), no bit in RTR.PSTSCFG is set, since the RMAP standard
[RMAP] does not specify that such an event should be recorded.
GR740-UM-DS, Jun 2016, Version 1.4
Action taken
203
www.cobham.com/gaisler
GR740
Table 197.RMAP target error detection order
Detection
Order
Error type
RMAP
error
code
Action taken
5
EEP immediately after
header
N/A
The RTR.PSTSCFG.EE bit is set in order to indicate that the error
occurred. No reply is sent.
6
Unused RMAP command
code
0x02
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
7
Invalid Target Logical
Address
0x0C
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
8
Invalid Key
0x03
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
9
Verify buffer overrun
0x09
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
10
RMW data length error
0x0B
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
11
RMAP command not
implemented or not
authorized.
0x0A
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
12
Early EOP / early EEP
(not immediately after
header)
0x05 /
0x07
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
13
Invalid Data CRC
0x04
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
14
EEP
0x07
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
15
Too much data
0x06
RMAP error code is saved in the RTR.PCTRLCFG.EC field. Reply is
sent if the Reply bit in the command’s Instruction field was set to 1.
Most of the errors listed in table 197 are errors that only occur in one specific way, and they are also
explained in the RMAP standard [RMAP]. Authorization failure (error code 0x0A) is however an
exception. All the cases that lead to an authorization failure are listed below:
•
A read command’s Data Length field exceed 128 B.
•
A command’s (read, write, or read-modify-write) Address field does not contain a 4 B aligned
address.
•
The access control features described in section 17.5.1.3 prevented the port from accessing the
RMAP target.
•
The Address field of a command (read, write, or read-modify-write) contains an address that is
outside of the configuration port’s memory space.
•
The Length field of a command (read, write, or read-modify write) is 0 nor 4.
•
A non-verified write command was received.
17.5.2 AMBA AHB slave interface
The configuration port provides an AMBA AHB slave interface, which makes the whole configuration port’s address space accessible from the AHB bus.
The routing table is shared between the ports, RMAP target and AHB slave, so accesses from the
AHB slave might be stalled because of accesses from the other sources. The priority order when
accessing the routing table, starting from the highest, is: router ports, AHB slave, RMAP target. Note
that since the AHB slave has higher priority than the RMAP target, it is possible to read and write to
GR740-UM-DS, Jun 2016, Version 1.4
204
www.cobham.com/gaisler
GR740
the configuration port’s registers in the middle of an RMAP write command. This needs to be considered in order to avoid a mismatch between the expected written value and actual written value.
None of the access control mechanisms mentioned in section 17.5.1.3 have any effect on the AHB
slave interface.
17.5.3 Registers
The configuration port’s registers listed in this section can be accessed either through the RMAP target, or the AMBA AHB slave interface. The address specified in table 198 is the RMAP address.
When accessed through the AHB slave interface, the address in the table should be added to the
router’s base address specified in table 7, section 2.3. Registers that exist in several identical copies,
corresponding to different addresses or ports, for example the RTR.RTPMAP registers, are only
described once. The register layout used is explained in section 1.10.
Table 198.GRSPWROUTER registers
RMAP address
Register name
Acronym
0x00000000
RESERVED *
0x00000004 0x00000030
Routing table port mapping, physical addresses 1-12
0x00000034 0x0000007C
RESERVED *
0x00000080 0x000003FC
Routing table port mapping, logical addresses 32-255
0x00000400
RESERVED *
0x00000404 0x00000430
Routing table address control, physical addresses 1-12
0x00000450 0x0000047C
RESERVED *
0x00000480 0x000007FC
Routing table address control, logical addresses 32-255
RTR.RTACTRL
0x00000800
Port control, port 0 (configuration port)
RTR.PCTRLCFG
0x00000804 0x00000830
Port control, port 1-12 (SpaceWire ports and AMBA ports)
RTR.PCTRL
0x00000850 0x0000087C
RESERVED
0x00000880
Port status, port 0 (configuration port)
RTR.PSTSCFG
0x00000884 0x000008B0
Port status, ports 1-12 (SpaceWire ports and AMBA ports)
RTR.PSTS
0x000008D0 0x000008FC
RESERVED
0x00000900 0x00000930
Port timer reload, ports 0-12
0x00000950 0x0000097C
RESERVED
0x00000980
Port control 2, ports 0 (configuration port)
RTR.PCTRL2CFG
0x00000984 0x00000930
Port control 2, ports 1-12 (SpaceWire ports and AMBA ports)
RTR.PCTRL2
0x000009D0 0x000009FC
RESERVED
RTR.RTPMAP
RTR.RTPMAP
RTR.RTACTRL
RTR.PTIMER
0x00000A00
Router configuration / status
RTR.RTRCFG
0x00000A04
Time-code
RTR.TC
0x00000A08
Version / instance ID
RTR.VER
GR740-UM-DS, Jun 2016, Version 1.4
205
www.cobham.com/gaisler
GR740
Table 198.GRSPWROUTER registers
RMAP address
Register name
Acronym
0x00000A0C
Initialization divisor
RTR.IDIV
0x00000A10
Configuration write enable
RTR.CFGWE
0x00000A14
Timer prescaler reload
RTR.PRESCALER
0x00000A18
Interrupt mask
RTR.IMASK
0x00000A1C
Interrupt port mask
RTR.IPMASK
0x00000A20
Port interrupt pending
RTR.PIP
0x00000A24
Interrupt code generation
RTR.ICODEGEN
0x00000A28
Interrupt code distribution ISR, interrupt 0-31
RTR.ISR0
0x00000A2C
Interrupt code distribution ISR, interrupt 32-63
RTR.ISR1
0x00000A30
Interrupt code distribution ISR timer reload
RTR.ISRTIMER
0x00000A34
Interrupt code distribution ACK-to-INT timer reload
RTR.AITIMER
0x00000A38
Interrupt code distribution ISR change timer reload
RTR.ISRCTIMER
0x00000A3C
RESERVED
0x00000A40
SpaceWire link running status
RTR.LRUNSTS
0x00000A44
Capability
RTR.CAP
0x00000A48 0x00000A4C
RESERVED
0x00000A50
SpaceWire Plug-and-Play - Device Vendor and Product ID
RTR.PNPVEND
0x00000A54
SpaceWire Plug-and-Play - Unit Vendor and Product ID
RTR.PNPUVEND
0x00000A58
SpaceWire Plug-and-Play - Unit Serial Number
RTR.PNPUSN
0x00000A5C 0x00000DFC
RESERVED
0x00000E00 0x00000E30
Maximum packet length, ports 0-12
0x00000E50 0x00000E7C
RESERVED
0x00000E84 0x00000E20
Credit counter, ports 1-8
0x00000E24 0x00000FFC
RESERVED
0x00001000
RESERVED**
RTR.MAXPLEN
RTR.CREDCNT
NOTE: Not available via on-chip AHB slave interface. Only available through
RMAP.
0x00001004 0x00001030
Routing table, combined port mapping and address control, addresses 1-12
0x00001050 0x0000107C
RESERVED**
0x00001080 0x000013FC
Routing table, combined port mapping and address control, addresses 32-255
RTR.RTCOMB
NOTE: Not available via on-chip AHB slave interface. Only available through
RMAP.
NOTE: Not available via on-chip AHB slave interface. Only available through
RMAP.
RTR.RTCOMB
NOTE: Not available via on-chip AHB slave interface. Only available through
RMAP.
* Physical address 0 (configuration port), and physical addresses 13-31 (non existing ports) does not have an RTR.RTPMAP or RTR.RTACTRL register, and are therefore RESERVED.
** Physical address 0 (configuration port), and physical addresses 13-31 (non existing ports) does not have an
RTR.RTCOMB register, and are therefore RESERVED.
GR740-UM-DS, Jun 2016, Version 1.4
206
www.cobham.com/gaisler
GR740
Table 199. 0x00000004-0x00000030, 0x00000080-0x000003FC - RTR.RTPMAP - Routing table port mapping, addresses
1-12 and 32-255
31
13 12
1
0
RESERVED
PE
PD
0x00000
N/R
N/R
r
rw
rw
31: 13
RESERVED
12: 1
Port enable bits (PE) - When set to 1, each bit enables packets with the physical / logical address corresponding
to this RTR.RTPMAP register to be sent on the port with the same number as the bit index. For physical
addresses, the bit index corresponding to the port with the same number as the physical address itself is always
1. For logical addresses, at least one bit in this field must be set in order for packets with the corresponding logical address to be routed.
0
Packet distribution (PD) - When set to 1, packet distribution is used for the physical / logical address corresponding to this RTR.RTPMAP register. When set to 0, group adaptive routing is used. See section 17.2.6 and
17.2.7 for more information.
NOTE: After reset, or after writing a RTR.RTPMAP register to zero, the register is considered invalid. For incoming packets with a logical address this means that the packet is spilled, and an invalid address error generated. For physical addresses
this means that the RTR.RTPMAP register will not be used when routing the packet, and the packet is routed to the port that
matches the physical address.
Table 200. 0x00000404-0x00000430, 0x00000480-0x000007FC - RTR.RTACTRL - Routing table address control,
addresses 1-12 and 32-255
31
4
RESERVED
3
2
1
0
SR EN PR HD
*
0x0000000
*
*
*
rw rw rw rw
r
31: 4
RESERVED
3
Spill-if-not-ready (SR) - When set to 1, an incoming packet with the corresponding physical / logical address is
immediately spilled if the selected output port’s link interface is not in run-state. If packet distribution is used for
the incoming packet, and this bit is set, the packet is spilled unless all output ports’ link interfaces are in run
state. For physical addresses, this bit is double mapped in the RTR.PCTRL.SR field. Reset value for physical
addresses is 0. Reset value for logical addresses is N/R.
2
Enable (EN) - Enables the routing table address control entry. Address control entries for physical addresses are
always enabled, and this field is constant 1. For logical addresses, this bit must be set to 1 in order for packets
with the corresponding logical address to be routed. Reset value for logical addresses is 0.
1
Priority (PR) - Sets the arbitration priority of this physical / logical address. 0 = low priority, 1 = high priority.
Used when more than one packet is competing for the same output port. For physical addresses, this bit is double
mapped in the RTR.PCTRL.PR field. Reset value for physical addresses is 0. Reset value for logical addresses is
N/R.
0
Header deletion (HD) - Enables / disabled header deletion for the corresponding logical address. For physical
addresses, header deletion is always enabled, and this bit is constant 1. Reset value for logical addresses is N/R.
Table 201. 0x00000800 - RTR.PCTRLCFG - Port control, port 0 (configuration port)
31
18 17 16 15
RESERVED
0x0000
r
31: 18
PL TS
0
0
rw rw
10
RESERVED
9
8
0
TR
RESERVED
0x00
1
0x000
r
rw
r
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
207
www.cobham.com/gaisler
GR740
Table 201. 0x00000800 - RTR.PCTRLCFG - Port control, port 0 (configuration port)
Packet length truncation (PL) - When set to 1, an RMAP / SpaceWire Plug-and-Play reply is spilled, and an EEP
written to the transmit FIFO of the output port, if the total length of the reply packet exceeds the maximum
length specified in the RTR.MAXPLEN register for port 0. See section 17.2.16 for more information on packet
length truncation.
17
16
Time-code / distributed interrupt code truncation (TS) - When set to 1, an ongoing RMAP / SpaceWire Plugand-Play reply is spilled, and an EEP written to the transmit FIFO of the output port, if a valid time-code or distributed interrupt code is received, and if the time-code / distribute interrupt code matches the codes selected by
the RTR.PCTRL2CFG.SV and RTR.PCTRL2CFG.SM fields. See section 17.2.21 for more information on timecode / distributed interrupt code spill.
15: 10
RESERVED
9
Timer enable (TR) - Enable data character timer for port 0. See section 17.2.15 for details.
8: 0
RESERVED
Table 202. 0x00000804-0x00000830 - RTR.PCTRL - Port control, ports 1-12 (SpaceWire ports and AMBA ports)
31
24 23 22 21 20 19 18 17 16 15 14
RD
RES
0x27
0x0
rw
r
11 10
9
8
7
6
5
4
3
2
1
0
ST SR AD LR PL TS IC ET RESERVED DI TR PR TF RS TE R CE AS LS LD
0
0
0
0
0
0
*
0
rw rw rw rw rw rw rw rw
0x0
r
1
0
1
rw rw rw rw rw rw
0
1
0
0
0
r
rw rw rw rw
1
0
0
31: 24
Run-state clock divisor (RD) - Clock divisor value used for the corresponding port’s link interface when in runstate. Field is only available for the SpaceWire ports.
23: 22
RESERVED
21
Static routing enable (ST) - When set to 1, incoming packets on this port are routed based on the physical
address specified in the corresponding RTR.PCTRL2.SD field, and the setting of the corresponding
RTR.PCTRL2.SC bit, instead of the packet’s first byte. Header deletion is not used when static routing is
enabled, which means that the first byte of the packet is always sent as well. This bit can only be set to 1 if the
RTR.RTRCFG.SR bit is set to 1. Note that when this bit is set to 1 it is not possible to access the configuration
port from this port.
20
Spill-if-not-ready (SR) - This bit is double mapping of the RTR.RTACTRL.SR bit. See table 200.
19
Auto-disconnect (AD) - When set to 1, the auto-disconnect feature described in section 17.2.14 is enabled. This
bit is only available for the SpaceWire ports.
18
Link-start-on-request (LR) - When set to 1, the link-start-on-request feature described in section 17.2.13 is
enabled. This bit is only available for the SpaceWire ports.
17
Packet length truncation (PL) - When set to 1, packets for which this port is the input port will be spilled, and an
EEP written to the transmit FIFO of the output port(s) if the packets exceed the maximum length specified in the
corresponding RTR.MAXPLEN register. See section 17.2.16 for more information on packet length truncation.
16
Time-code / distributed interrupt code truncation (TS) - When set to 1, ongoing packets for which this port is the
input port will be spilled, and an EEP written to the transmit FIFO of the output port(s), if a valid time-code /
distributed interrupt code is received, and if the time-code / distributed interrupt code also matches the codes
selected by the RTR.PCTRL2.SV and RTR.PCTRL2.SM fields. See section 17.2.21 for more information on
time-code / distributed interrupt code spill.
15
Distributed interrupt codes enable (IC) - When set to 0, all incoming distributed interrupt codes on this port are
discarded, and no distributed interrupt codes are sent out on the port. When set to 1, the four bits
RTR.PCTRL2.IR, RTR.PCTRL2.IT, RTR.PCTRL2.AR, RTR.PCTRL2.AT are used to enable / disable distributed interrupt code transmit and receive. Note that the global distributed interrupt code enable bit, RTRTCFG.IE,
also must be set to 1 for distributed interrupt codes to be sent / received. See section 17.2.18 for a description of
distributed interrupt code distribution. Reset value depends on bootstrap signals, as described in section 3.1.
14
Enable external time (ET) - When a time-code is received on the port and this bit is set to 0, the router discards
the received time-code value and instead increments its internal time-counter value (RTR.TC.TC), and forwards
a time-code with the new value to the other ports. If this bit is set to 1 when the time-code is received, the timecode is processed according to the rules described in section 17.2.17. This bit is only available for the AMBA
ports.
13: 11
RESERVED
10
Disable port (DI) - When set to 1, data transfers to and from this port are disabled. See section 17.2.8 for details.
9
Packet timer enable (TR) - Enable the data character timer for incoming packets. See section 17.2.15 for details.
GR740-UM-DS, Jun 2016, Version 1.4
208
www.cobham.com/gaisler
GR740
8
Table 202. 0x00000804-0x00000830 - RTR.PCTRL - Port control, ports 1-12 (SpaceWire ports and AMBA ports)
Priority (PR) - This bit is double mapping of the RTR.RTACTRL.SR bit. See table 200.
7
Transmit FIFO reset (TF) - Resets the transmit FIFO on this port. This means that the FIFO is emptied (counters
and pointers set to 0), and an EEP is written to the FIFO to ensure that any incomplete packet is detected by the
receiver. If a packet transmission is active (another port is using this port as output port) when this bit is set, the
remainder of that packet will be spilled before the EEP is inserted. This bit is self-clearing, and should not be
written with 0 while it is 1, since that could abort the ongoing transmit FIFO reset.
6
Receive FIFO spill (RS) - Spills the receive FIFO for this port, meaning that the packet currently being received
is spilled. The output port(s) used for the packet will have an EEP written to the transmit FIFO to indicate that
the packet was ended prematurely. If no packet is received, setting this bit has no effect. This bit is self-clearing,
and should not be written with 0 while it is 1, since that could abort the ongoing receive FIFO spill.
5
Time-code enable (TE) - Enables time-codes to be received and transmitted on this port. When set to 1, received
time-codes are processed according to the rules described in section 17.2.17. If this bit is set to 0, all received
time-codes on this port are ignored.
4
RESERVED
3
Configuration port access enable (CE) - Enable accesses to the configuration port from this port. If set to 0,
incoming packets with physical address 0 will be spilled.
2
Autostart (AS) - Enable the link interface FSM’s autostart feature, as defined in [SPW]. This bit is only available
for the SpaceWire ports.
1
Link start (LS) - Start the link interface FSM. This bit is only available for the SpaceWire ports.
0
Link disabled (LD) - Disable the link interface FSM. This bit is only available for the SpaceWire ports.
Table 203. 0x00000880 - RTR.PSTSCFG - Port status, port 0 (configuration port)
31 30 29 28 27 26 25 24 23
EO EE PL TT PT HC PI CE
0
0
0
0
0
0
0
0
wc wc wc wc wc wc wc rw*
20 19 18 17
EC
R
0x0
0
r
r
TS ME
0
0
wc wc
12 11
7
6
5
RESERVED
IP
RES
0x00
0x0
0x0
r
r
r
4
CP
3
0
PC
0x0
rw*
r
31
Early EOP (EO) - Set to 1 when an RMAP / SpaceWire Plug-and-Play command with an early EOP was
received by the configuration port. See section 17.5.1.4 for error detection order.
30
Early EEP (EE) - Set to one when an RMAP / SpaceWire Plug-and-Play command with an early EEP was
received by the configuration port. See section 17.5.1.4 for error detection order.
29
Packet length truncation (PL) - Set to 1 when an RMAP / SpaceWire Plug-and-Play reply packet has been
spilled due to a maximum length violation. See section 17.2.16 for details.
28
Time code / distributed interrupt code tick truncation (TT) - Set to one when an RMAP / SpaceWire Plug-andPlay reply packet has been spilled due to a time code / distributed interrupt code. See section 17.2.21 for details.
27
Packet type error (PT) - Set to one if an RMAP / SpaceWire Plug-and-Play packet with correct header CRC, but
with the packet type bits set to the reserved values “10” or “11”, was received by the configuration port. See section 17.5.1.4 for error detection order.
26
Header CRC Error (HC) - Set to one if a Header CRC error is detected in an RMAP / SpaceWire Plug-and-Play
command received by the configuration port. See section 17.5.1.4 for error detection order.
25
Protocol ID Error (PI) - Set to one if a packet received by the configuration port had the wrong protocol ID. Supported protocol ID:s are 0x01 (RMAP), and 0x03 (SpaceWire Plug-and-Play). See section 17.5.1.4 for error
detection order.
24
Clear error code (CE) - Write with a 1 to clear the RTR.PCTRLCFG.EC field. This bit is self clearing and always
reads 0. Writing 0 has no effect.
23: 20
Error code (EC) - Shows the four least significant bits of the latest non-zero RMAP status code. If zero, no error
has occurred.
19
RESERVED
18
Timeout spill (TS) - Set to one when an RMAP reply was spilled due to a packet timeout. See section 17.2.15 for
details.
17
Memory error (ME) - Set to one when a memory error occur while accessing the on-chip memory used as buffer
for RMAP connands handled by the configuration port.
16: 12
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
209
www.cobham.com/gaisler
GR740
Table 203. 0x00000880 - RTR.PSTSCFG - Port status, port 0 (configuration port)
Input port (IP) - The number of the last port from which a packet was routed to the configuration port. This field
is updated even if an operation is not performed, for example due to an incorrect RMAP packet.
11: 7
6: 5
RESERVED
4
Clear SpaceWire Plug-and-Play error code (CP) - Write with a 1 to clear the RTR.PCTRLCFG.PC field. This bit
is self clearing and always reads 0. Writing 0 has no effect.
3: 0
SpaceWire Plug-and-Play Error code (PC) - Shows the four least significant bits of the latest non-zero SpaceWire Plug-and-Play status code. If zero, no error has occurred.
Table 204. 0x00000884-0x000008B0 - RTR.PSTS - Port status, ports 1-12 (SpaceWire ports and AMBA ports)
31 30 29 28 27 26 25
PT
*
r
23 22 21 20 19 18 17 16 15 14
PL TT RS SR RESERVED LR SP AC R
0
0
0
0
wc wc wc wc
0x0
0
0
0
0
r
r
r
r
r
TS ME TF RE
0
0
wc wc
12 11
7
LS
IP
6
5
4
3
2
1
0
PR PB IA CE ER DE PE
0
1
000
00000
0
0
r
r
r
r
r
r
0
0
0
0
0
wc wc wc wc wc
31: 30
Port type (PT) - The type of this port. Constant value of “00” for the SpaceWire ports, and constant value of “01”
for the AMBA ports.
29
Packet length truncation (PL) - Set to 1 when a packet for which this port was the input port has been spilled due
to the packet length truncation feature. See section 17.2.16 for details.
28
Time-code / distributed interrupt code truncation (TT) - Set to 1 when a packet for which this port was the input
port has been spilled due to the time-code / distributed interrupt code truncation feature. See section 17.2.21 for
details.
27
RMAP / SpaceWire Plug-and-Play spill (RS) - Set to 1 when an RMAP / SpaceWire Plug-and-Play command
received on this port was spilled by the configuration port.
26
Spill-if-not-ready spill (SR) - Set to 1 when a packet received on this port was spilled due to the spill-if-notready feature. See section 17.2.10.
25: 23
RESERVED
22
Link-start-on-request status (LR) - Set to 1 when this port either was started, or currently is trying to start, due to
the link-start-on-request feature, described in section 17.2.13. This bit is only available for the SpaceWire ports.
21
Spill status (SP) - This bit is 1 when a packet that is incoming on this port currently is being spilled. Otherwise,
this bit is 0.
20
Active status (AC) - Set to 1 when a packet arrives at this port and the port has been given access to the routing
table. Cleared when the packet has been transmitted or spilled.
19
RESERVED
18
Timeout spill (TS) - Set to 1 when a packet for which this port was the input port was spilled due to a packet timeout. See section 17.2.15 for details.
17
Memory error (ME) - Set to one when a memory error occur while accessing the on-chip memory in the ports.
16
Transmit FIFO full (TF) - Set to 1 when the transmit FIFO on this port is full.
15
Receive FIFO empty (RE) - Set to 1 when the receive FIFO on this port is empty. This bit is only available for
the SpaceWire ports.
14: 12
Link state (LS) - Current link state. 000 = Error reset. 001 = Error wait, 010 = Ready, 011 = Started, 100 = Connecting, 101 = Run state. This field is only available for the SpaceWire ports.
11: 7
Input port (IP) - This field shows the number of the input port for either the currently ongoing packet transfer on
this port (if RTR.PSTS.PB = 1), or for the last packet transfer on this port (if RTR.PSTS.PB = 0).
6
Port receive busy (PR) - Set to 1 when this port is the input port of an ongoing packet transfer.
5
Port transmit busy (PB) - Set to 1 when this port is the output port of an ongoing packet transfer.
4
Invalid address (IA) - Set to 1 when an invalid address error occurred on this port. See section 17.2.12 for
details.
3
Credit error (CE) - Set to 1 when a credit error has occurred. This bit is only available for the SpaceWire ports.
2
Escape error (ER) - Set to 1 when an escape error has occurred. This bit is only available for the SpaceWire
ports.
GR740-UM-DS, Jun 2016, Version 1.4
210
www.cobham.com/gaisler
GR740
Table 204. 0x00000884-0x000008B0 - RTR.PSTS - Port status, ports 1-12 (SpaceWire ports and AMBA ports)
Disconnect error (DE) - Set to 1 when a disconnect error has occurred. This bit is only available for the SpaceWire ports.
1
0
Parity error (PE) - Set to 1 when a parity error has occurred on. This bit is only available for the SpaceWire
ports.
Table 205. 0x00000900-0x00000930 - RTR.PTIMER - Port timer reload, ports 0-12
31
16 15
0
RESERVED
RL
0x0000
0xFFFF
r
rw*
31: 16
RESERVED
15: 0
Timer reload (RL) - Port timer reload value, counted in prescaler ticks. This value is used to reload the corresponding port timer used for packet transfer timeouts, and auto-disconnect. The minimum value of this field is 1.
Trying to write 0 will result in 1 being written.
Table 206. 0x00000980 - RTR.PCTRL2CFG - Port control 2, port 0 (configuration port)
31
24 23
16 15 14 13 12 11 10
SM
SV
0xC0
rw
9
8
6
OR
RESERVED
0x00
1
0x0000
rw
rw
r
5
1
0
31: 24
Time-code / distributed interrupt code truncation mask (SM) - Defines which bits of a time-code / distributed
interrupt code that must match the value specified in RTR.PCTRL2CFG.SV in order for an RMAP / SpaceWire
Plug-and-Play reply packet to be spilled. If a bit in this field is set to 1, the corresponding bit in
RTR.PCTRL2.SV must match the time-code / distributed interrupt code. If a bit in this field is set to 0, the corresponding bit in RTR.PCTRL2.SV does not have to match the time-code / distributed interrupt code.
23: 16
Time-code / distributed interrupt code truncation value (SV) - Defines the value to use together with the
RTR.PCTRL2CFG.SM field when checking if a received time-core / distributed interrupt code should spill an
ongoing RMAP / SpaceWire Plug-and-Play reply.
15
Overrun timeout enable (OR) - Enables spilling due to overrun timeouts for RMAP / SpaceWire Plug-and-Play
replies. See section 17.2.15 for details.
14: 0
RESERVED
Table 207. 0x00000984-0x00000930 - RTR.PCTRL2 - Port control 2, ports 1-12 (SpaceWire ports and AMBA ports)
31
24 23
16 15 14 13 12 11 10
SM
SV
0xC0
0x00
rw
rw
OR UR R
1
AT AR IT
1
1
9
8
6
IR RESERVED
1
0
1
rw rw
r
rw rw rw rw
1
5
1
SD
0
SC
0x0
0x00
0
r
rw
rw
31: 24
Time-code / distribute interrupt code truncation mask (SM) - Defines which bits of a time-code / distributed
interrupt code that must match the value specified in RTR.PCTRL2.SV in order for a packet, for which this port
is the input port, to be spilled. If a bit in this field is set to 1, the corresponding bit in RTR.PCTRL2.SV must
match the time-code / distribute interrupt code. If a bit in this field is set to 0, the corresponding bit in
RTR.PCTRL2.SV does not have to match the time-code / distributed interrupt code.
23: 16
Time-code / distribute interrupt code truncation value (SV) - Defines the value to use together with the
RTR.PCTRL2.SM field when checking if a time-code / distributed interrupt code should spill a packet for which
this port is the input port.
15
Overrun timeout enable (OR) - Enables spilling due to overrun timeouts for packets for which this port is the
input port. See section 17.2.15 for details.
GR740-UM-DS, Jun 2016, Version 1.4
211
www.cobham.com/gaisler
GR740
14
Table 207. 0x00000984-0x00000930 - RTR.PCTRL2 - Port control 2, ports 1-12 (SpaceWire ports and AMBA ports)
Underrun timeout enable (UR) - Enables spilling due to unerrun timeouts for packets for which this port is the
input port. See section 17.2.15 for details.
13
RESERVED
12
Interrupt acknowledgement code transmit enable (AT) - Enables the transmission of interrupt acknowledgement
codes on this port. If set to 0, no interrupt acknowledgement codes will be forwarded to this port.
11
Interrupt acknowledgement code receive enable (AR) - Enabled the reception of interrupt acknowledgement
codes on this port. If set to 0, all received interrupt acknowledgement codes on this port will be silently discarded.
10
Interrupt code transmit enable (IT) - Enables the transmission of interrupt codes on this port. If set to 0, no interrupt codes will be forwarded to this port.
9
Interrupt code receive enable (IR) - Enabled the reception of interrupt codes on this port. If set to 0, all received
interrupt codes on this port will be silently discarded.
8: 6
RESERVED
5: 1
Static route destination (SD) - When RTR.PCTRL.ST is set to 1, incoming packets on this port will be routed
based on the value of this field, and the setting of RTR.PCTRL2.SC, instead of the packet’s first byte.
0
Static route configuration (SC) - When this bit is set to 1, the RTR.RTPMAP register corresponding to the physical address specified by the RTR.PCTRL2.SD field will be used when routing packets, if RTR.PCTRL.ST is set
to 1.
Table 208. 0x00000A00 - RTR.RTRCFG - Router configuration / status
31
31: 27
27 26
22 21
17 16 15 14 13 12 11 10
SR PE IC
9
8
7
6
SP
AP
RESERVED
0x08
0x04
0x00
1
1
0
r
r
r
r
r
rw rw rw rw rw rw rw rw
5
4
3
2
1
0
IS
IP
AI AT IE RE EE R SA TF ME TA PP
0
0
*
1
*
0
*
0
0
r
rw rw wc
*
0
1
1
r
r
SpaceWire ports (SP) - Set to the number of SpaceWire ports in the router. Constant value of 0x08.
26: 22
AMBA ports (AP) - Set to the number of AMBA ports in the router. Constant value of 0x04
21: 16
RESERVED
15
Static routing enable (SR) - This read-only bit specifies that the router’s static routing feature is always globally
enabled. See section 17.2.9 for details.
14
SpaceWire Plug-and-Play enable (PE) - This read-only bit specifies that the router’s SpaceWire Plug-and-Play
features are always globally enabled. See section 17.5.4 for details.
13
ISR change timer enable (IC) - If set to 1, the router will wait for the time period specified by the RTR.IRCTIMER register after an ISR bit change value, before it allows an incoming interrupt code / interrupt acknowledgement code to change the value of the same ISR bit. If set to 0, the ISR change timers are not used, and an
ISR bit is allowed to change value again as soon as the previous interrupt code / interrupt acknowledgement
code has been distributed.
12
Distributed interrupt code selection routine (IS) - If set to 0, the router uses round-robin on the interrupt numbers
when deciding which distribute interrupt code to distribute next. If set to 1, the router gives priority to lower
interrupt numbers when deciding which interrupt code or interrupt acknowledgement code / extended interrupt
code to distribute.
11
Distributed interrupt code priority (IP) - When set to 0, all interrupt codes have priority over all interrupt
acknowledgement codes / extended interrupt codes, and will be distributed first. When set to 1, all interrupt
acknowledgement codes / extended interrupt codes have priority over all interrupt codes.
10
Auxiliary distributed interrupt code enable (AI) - If set to 1, distributed interrupt codes can be sent and received
on the auxiliary time code / distributed interrupt code interface. If set to 0, all distributed interrupt codes
received on the auxiliary interface are silently discarded, and no distributed interrupt code codes will be transmitted on the interface. Reset value depends on bootstrap signals, as described in section 3.1.
9
Auxiliary time-code enable (AT) - If set to 1, time-codes can be sent and received on the auxiliary time code /
distributed interrupt code interface. If set to 0, all time-codes received on the auxiliary interface are silently discarded, and no time-codes will be transmitted on the interface.
GR740-UM-DS, Jun 2016, Version 1.4
212
www.cobham.com/gaisler
GR740
8
Table 208. 0x00000A00 - RTR.RTRCFG - Router configuration / status
Distributed interrupt code enable (IE) - Global enable / disable for the distributed interrupt codes. If set to 0, all
received distributed interrupt codes will either be silently discarded (if RTRCFG.TF = 1), or handled as timecodes (if RTRCFG.TF = 0). When set to 1, whether or not distributed interrupt codes are received or transmitted
on a port depends on the setting of the register bits RTR.PCTRL.IC, RTR.PCTRL2.IR, RTR.PCTRL2.IT,
RTR.PCTRL2.AR, and RTR.PCTRL2.AT. Reset value depends on bootstrap signals, as described in section 3.1.
7
Reset (RE) - Resets the complete router when written with a 1. When this bit is written through RMAP, an
RMAP reply will not be sent, even if the reply bit in the RMAP commands Instruction field is set to 1. This bit is
self-clearing.
6
Enable extended interrupts (EE) - If set to 0, all distributed interrupt codes with bit 5 set to 1 are handled as
interrupt acknowledgement codes. If set to 1, all distributed interrupt codes with bit 5 set to 1 are handled as
extended interrupt codes, i.e with interrupt identifiers 32-63.
5
RESERVED
4
Self addressing enable (SA) - If set to 1, ports are allowed to send packets to themselves. If set to 0, packets with
the same input port as output port are spilled, and an invalid address error is asserted for that port.
3
Time-code control flag mode (TF) - When set to 0, all received time code / distributed interrupt codes are handled as time-codes, no matter the value of the control flags (bits 7:6 of the code). When set to 1, the time-code
control flags must have value “00” to be considered valid time-codes. Note that the RTRCFG.IE bit has priority
over this bit, which means that if RTRCFG.IE is 1, then setting this bit to 0 has no impact. Reset value depends
on bootstrap signals, as described in section 3.1.
2
Memory error (ME) - Set to 1 when a memory error occur while accessing the on-chip memory used for the
routing table.
1
Timers available (TA) - Constant value 1. Indicates that the router has support for timers, as described in section
17.2.15.
0
SpaceWire Plug and Play available (PP) - Constant value 1. Indicates that the router supports SpaceWire Plug
and Play, as described in section 17.5.4.
Table 209. 0x00000A04 - RTR.TC - Time-code
31
10
RESERVED
9
8
7
RE EN
0x000000
0
r
1
6
5
0
CF
TC
0x0
0x00
r
r
rw* rw
31: 10
RESERVED
9
Reset time-code (RE) - When this field is written to 1, the RTR.TC.CF and RTR.TC.TC fields are reset. This bit
is self-clearing, and always reads 0. Writing 0 has no effect.
8
Enable time-codes (EN) - When set to 1, received time-codes are handled by the router according to the rules
described in 17.2.17. When set to 0, all received time-codes are silently discarded.
7: 6
Time-control flags (CF) - The current value of the router’s time-code control flags (bits 7:6 of the latest valid
time-code received).
5: 0
Time-counter (TC) - Current value of the router’s time counter.
Table 210. 0x00000A08 - RTR.VER - Version / instance ID
31
31: 24
24 23
16 15
8
7
2
1
0
MA
MI
PA
ID
0x01
0x02
0x00
0x04
*
r
r
r
rw
rw
Major version (MA) - Holds the major version number of the router. Constant value 0x01.
23: 16
Minor version (MI) - Holds the minor version number of the router. Constant value 0x02.
15: 8
Patch (PA) - Holds the patch number of the router. Constant value 0x00.
7: 0
Instance ID (ID) - Holds the instance ID number of the router. Reset value for bits 1:0 depends on bootstrap signals, as described in section 3.1.
GR740-UM-DS, Jun 2016, Version 1.4
213
www.cobham.com/gaisler
GR740
Table 211. 0x00000A0C - RTR.IDIV - Initialization divisor
31
8
7
0
RESERVED
ID
0x000000
0x27
r
rw
31: 8
RESERVED
7: 0
Initialization clock divisor (ID) - Clock divisor value used by all the SpaceWire links to generate the 10 Mbit/s
rate during initialization.
Table 212. 0x00000A10 - RTR.CFGWE - Configuration port write enable
31
1
RESERVED
0
WE
0x00000000
1
r
rw
31: 1
RESERVED
0
Configuration port write enable (WE) - When set to 1, write accesses to the configuration port area are allowed.
When set to 0, write accesses are only allowed to this register. RMAP write and RMAP read-modify-write commands will be replied to with the Status field set to 0x0A (authorization failure), if a reply was requested. The
value of this bit has no effect for SpaceWire Plug-and-Play commands.
Table 213. 0x00000A14 - RTR.PRESCALER - Timer prescaler reload
31
16 15
0
RESERVED
RL
0x0000
0xFFFF
r
rw*
31: 16
RESERVED
15: 0
Timer prescaler reload (RL) - Global prescaler reload value used for generating a common tick for the data character timers, auto-disconnect timers, and interrupt code distribution timers. The prescaler runs on the system
clock, and a tick is generated every RTR.PRESCALER.RL+1 CLK cycle. The minimum value of this field is
250. Trying to write a value less than that will result in 250 being written.
Table 214. 0x00000A18 - RTR.IMASK - Interrupt mask
31
11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
PE SR RS TT PL TS AC RE IA LE ME
0x00000
0
r
0
0
0
0
0
0
0
0
0
0
rw rw rw rw rw rw rw rw rw rw rw
31: 11
RESERVED
10
SpaceWire Plug-and-Play error (PE) - Generate an interrupt when a SpaceWire Plug and Play error has been
detected in the configuration port. The different errors are described in 17.5.4.
9
Spill-if-not-ready (SR) - Generate an interrupt when a packet has been spilled because of the spill-if-not-ready
feature described in section 17.2.10.
8
Run-state entry (RS) - Generate an interrupt when a SpaceWire link enters run-state.
7
Time code / distributed interrupt code tick truncation (TT) - Generate an interrupt when a packet has been spilled
because of the time code / distributed interrupt code truncation feature described in section 17.2.21.1.
GR740-UM-DS, Jun 2016, Version 1.4
214
www.cobham.com/gaisler
GR740
6
Table 214. 0x00000A18 - RTR.IMASK - Interrupt mask
Packet length truncation (PL) - Generate an interrupt when a packet has been spilled due to the packet length
truncation feature described in section 17.2.16.
5
Timeout spill (TS) - Generate an interrupt when a packet has been spilled due to the timeout mechanism.
4
Auxiliary configuration port error (AC) - Generate an interrupt when either a header CRC error, protocol ID
error, packet type error, early EOP, or early EEP has been detected in the configuration port.
3
RMAP error (RE) - Generate an interrupt when an error has been detected in the configuration port for an
RMAP command such that the PSTS.EC field is set to a non-zero value.
2
Invalid address (IA)- Generate an interrupt when an invalid address error has occurred on a port. See
RTR.PSTS:IA bit and section 17.2.12 for a definition of invalid address.
1
Link error (LE) - Generate an interrupt when a link error has been detected on a SpaceWire port.
0
Memory error (ME) - Generate an interrupt when a memory error occur in any of the router’s on-chip memories.
Table 215. 0x00000A1C - RTR.IPMASK - Interrupt port mask
31
13 12
0
RESERVED
IE
0x000
0x0000
r
rw
31: 20
RESERVED
19: 0
Port interrupt enable (IE) - Set a bit to 1 to enable interrupts to be generated for an error detected in the port with
the same number as the bit index. An interrupt is optionally signalled through a distributed interrupt code.
Table 216. 0x00000A20 - RTR.PIP - Port interrupt pending
31
13 12
0
RESERVED
IP
0x000
0x0000
r
wc
31: 20
RESERVED
19: 0
Interrupt pending (IP) - When a bit is set to 1, the port with the same number as the bit index was the source of
an interrupt. A bit in this field will only be set to 1 for a generated interrupt if the port’s corresponding bit in
RTR.IPMASK is set, as well as the error types corresponding bit in RTR.IMASK, are set.
Table 217. 0x00000A24 - RTR.ICODEGEN - Interrupt code generation
31
21 20 19 18 17 16 15
RESERVED
0x000
r
UA AH IT TE EN
0
0
0
1
0
rw rw rw rw rw
5
4
0
RESERVED
IN
0x000
0x00
r
rw
0
31: 21
RESERVED
20
Interrupt code generation un-acknowledge mode (UA) - If this bit is set to 1, an ISR timeout for a distributed
interrupt that was generated by the router will clear the bits in the RTR.PIP register that were set when the interrupt was generated. If this bit is set to 0, no extra handling is done on an ISR timeout event, and the bits in
RTR.PIP will stay set. See section 17.2.18.
19
Interrupt acknowledgement code handling (AH) - When set to 1, and the router has generated an interrupt code,
a received interrupt acknowledgement code with the interrupt number matching the RTR.ICODEGEN.IN field
will clear the bits in the RTR.PIP register that were set when the interrupt code was generated. If set to 0, no
extra handling of a received interrupt acknowledgement code is done and the bits in RTR.PIP will stay set. This
bit is unused when the distributed interrupts are operating in the extended interrupt mode. See section 17.2.18.
GR740-UM-DS, Jun 2016, Version 1.4
215
www.cobham.com/gaisler
GR740
18
Table 217. 0x00000A24 - RTR.ICODEGEN - Interrupt code generation
Interrupt type (IT) - 0 = Level. 1 = Edge. When set to 0, a new interrupt code is distributed as long as RTR.PIP
register is non zero. When set to 1, a new interrupt code is distributed only when a bit in RTR.PIP toggles from
0 to 1. See section 17.2.18.
17
Interrupt acknowledgement code to interrupt code timer enable (TE) - If set to 1, the router will wait for the time
period specified by the RTR.AITIMER register after the reception of an interrupt acknowledgement code (for
which the router generated the corresponding interrupt code) until a new interrupt code is allowed to be generated. If set to 0, the timer is not used, and a new interrupt code is allowed to be generated as soon as the interrupt
acknowledgement code has been distributed.
16
Interrupt code generation enable (EN) - When 1, interrupt code generation is enabled, and an interrupt code can
be distributed when an internal error event occurs. See section 17.2.18.
15: 6
RESERVED
4: 0
Interrupt number (IN) - Sets the interrupt identifier of the interrupt code that will be distributed when the interrupt code generation feature is enabled (RTR.ICODEGEN.EN = 1). See section 17.2.18.
Table 218. 0x00000A28 - RTR.ISR0 - Interrupt code distribution ISR register, interrupt 0-31
31
0
IB
0x00000000
wc
31: 0
Distributed interrupt code ISR bits (IB) - The current value of the interrupt code distribution ISR register for
interrupt numbers 0 - 31. Each bit index corresponds to the ISR bit value for the corresponding interrupt number.
A bit value of 1 indicates that an interrupt code with the corresponding interrupt number has been received, but
not yet acknowledged. A bit value of 0 indicates either that no interrupt code with the corresponding interrupt
number has been received, or that the previous interrupt code was either acknowledged or timed out. This register should be normally only be used for diagnostics and / or FDIR.
Table 219. 0x00000A2C - RTR.ISR1 - Interrupt code distribution ISR register, interrupt 32-63
31
0
IB
0x00000000
wc
31: 0
Distributed interrupt code ISR bits (IB) - The current value of the interrupt code distribution ISR register for
interrupt numbers 32 - 63. Each bit index corresponds to the ISR bit value for the corresponding interrupt number, minus 32, i.e bit 0 corresponds to interrupt number 32, bit 1 to interrupt number 33 etc. A bit value of 1 indicates that an interrupt code with the corresponding interrupt number has been received, but not yet
acknowledged. A bit value of 0 indicates either that no interrupt code with the corresponding interrupt number
has been received, or that the previous extended interrupt code timed out. This register should be normally only
be used for diagnostics and / or FDIR.
Table 220. 0x00000A30 - RTR.ISRTIMER - Interrupt code distribution ISR timer reload
31
16 15
0
RESERVED
RL
0x0000
0xFFFF
r
rw
GR740-UM-DS, Jun 2016, Version 1.4
216
www.cobham.com/gaisler
GR740
31: 16
15: 0
Table 220. 0x00000A30 - RTR.ISRTIMER - Interrupt code distribution ISR timer reload
RESERVED
Interrupt code distribution ISR timer reload (RL) - Interrupt code distribution ISR timer reload value, counted in
prescaler ticks. Each ISR bit has its own timer, which is started and reloaded with the value of this field when an
interrupt code with the corresponding interrupt number is received (or generated by the router). See section
17.2.18 for details on interrupt code distribution.
Table 221. 0x00000A34 - RTR.AITIMER - Interrupt code distribution ACK-to-INT timer reload
31
16 15
0
RESERVED
RL
0x0000
0x0000
r
rw
31: 16
RESERVED
15: 0
Interrupt acknowledgement code to interrupt code timer reload (RL) - Interrupt code distribution interrupt
acknowledgement code to interrupt code timer reload value, counted in prescaler ticks. When an interrupt
acknowledgement code is receive - for which the router generated the corresponding interrupt code - the interrupt acknowledgement code to interrupt code timer is started and reloaded with the value of this field. See section 17.2.18 for details on interrupt code distribution.
Table 222. 0x00000A38 - RTR.ISRCTIMER - Interrupt code distribution ISR change timer reload
31
5
4
0
RESERVED
RL
0x0000000
0x00
r
rw
31: 5
RESERVED
4: 0
Interrupt code distribution ISR change timer reload (RL) - Interrupt code distribution ISR change timer reload
value, counted in prescaler ticks. Each time an ISR bit change value, the corresponding ISR change timer is
started and reloaded with the value of this field. See section 17.2.18 for details on interrupt code distribution.
Table 223. 0x00000A40 - RTR.LRUNSTAT - Link running status
31
9
8
1
0
RESERVED
LR
R
0x000000
0x00
0
r
r
r
31: 19
RESERVED
18: 1
Link running status (LR)- Each bit is set to 1 when the link interface for the SpaceWire port with the same number as the bit index is in run-state. If the link interface is not in run-state, the bit is set to 0.
0
RESERVED
Table 224. 0x00000A44 - RTR.CAP - Capability
31
26 25 24 23 22
20 19 18
16 15 14 13 12 11 10
RESERVED
AF
R
PF
R
RM
R AS AX DP ID SD
0x00
0x3
0
0x3
0
0x0
0
0
1
0
1
r
r
r
r
r
r
r
r
r
r
r
31: 26
9
5
4
0
PC
CC
1
0x00
0x00
r
r
r
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
217
www.cobham.com/gaisler
GR740
25: 24
Table 224. 0x00000A44 - RTR.CAP - Capability
AMBA port word FIFO size (AF) - Shows the number of entries in the AMBA port’s 32-bit FIFOs. The number
of entries is determined by the value of this field, according to the formula: Entries = 2(RTR.CAP.PF+2). Constant
value of 0x3 = 32 entries.
23
RESERVED
22: 20
Port N-char FIFO size (PF) - Shows the number of entries in the port’s N-char FIFOs. The number of entries is
determined by the value of this field, according to the formula: Entries = 2(RTR.CAP.PF+4). Constant value of 0x3
= 128 entries.
19
RESERVED
18: 16
RMAP maximum data length (RM) - This field specifies the maximum data length in an RMAP read / write
command that the configuration port can handle. The length can be determined according to the formula: Length
= 2(RTR.CAP.RM+2). Constant value of 0x0 = 4 bytes.
15
RESERVED
14
Synchronous / asynchronous auxiliary interface (AS) - Constant value of 0, indicating that the auxiliary timecode / distributed interrupt code interface is synchronous to the system clock.
13
Auxiliary time-code / distributed interrupt code support (AX) - Specifies that the router has support for the auxiliary time-code / distributed interrupt code feature described in 17.2.19. Constant value of 1.
12
SpaceWire dual port (DP) - Constant value of 0, indicating that SpaceWire ports are not implemented with dual
ports.
11
Distributed interrupt support (ID) - Specifies that the router has support for the interrupt code distribution
scheme, described in 17.2.18. Constant value of 1.
10
SpaceWire-D support (SD) - Specifies that the router has support for the SpaceWire-D, described in section
17.2.21. Constant value of 1.
9: 5
Port packet counter bits (PC) - Specifies the number of bits in the port’s incoming / outgoing packet counters.
Constant value of 0x00 = no packet counters.
4: 0
Port character counter bits (CC) - Specifies the number of bits in the port’s incoming / outgoing character counters. Constant value of 0x00 = no character counters.
Table 225. 0x00000A50 - RTR.PNPVEND - SpaceWire Plug-and-Play - Device Vendor and Product ID
31
16 15
0
VI
PI
0x0003
0x0718
r
r
31: 16
SpaceWire Plug-and-Play Vendor ID (VI) - Double mapping of the VEND bits from the SpaceWire Plug-andPlay Device Vendor and Product ID field. See table 234.
25: 0
SpaceWire Plug-and-Play Product ID (PI) - Double mapping of the PROD bits from the SpaceWire Plug-andPlay Device Vendor and Product ID field. See table 234.
Table 226. 0x00000A54 - RTR.PNPUVEND - SpaceWire Plug-and-Play - Unit Vendor and Product ID
31
16 15
0
VI
PI
0x0000
0x0000
rw
rw
31: 16
SpaceWire Plug-and-Play Unit vendor ID (VI) - Double mapping of the VEND bits from the SpaceWire Plugand-Play Unit Vendor and Product ID field (see table 243).
25: 0
SpaceWire Plug-and-Play Unit product ID (PI) - Double mapping of the PROD bits from the SpaceWire Plugand-Play Unit Vendor and Product ID field (see table 243).
GR740-UM-DS, Jun 2016, Version 1.4
218
www.cobham.com/gaisler
GR740
Table 227. 0x00000A58 - RTR.PNPUSN - SpaceWire Plug-and-Play - Unit Serial Number
31
2
1
0
SN
0x00000000
*
rw
31: 0
SpaceWire Plug-and-Play Unit serial number (SN) - Double mapping of the SpaceWire Plug-and-Play Unit
Serial Number field (see table 244). Reset value for bits 1:0 depends on bootstrap signals, as described in section
3.1.
Table 228. 0x00000E00-0x00000E30 - RTR.MAXPLEN - Maximum packet length, ports 0-12
31
25 24
0
RESERVED
ML
0x00
0x000000
r
rw
31: 25
RESERVED
24: 0
Maximum packet length (ML) - Maximum length of packets for which the corresponding port is the input port.
This field is only used when the RTR.PCTRL.PL bit (RTR.PCTRLCFG.PL for port 0) is set to 1. See section
17.2.16 for details.
Table 229. 0x00000E84-0x00000EA0 - RTR.CREDCNT - Credit counter, ports 1-8
31
12 11
6
5
0
RESERVED
OC
IC
0x00000
0
0
r
r
r
31: 12
RESERVED
11: 6
Out credit counter (OC) - Number of outgoing credits. For each credit, the other end of the link is allowed to
send one N-Char.
5: 0
In credit counter (IC) - Number of incoming credits. For each credit, the port is allowed to transmit one N-Char.
NOTE: This register is only available for SpaceWire ports.
Table 230. 0x00001004-0x000013FC - RTR.RTCOMB - Routing table, combined port mapping and address control,
addresses 1-255
31 30 29 28 27
20 19
1
0
SR EN PR HD
RESERVED
PE
PD
N/R 0 N/R N/R
0x00
N/R
N/R
rw rw rw rw
r
rw
rw
31
Spill-if-not-ready (SR) - This bit is a double mapping of the RTR.RTACTRL.SR bit. See table 200.
30
Enable (EN) - This bit is a double mapping of the RTR.RTACTRL.EN bit. See table 200.
29
Priority (PR) - This bit is a double mapping of the RTR.RTACTRL.PR bit. See table 200.
28
Header deletion (HD) - This bit is a double mapping of the RTR.RTACTRL.HD bit. See table 200.
27: 20
RESERVED
19: 1
Port enable bits (PE) - This field is a double mapping of the RTR.RTPMAP.PE field. See table 199.
NOTE: This register is not available via on-chip AHB slave interface. Only available through RMAP.
GR740-UM-DS, Jun 2016, Version 1.4
219
www.cobham.com/gaisler
GR740
Table 230. 0x00001004-0x000013FC - RTR.RTCOMB - Routing table, combined port mapping and address control,
addresses 1-255
0
Packet distribution (PD) - This field is a double mapping of the RTR.RTPMAP.PD field. See table 199.
NOTE: See note for RTR.RTPMAP (table 199).
GR740-UM-DS, Jun 2016, Version 1.4
220
www.cobham.com/gaisler
GR740
17.5.4 SpaceWire Plug-and-Play interface
The configuration port supports parts of the SpaceWire Plug-and-Play protocol [SPWPNP]. The supported fields are listed in table 233, and explained in more detail in tables 234 through 248.
The SpaceWire Plug-and-Play protocol uses standard RMAP commands and replies with the same
requirements as presented in section 17.5.1, but with the following differences:
•
Protocol Identifier field of a command shall be set to 0x03.
•
A command’s address fields shall contain a word address. The SpaceWire Plug-and-Play
addresses are encoded as shown in table 231.
•
The increment bit in the command’s instruction field shall be set to 1, otherwise a reply with Status field set to 0x0A (authorization failure) is sent.
•
RMAP Read-modify-write command is replaced by a compare-and-swap operation. The command’s data fields shall contain the new data to be written, while the mask fields shall contain the
value that the current data must match in order for the new data to be written. If there is a mismatch, a reply with Status field set to 0x0A (authorization failure) is sent.
•
The reply packet’s Status field can contain the additional status codes described in table 232.
Table 231. SpaceWire Plug-and-Play address encoding
31
24 23
Application Index
19 18
Protocol Index
14 13
FieldSet ID
0
Field ID
Table 232. SpaceWire Plug-and-Play status codes
Value
Description
0xF0
Unauthorized access - A write, or compare-and-swap command arrived either when the router was
not configured (Device ID field = 0), or the command did not match the owner information saved in
the Link Information field and Owner Address fields.
0xF1
Reserved field set - A read, write, or compare-and-swap command’s address field points to a non
existing field set.
0xF2
Read-only field - A write, or compare-and-swap command’s address points to a read-only field.
0xF3
Compare-and-swap-only-field - A write command’s address points to a compare-and-swap-only
field.
Note that it is not possible to access the SpaceWire Plug-and-Play fields through the AHB slave interface, except for the fields that are double mapped into the configuration port’s address space (see section 17.5.3).
An access (read, write, or compare-and-swap) made either to a field outside the Device Information
service, or to a field in an undefined field set within the Device Information service, will generate a
reply with the Status field set to 0xF1. An access (read, write, or compare-and-swap) to an undefined
or unsupported field in one of the defined field sets, within the Device Information service, is not
GR740-UM-DS, Jun 2016, Version 1.4
221
www.cobham.com/gaisler
GR740
treated as an error, and the Status field of the reply will be 0x00. Possible write-data for such an access
is discarded, and possible read-data returned is always 0.
Table 233.SpaceWire Plug-and-Play support
SpW PnP
Address
Register name
Acronym
Service - Field set - Field
0x00000000
SpaceWire Plug-and-Play - Device Vendor
and Product ID
RTR.PNPVEND
Device Information - Device
Identification - Device Vendor
and Product ID
0x00000001
SpaceWire Plug-and-Play - Version
RTR.PNPVER
Device Information - Device
Identification - Version
0x00000002
SpaceWire Plug-and-Play - Device Status
RTR.PNPDEVSTS
Device Information - Device
Identification - Device Status
0x00000003
SpaceWire Plug-and-Play - Active Links
RTR.PNPACTLNK
Device Information - Device
Identification - Active Links
0x00000004
SpaceWire Plug-and-Play - Link Information
RTR.PNPLNKINFO
Device Information - Device
Identification - Link Information
0x00000005
SpaceWire Plug-and-Play - Owner Address 0
RTR.PNPOA0
Device Information - Device
Identification - Owner Address 0
0x00000006
SpaceWire Plug-and-Play - Owner Address 1
RTR.PNPOA1
Device Information - Device
Identification - Owner Address 1
0x00000007
SpaceWire Plug-and-Play - Owner Address 2
RTR.PNPOA2
Device Information - Device
Identification - Owner Address 2
0x00000008
SpaceWire Plug-and-Play - Device ID
RTR.PNPDEVID
Device Information - Device
Identification - Device ID
0x00000009
SpaceWire Plug-and-Play - Unit Vendor and
Product ID
RTR.PNPUVEND
Device Information - Device
Identification - Unit Vendor and
Product ID
0x0000000A
SpaceWire Plug-and-Play - Unit Serial Number
RTR.PNPUSN
Device Information - Device
Identification - Unit Serial Number
0x00004000
SpaceWire Plug-and-Play - Vendor String
Length
RTR.PNPVSTRL
Device Information - Vendor /
Product String - Vendor String
Length
0x00006000
SpaceWire Plug-and-Play - Product String
Length
RTR.PNPPSTRL
Device Information - Vendor /
Product String - Product String
Length
0x00008000
SpaceWire Plug-and-Play - Protocol Count
RTR.PNPPCNT
Device Information - Protocol
Support - Protocol Count
0x0000C000
SpaceWire Plug-and-Play - Application
Count
RTR.PNPACNT
Device Information - Application
Support- Application Count
Table 234. 0x00000000 - RTR.PNPVEND - SpaceWire Plug-and-Play - Device Vendor and Product ID
31
16 15
0
VEND
PROD
0x0003
0x0740
r
r
31: 16
Vendor ID (VEND) - SpaceWire vendor ID assigned to Cobham Gaisler. Constant value of 0x0003.
15: 0
Product ID (PROD) - Product ID assigned to GR740. Constant value of 0x0740
GR740-UM-DS, Jun 2016, Version 1.4
222
www.cobham.com/gaisler
GR740
Table 235. 0x00000001 - RTR.PNPVER - SpaceWire Plug-and-Play - Version
31
24 23
16 15
8
7
0
MAJOR
MINOR
PATCH
RESERVED
0x01
0x02
0x00
0x00
r
r
r
r
31: 24
Major version number (MAJOR) - Constant value of 0x01.
23: 16
Minor version number (MINOR) - Constant value of 0x02
15: 8
Patch / Build number (PATCH) - Constant value of 0x00.
7: 0
RESERVED
Table 236. 0x00000002 - RTR.PNPDEVSTS - SpaceWire Plug-and-Play - Device Status
31
8
7
0
RESERVED
STATUS
0x000000
0x00
r
r
31: 8
RESERVED
7: 0
Device status (STATUS) - Constant value of 0x00.
Table 237. 0x00000003 - RTR.PNPACTLNK - SpaceWire Plug-and-Play - Active Links
31
9
8
1
0
RESERVED
ACTIVE
R
0x000
0x00
0
r
r
r
31: 13
RESERVED
12: 1
Link active (ACTIVE) - If set to 1, the port with the same number as the bit index is running. If set to 0, the port
is not running. For the SpaceWire ports (ports 1-8), the corresponding bit will be set to 1 if the link interface is in
run-state and the port is not disabled through the Port Control register (RTR.PCTRL.DI = 0). For the AMBA
ports, the bit is set to 1 if RTR.PCTRL.DI = 0.
0
RESERVED
Table 238. 0x00000004 - RTR.PNPLNKINFO -SpaceWire Plug-and-Play - Link Information
31
31: 24
7
6
5
OLA
24 23 22 21 20
OAL
R
OL
16 15
RES
13 12
RL
8
T
U
R
4
LC
0
0x00
0x0
0
0x0
0x0
0x0
1
0
0
0xC
r
r
r
r
r
r
r
r
r
r
Owner logical address (OLA) - Shows the value of the Initiator Logical Address field from the last successful
compare-and-swap command that set the Device ID field.
23: 22
Owner address length (OAL) - Shows how many of the three Owner Address fields that contain valid data.
21
RESERVED
20: 16
Owner link (OL) - Shows the number of the port which was used for the last successful operation to set the value
of the Device ID field.
15: 13
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
223
www.cobham.com/gaisler
GR740
12: 8
Table 238. 0x00000004 - RTR.PNPLNKINFO -SpaceWire Plug-and-Play - Link Information
Return link (RL) - Shows the number of the port through which the reply to the current read command will be
transmitted.
7
Device type (T) - Constant value of 1, indicating that this device is a router.
6
Unit information (U) - Indicates if the unit identification information (Unit Vendor and Product ID field, and
Unit Serial Number field) are valid. 0 = invalid, 1 = valid. This bit will be 0 after reset / power-up. Once the Unit
Vendor and Product ID field has been written with a non-zero value, this bit will be set to 1.
5
RESERVED
4: 0
Link count (LC) - Shows the number of router ports. Constant value of 0xC.
Table 239. 0x00000005 - RTR.PNPOA0 - SpaceWire Plug-and-Play - Owner Address 0
31
0
RA
0x00000000
r
31: 0
Reply address (RA) - Shows byte 0-3 of the Reply Address from the last successful compare-and-swap command that set to the Device ID field. If there was no Reply Address, then this field is zero.
Table 240. 0x00000006 - RTR.PNPOA1 - SpaceWire Plug-and-Play - Owner Address 1
31
0
RA
0x00000000
r
31: 0
Reply address (RA) - Shows byte 4-7 of the Reply Address from the last successful compare-and-swap command that set to the Device ID field. If the Reply Address was four bytes or less, then this field is zero.
Table 241. 0x00000007 - RTR.PNPOA2 - SpaceWire Plug-and-Play - Owner Address 2
31
0
RA
0x00000000
r
31: 0
Reply address (RA) - Shows byte 8-11 of the Reply Address from the last successful compare-and-swap command that set to the Device ID field. If the Reply Address was eight bytes or less, then this field is zero.
Table 242. 0x00000008 - RTR.PNPDEVID - SpaceWire Plug-and-Play - Device ID
31
0
DID
0x00000000
cas
31: 0
Device ID (DID) - Shows the device identifier. After reset / power-up, or when this field is written to zero, the
router is not considered to have an owner. The same applies to the case when the port indicated by the OL bits in
the Link Information field is either disconnected, or disabled by setting the RTR.PCTRL.DI bit to 1. This field is
only writable through a compare-and-swap operation.
GR740-UM-DS, Jun 2016, Version 1.4
224
www.cobham.com/gaisler
GR740
Table 243. 0x00000009 - RTR.PNPUVEND - SpaceWire Plug-and-Play - Unit Vendor and Product ID
31
16 15
0
VEND
PROD
0x0000
0x0000
r
r
31: 16
Unit vendor ID (VEND) - Shows the unit vendor identifier. This field is read-only through the SpaceWire Plugand-Play protocol, however it is writable through RMAP and AHB (see section 17.5.3). When this field, or the
PROD field, is written with a non-zero value, the U bit in the Link Information field is set to 1.
15: 0
Unit product ID (VEND) - Shows the unit product identifier. This field is read-only through the SpaceWire
Plug-and-Play protocol, however it is writable through RMAP and AHB (see section 17.5.3). When this field, or
the VEND field, is written with a non-zero value, the U bit in the Link Information field is set to 1.
Table 244. 0x0000000A - RTR.PNPUSN - SpaceWire Plug-and-Play - Unit Serial Number
31
0
USN
0x00000000
r
31: 0
Unit serial number (USN) - Shows the unit serial number. This field is read-only through the SpaceWire Plugand-Play protocol, however it is writable through RMAP and AHB (see section 17.5.3).
Table 245. 0x00004000 - RTR.PNPVSTRL - SpaceWire Plug-and-Play - Vendor String Length
31
15 14
0
RESERVED
LEN
0x00000
0x0000
r
r
31: 15
RESERVED
14: 0
Vendor string length (LEN) - Constant value of 0, indicating that no vendor string is present.
Table 246. 0x00006000 - RTR.PNPPSTRL - SpaceWire Plug-and-Play - Product String Length
31
15 14
0
RESERVED
LEN
0x00000
0x0000
r
r
31: 15
RESERVED
14: 0
Product string length (LEN) - Constant value of 0, indicating that no product string is present.
Table 247. 0x00008000 - RTR.PNPPCNT - SpaceWire Plug-and-Play - Protocol Count
31
GR740-UM-DS, Jun 2016, Version 1.4
5
4
0
RESERVED
PC
0x0000000
0x00
r
r
225
www.cobham.com/gaisler
GR740
31: 5
4: 0
Table 247. 0x00008000 - RTR.PNPPCNT - SpaceWire Plug-and-Play - Protocol Count
RESERVED
Protocol count (PC) - Constant value of 0, indicating that no protocols can be managed by using SpaceWire
Plug-and-Play.
Table 248. 0x0000C000 - RTR.PNPACNT - SpaceWire Plug-and-Play - Application Count
31
8
7
0
RESERVED
AC
0x000000
0x00
r
r
31: 8
RESERVED
7: 0
Application count (AC) - Constant value of 0, indicating that no applications can be managed by using SpaceWire Plug-and-Play.
GR740-UM-DS, Jun 2016, Version 1.4
226
www.cobham.com/gaisler
GR740
18
IOMMU - Bridge connecting Master I/O AHB bus
18.1
Overview
The IOMMU is a bridge that connects the Master I/O AHB bus to the Processor AHB bus and to the
Memory AHB bus. AHB transfer forwarding is performed in one direction, where AHB transfers to
the slave interface are forwarded to one of the master interfaces. The core can be configured to provide access protection and address translation for AMBA accesses traversing over the core. Access
protection can be provided using a bit vector to restrict access to memory. Access protection and
address translation can also be provided using page tables in main memory, providing full IOMMU
functionality. Both protection strategies allow devices to be placed into eight groups that share data
structures located in main memory. The protection and address translation functionality provides protection for memory assigned to processes and operating systems from unwanted accesses by units
capable of direct memory access.
Applications of the core include system partitioning, clock domain partitioning, system expansion and
secure software partitioning.
Features offered by the core include:
18.2
•
Single and burst AHB transfer forwarding
•
Access protection and address translation that can provide full IOMMU functionality
•
Devices can be placed into groups where a group shares page tables / access restriction vectors
•
Hardware table-walk
•
Efficient bus utilization through data prefetching and posted writes
•
Read and write combining, improves bus utilization and allows connecting cores with differing
AMBA access size restrictions.
Bridge operation
18.2.1 General
The first sub sections below describe the general AHB bridge function. The functionality providing
access restriction and address translation is described starting with section 18.3. In the description of
AHB accesses below the core propagates accesses from the Master I/O AHB bus to one of its master
interfaces (Processor AHB bus or Memory AHB bus).
The core occupies the full 4 GiB AMBA address space on the Master I/O AHB bus and is capable of
handling single and burst transfers generated by the AHB masters on the Master I/O bus.
For AHB write transfers write data is always buffered in an internal FIFO implementing posted
writes. For AHB read transfers the core uses GRLIB’s AMBA Plug&Play information to determine
whether the read data will be prefetched and buffered in an internal FIFO. If the target address for an
AHB read burst transfer is a prefetchable location the read data will be prefetched and buffered.
The core will insert wait states when handling an access. The core will still issue RETRY when the
core is busy emptying it’s write buffer on the master side.
18.2.2 Multi-bus bridge
The bridge has two AHB master interfaces connected to separate AHB buses. The bus select fields in
the bridge’s Master configuration registers allows the user to select which AHB master interface that
should be used for accesses initiated by a specific master on the Master I/O AHB bus. This selection
can be overriden by a field in the IOPTE when IOMMU protection is enabled.
GR740-UM-DS, Jun 2016, Version 1.4
227
www.cobham.com/gaisler
GR740
The control register field LB selects which AHB master interfaces that should be used when the core
fetches IOPTEs or APV bit vector data from memory (protection data structures described under sections 18.4 and 18.5).
18.2.3 AHB read transfers
When a read transfer is registered on the slave interface connected to the Master I/O AHB bus, the
core will insert wait states. The master interface then requests the bus and starts the read transfer on
the master side. Single transfers on the slave side are normally translated to single transfers with the
same AHB address and control signals on the master side.
If the transfer is a burst transfer to a prefetchable location, the master interface will prefetch data in
the internal read FIFO. If the burst on the slave side was an incremental burst of unspecified length
(INCR), the length of the burst is unknown. In this case the master interface performs an incremental
burst up to a 32-byte address boundary. When the burst transfer is completed on the master side, the
core will return data with zero wait states.
If the burst is to a non-prefetchable area, the burst transfer on the master side is performed using
sequence of NONSEQ, BUSY and SEQ transfers. The first access in the burst on the master side is of
NONSEQ type. Since the master interface can not decide whether the burst will continue on the slave
side or not, the system bus is held by performing BUSY transfers. On the slave side, the master that
initiated the transfer is allowed in bus arbitration. The first access in the transfer is completed by
returning read data. The next access in the transfer on the slave side is extended by asserting
HREADY low. On the master side the next access is started by performing a SEQ transfer (and then
holding the bus using BUSY transfers). This sequence is repeated until the transfer is ended on the
slave side.
In case of an ERROR response on the master side the ERROR response will be given for the same
access (address) on the slave side. SPLIT and RETRY responses on the master side are re-attempted
until an OKAY or ERROR response is received.
18.2.4 AHB write transfers
The core implements posted writes. During the AHB write transfer on the slave side the data is buffered in the internal write FIFO and the transfer is completed on the slave side by always giving an
OKAY response. The master interface requests the bus and performs the write transfer when the master bus is granted. If the burst transfer crosses the 32-byte write burst address boundary, a RETRY
response is given. When the core has written the contents of the FIFO out on the master side, the core
will allow the master on the slave side to perform the remaining accesses of the write burst transfer.
18.2.5 Read and write combining
Read and write combining allows the core to assemble or split AMBA accesses on the core’s slave
interface into one or several accesses on the master interface. The effects of read and write combining
is shown in the table below.
Table 249.Read and write combining
Access on slave interface
Resulting access(es) on master interface
BYTE or HALF-WORD single read
access to any area
Single access of same size
BYTE or HALF-WORD read burst
to prefetchable area
Incremental read burst of same access size as on slave interface, the length is the
same as the number of 32-bit words in the read buffer, but will not cross the read
burst boundary.
BYTE or HALF-WORD read burst
to non-prefetchable area
Incremental read burst of same access size as on slave interface, the length is the
same as the length of the incoming burst. The master interface will insert BUSY
cycles between the sequential accesses.
GR740-UM-DS, Jun 2016, Version 1.4
228
www.cobham.com/gaisler
GR740
Table 249.Read and write combining
Access on slave interface
Resulting access(es) on master interface
BYTE or HALF-WORD single
write
Single access of same size
BYTE or HALF-WORD write burst Incremental write burst of same size and length, the maximum length is the number
of 32-bit words in the write FIFO.
Single read access to any area
Single access of same size
Read burst to prefetchable area
Burst of 128-bit accesses up to 32-byte address boundary.
Read burst to non-prefetchable area
Incremental read burst of same access size as on slave interface, the length is the
same as the length of the incoming burst. The master interface will insert BUSY
cycles between the sequential accesses.
Single write
Single write access of same size
Write burst
Burst write of maximum possible size. The core will use the maximum size (up to
128-bit) that it can use to empty the write buffer.
Read and write combining is disabled for accesses to the area 0xF0000000 - 0xFFFFFFFF to prevent
accesses wider than 32 bits to register areas.
18.2.6 Core latency
This section deals with latencies in the core’s bridge function. Access protection mechanisms may add
additional delays, please refer to the description of access protection for a description of additional
delays when access protection and/or address translation is enabled.
Table 250 further down shows core behaviour for a single read access.
Table 250.Example of single read
Clock cycle
Core slave side activity
Core master side activity
0
Discovers access and transitions from idle state
Idle
1
Slave side waits for master side, wait states are
inserted on the AMBA bus.
Discovers slave side transition. Master interface output
signals are assigned.
2
Bus access is granted, perform address phase.
3
Register read data and transition to data ready state.
4
Discovers that read data is ready, assign
HREADY output register and data output register.
5
HREADY is driven on AMBA bus. Core has
returned to idle state
Idle
While the transitions shown in table 250 are simplified they give an accurate view of the core delay. If
the master interface needs to wait for a bus grant or if the read operation receives wait states, these
cycles must be added to the cycle count in the tables.
Table 251 below lists the delays incurred for single operations that traverse the bridge while the bridge
is in its idle state. The second column shows the number of cycles it takes the master side to perform
the requested access, this column assumes that the master slave gets access to the bus immediately
and that each access is completed with zero wait states. The table only includes the delay incurred by
traversing the core. For instance, when the access initiating master reads the core’s prefetch buffer,
each additional read will consume one clock cycle. However, this delay would also have been present
if the master accessed any other slave.
Write accesses are accepted with zero wait states if the bridge is idle, this means that performing a
write to the idle core does not incur any extra latency. However, the core must complete the write
operation on the master side before it can handle a new access on the slave side. If the core has not
GR740-UM-DS, Jun 2016, Version 1.4
229
www.cobham.com/gaisler
GR740
transitioned into its idle state, pending the completion of an earlier access, the delay suffered by an
access be longer than what is shown in the tables in this section. Accesses may also suffer increased
delays during collisions when the core has been instantiated to form a bi-directional bridge. Locked
accesses that abort on-going read operations will also mean additional delays.
Note that since the core has support for read and/or write combining, the number of cycles required
for the master will change depending on the access size and length of the incoming burst access.
Table 251.Access latencies
Access
Master acc. cycles Slave cycles
Delay incurred by performing access over core
Single read
3
1
4 * clkmst
Burst read with prefetch
2 + (burst length)x
2
2 * clkslv + (2 + burst length)* clkmst
Single writexx
(2)
0
0
Burst writexx
(2 + (burst length))
0
0
x A prefetch
xx The
18.3
operation ends at the address boundary defined by the prefetch buffer’s size
core implements posted writes, the number of cycles taken by the master side can only affect the next access.
General access protection and address translation
18.3.1 Overview
The core provides two types of access protection. The first option is to use a bit vector to implement
access restriction on a memory page basis. The second option is to use a page-table to provide access
restriction and address translation. Regardless of the protection strategy, the core provides means to
assign masters on the Master I/O AHB bus in groups where each group can be associated with a data
structure (access restriction vector or page table) in memory. The core supports a dynamically configurable page size from 4 to 512 KiB.
When a master on the Master I/O AHB bus initiates an access to be propagated, the bridge will first
look at the incoming master’s group assignment setting to determine to which group the master
belongs. When the group is known, the bridge can propagate or inhibit the access based on the group’s
attributes, or determine the address of the in-memory data structures to use for access checks (and
possibly address translation). The in-memory data structure may be cached by the bridge, otherwise
the information will be fetched from main memory.
Once the bridge has the necessary information to process the incoming access, the access will be
either allowed to propagate through the core or, in case the access is to a restricted memory location,
be inhibited. If the access is inhibited, the bridge will issue an AMBA ERROR response to the master
if the incoming access is a read access. The bridge implements posted writes, therefore write operations will not receive an AMBA ERROR response. An interrupt can, optionally, be asserted when an
access is inhibited. The AHB failing access register can be configured to log the first or most recent
access that was inhibited.
It is possible for masters to access the bridge’s register interface through the bridge. In this case the
bridge will perform an access to itself over the Processor and Slave I/O AHB buses.
GR740-UM-DS, Jun 2016, Version 1.4
230
www.cobham.com/gaisler
GR740
18.3.2 Delays incurred from access protection
The time required for the core’s master interface to start an access may be delayed by access protection checks. Table 252 below shows the added delays, please refer to section 18.2.6 for a description
of delays from the core’s bridge operation.
Table 252.Access protection check latencies
Protection mode
Delay in clock cycles on master side
Disabled
0
Write-protection only and read access
0
Master assigned to group in passthrough or inactive group
1
Access Protection Vector, cache hit
1
Access Protection Vector cache miss, cache disabled/not implemented
Minimumx 4 clock cycles
IOMMU Protection, cache hit
1
IOMMU Protection, TLB miss, TLB disabled/not implemented
Minimumx 4 clock cycles
x The core may suffer additional AMBA bus delays when accessing the vector in memory. 4 cycles is the minimum time
required and assumes that the core is instantly granted access to the bus and that data is delivered with zero wait states.
18.4
Access Protection Vector
The Access Protection Vector (APV) consists of a continuous bit vector where each bit determines the
access rights to a memory page. The bit vector provides access restriction on the full 4 GiB AMBA
address space. The required size of the bit vector depends on the page size used by the core, see table
below:
Table 253.Bit vector size vs. page size
Page size
Bit vector size
4 KiB
128 KiB
8 KiB
64 KiB
16 KiB
32 KiB
32 KiB
16 KiB
64 KiB
8 KiB
128 KiB
4 KiB
256 KiB
2 KiB
512 KiB
1 KiB
Each group can have a bit vector with a base address specified by a field in the group’s Group Control
Register. When a master performs an access to the core, the master’s group number is used to select
one of the available bit vectors. The AMBA access size used to fetch the vector is fixed to quad-word
(128-bits) and can be read out from the core’s Capability register 1. When the AMBA access size to
use is 128-bits and the page size is 4 KiB, bits 31:19 of the incoming address (HADDR) are used to
index a word in the bit vector, and bits HADDR[18:12] are used to select one of the 128 bits in the
fetched data. For each increase in page size one bit less of the physical address is used.
The lowest page is protected by the most significant bit in the bit vector. This means that page 0 is
protected by the most significant bit in byte 0 read from the bit vector’s base address (using big endian
addressing). When performing WORD accesses, the lowest page is protected by bit 31 in the accessed
word (using the bit numbering convention used throughout this document). When performing
4WORD (128-bit) accesses, the lowest page is protected by bit 127 in the accessed word. This allows
the same bit vector layout regardless of access size used by the IOMMU to fetch bit vector data.
GR740-UM-DS, Jun 2016, Version 1.4
231
www.cobham.com/gaisler
GR740
If the bit at the selected position is ‘0’, the access to the page is allowed and the core will propagate
the access. If the selected bit is ‘1’, and the access is an read access, an AMBA ERROR response is
given to the master initiating the access. If the selected bit is ‘1’, and the access is a write access, the
write is inhibited (not propagated through the bridge).
18.4.1 Access Protection Vector cache
The core has internal memory that can cache the Access Protection Vector. The cache has 32 lines
where each line is 16 bytes. These parameters can be read via Capability registers 0 and 1. The RAMs
in the APV cache are shared with the IOMMU TLB.
The cache is implemented as a direct-mapped cache built up of one data RAM and one tag RAM. The
number of locations in each RAM is the number of lines in the cache. The width of the data RAM
(cache line size) is the same as the size of the AMBA accesses used to fetch the APV from main memory. The address used to select a position in the RAMs, called the set address, has log2(number of
lines in the cache) = 5 bits.
The core will only cache bit vector data for accesses to the memory area 0x00000000 - 0x7FFFFFFF
(SDRAM memory area). Capability register 1 contains an address and a mask that describes this area.
Bit vector data for the specified memory range will be cached by the core. Bit vector data for accesses
made outside the memory range will not be placed in the cache, and will instead be fetched for memory on each access.
The number of address bits taken from the physical address required to uniquely address one position
in the bit vector depends on the cache line size and the page size. The number of required bits is
shown in table 254 below.
Table 254.Cache line size vs. physical address bits
Bits of physical address needed to identify one position depending on page size
Cache line
size in bits
4 KiB
8 KiB
16 KiB 32 KiB 64 KiB 128 KiB
256 KiB
512 KiB
128
12
11
10
6
5
9
8
7
As the cache is not large enough to hold a copy of each position in the bit vector, part of the physical
address and group will be placed in the cache tag RAM instead. The arrangement will be:
Table 255. Set address/ TAG arrangement
Set address:
31
4
Not present
0
Low bits of physical
address
Contents of Tag RAM:
10
Not present
0
8
7
Group ID
1
0
High bits of physical address
V
Valid (V) - Signals that addressed position in cache contains valid data
Since the physical address is used as the set address, accesses from a master assigned to one group
may evict cached bit vector data belonging to another group. This may not be wanted in systems
where interference between groups of masters should be minimized. In order to minimize inter-group
interference, the core can use the group ID in the set address, this functionality is called group-setaddressing:
Table 256. Group set addressing: Set address/TAG arrangement
Set address:
31
GR740-UM-DS, Jun 2016, Version 1.4
4
232
3
2
0
www.cobham.com/gaisler
GR740
Table 256. Group set addressing: Set address/TAG arrangement
Not present
Low
phys.
Group ID
Contents of Tag RAM:
31
10
Not present
0
1
High bits of physical address
0
V
Valid (V) - Signals that addressed position in cache contains valid data
Group-set-addressing is enabled via the GS field in the core’s Control register.
18.4.2 Access Protection Vector cache flush operation
If the contents of a vector is modified the core cache must be flushed by writing to the TLB/Cache
Flush Register. The TLB/Cache Flush register contains fields to flush the entire cache or to flush the
lines belonging to a specified group. In order to flush entries for a specific group, group-set-addressing must be implemented and enabled. Performing a group flush without group-set-addressing may
only flush part of the cache and can lead to unexpected behavior.
The core will not propagate any transfers while a cache flush operation is in progress.
18.5
IO Memory Management Unit (IOMMU) functionality
The IOMMU functionality of the core provides address translation and access protection on the full 4
GiB AMBA address space. The size of the address range where addresses are translated is specified
by the IOMMU Translation Range (ITR) field in the core’s Control register:
Size of translated address range in MiB = 16 MiB * 2ITR
The maximum allowed value of the ITR field is eight, which means that the IOMMU can provide
address translation to an area of size 16*28 = 4096 MiB, which is the full 32-bit address space. When
ITR is set to eight and a page size of 4 KiB is used, bits 31:12 of the incoming IO address are translated to physical addresses, using IO Page Tables entries describes below. Bits 11:0 of the incoming
access are propagated through the IOMMU. For each increase in page size one more bit will be
directly propagated through the IOMMU instead of being translated.
If ITR is less then eight then the most significant bits of the IO address must match the value of the
TMASK field in Capability register 2. If an access is outside the range specified by TMASK the
access will be inhibited. Table 257 shows the the effect of different ITR values. As an example, with
ITR set to 2, the IOMMU will perform address translation for a range that spans 64 MiB. This range
will be located at offset TMASK[31:26]. Accesses to addresses that do not have their most significant
bits set to match TMASK[31:26] will be inhibited. The table also shows the number of pages within
the decoded range and the memory required to hold the translation information (page tables) in main
memory. The pgsz value is the value of the PGSZ field in the control register.
GR740-UM-DS, Jun 2016, Version 1.4
233
www.cobham.com/gaisler
GR740
Table 257.Effects of IOMMU Translation Range setting
ITR
Size of translated range
TMASK bits used
0
16 MiB
1
Number of pages
Size of page tables
TMASK[31:24]
pgsz
4096 / 2
16 / 2pgsz KiB
32 MiB
TMASK[31:25]
8192 / 2pgsz
32 / 2pgsz KiB
2
64 MiB
TMASK[31:26]
16384 / 2pgsz
64 / 2pgszKiB
3
128 MiB
TMASK[31:27]
32768 / 2pgsz
128 / 2pgsz KiB
4
256 MiB
TMASK[31:28]
655536 / 2pgsz
256 / 2pgsz KiB
5
512 MiB
TMASK[31:29]
131072 / 2pgsz
512 / 2pgsz KiB
6
1024 MiB
TMASK[31:30]
262144 / 2pgsz
1 / 2pgsz MiB
7
2048 MiB
TMASK[31]
524288 / 2pgsz
2 / 2pgsz MiB
8
4096 MiB
TMASK not used
1048576 / 2pgsz
4 / 2pgsz MiB
18.5.1 IO Page Table Entry
Address translation is performed by looking up translation information in a one-level table present in
main memory. Part of the incoming address is used to index the table that consists of IO Page Table
Entries. The format of an IO Page Table Entry (IOPTE) is shown in table 258 below.
Table 258. IOMMU Page Table Entry (IOPTE)
31
8
PPAGE
31:8
7
C
6
5
R
2
1
0
BO BS W
4
3
V
R
Physical Page (PPAGE) - Bits 27:8 of this field corresponds to physical address bits 31:12 of the
page. With a 4 KiB page size, PPAGE[27:8] is concatenated with the incoming IO address bits [11:0]
to form the translated address. For each increase in page size one bit less of PPAGE is used and one
bit more of the incoming IO address is used: this means that with a 16 KiB page size ,
PPAGE[27:10] will be concatenated with the incoming IO address bits [13:0] to form the translated
address.
Bits 31:27 of this field are currently discarded by the IOMMU and are present in the data structure
for forward compatibility with systems using 36-bit AMBA address space.
7
Cacheable (C) - This field is currently not used by the IOMMU
6:5
RESERVED
4
Bus select Override (BO) - If this field is set to ‘1’ then the bus selection is made via the IOPTE.BS
field instead of the per master selection in the Master Configuration register.
3
Bus Select (BS) - Overrides master configuration register BS field when BO field in this IOPTE is
set.
BS = ‘0’ routes traffic over the Processor AHB bus. BS = ‘1’ routes traffic to the Memory AHB bus.
2
Writeable (W) - If this field is ‘1’ write access is allowed to the page. If this field is ‘0’, only read
accesses are allowed.
1
Valid (V) - If this field is ‘1’ the PTE is valid. If this field is ‘0’, accesses to the page covered by this
PTE will be inhibited.
0
RESERVED
When the core has IOMMU protection enabled all, incoming accesses from masters belonging to an
active group, which is not in pass-through mode, will be matched against TMASK. If an access is outside the range specified by ITR/TMASK, the access will be inhibited and may receive an AMBA
ERROR response (not applicable when the access is a posted write).
If the incoming access is within the range specified by ITR/TMASK, the core will use the incoming
IO address to index the page table containing the address translation information for the master/IO
GR740-UM-DS, Jun 2016, Version 1.4
234
www.cobham.com/gaisler
GR740
address. The core may be implemented with an Translation Lookaside Buffer (TLB) that may hold a
cached copy of the translation information. Otherwise the translation information will be fetched from
main memory. The base address of the page table to use is given by the Group Configuration register
to which the master performing the access is assigned. Please see the register description of the Group
Configuration register for constraints on the page table base address. The core will use bits X:Y to
index the table, where X depends on the value of the ITR field in the core’s Control register, and Y
depends on the page size (Y = 12 + PGSZ field in Control register).
When the core has fetched the translation information (IOPTE) for the accesses page it will check the
IOPTE’s Valid (V) and Writeable (W) fields. If the IOPTE is invalid, the access will be inhibited. If
the Writeable (W) field is unset and the access is a write access, the access will be inhibited. Otherwise the core will, for a page size of 4 KiB, use the IOPTE field PPAGE, bits 27:8, and bits 11:0 of the
incoming IO address to form the physical address to use when the access is propagated by the core
(physical address: PPAGE[27:8] & IOADDR[11:0]).
If the valid (V) bit of the IOPTE is ‘0’ the core may or may not store the IOPTE in the TLB. This is
controlled via the SIV field in the core’s Control register.
18.5.2 Prefetch operations and IOMMU protection
During normal bridge operation, and with Access Protection Vector protection, the core determines if
data for an access can be prefetched by looking at the IO address and the System bus plug and play
information. This operation cannot be done without introducing additional delays when the core is
using IOMMU protection. The incoming IO address must first be translated before it can be determined if the access is to a memory area that can be prefetched. In order to minimize delays the core
makes the assumption that any incoming burst access is to a prefetchable area. The result is that when
using IOMMU protection all burst accesses will result in the core performing a prefetch operation.
18.5.3 Translation Lookaside Buffer operation
The TLB is implemented as a direct-mapped cache with 32 entries, where each entry is 16 bytes, built
up of one data RAM and one tag RAM. The number of locations in each RAM is the number of
entries in the TLB. The width of the data RAM (entry size) is the same as the size of the AMBA
accesses used to fetch page table entries from main memory.
The address used to select a position in the RAMs, called the set address, has log2(number of entries
in the TLB) = 5 bits. The number of address bits taken from the physical address required to uniquely
address one position in the TLB depends on the page size. The number of required bits for each
allowed page size is shown in table 254 below, the values in the third to tenth column is the number of
address bits that must be used to accommodate the largest translatable range (maximum value of ITR
field in the core’s Control register). Note that an entry size larger than 32 bits results in an TLB that
holds multiple IOPTEs per entry.
Table 259.TLB entry size, page size
Entry
size in
bits
Entry
size in
IOPTEs
Bits of physical address needed to identify one position depending on page size
4 KiB
8 KiB
16 KiB
32 KiB
64 KiB
128 KiB
256 KiB
512 KiB
128
4
18
17
16
15
14
13
12
11
As the TLB is not large enough to hold a copy of each position in the page table, part of the physical
address and group will be placed in the tag RAM, the arrangement will be:
Table 260. Set address/TAG arrengement
Set address:
31
4
Not present
GR740-UM-DS, Jun 2016, Version 1.4
235
0
Low bits of physical
address
www.cobham.com/gaisler
GR740
Table 260. Set address/TAG arrengement
Contents of Tag RAM:
31
16
Not present
0
14 13
Group ID
1
High bits of physical address
0
V
Valid (V) - Signals that addressed position in cache contains valid data
Since the physical address is used as the set address, accesses from a master assigned to one group
may evict cached IOPTE’s belonging to another group. This may not be wanted in systems where
interference between groups of masters should be minimized. In order to minimize inter-group interference, the core can be implemented with support for using as much of the group ID as possible in
the set address, this functionality is called group-set-addressing:
Table 261. Group set address: Set address bits < (group ID bits) + (Physical address bits)
Set address:
31
Not present
4
2
Low
phys
Group ID
0
Contents of Tag RAM:
31
16
Not present
0
16
High bits of physical address
1
0
V
Valid (V) - Signals that addressed position in cache contains valid data
Group-set-addressing is enabled via the GS field in the core’s Control register.
18.5.4 TLB flush operation
If the contents of a page table is modified the TLB must be flushed by writing to the TLB/Cache Flush
Register. The TLB/Cache Flush register contains fields to flush the entire TLB or to flush the entries
belonging to a specified group. In order to flush entries for a specific group, group-set-addressing
must be implemented and enabled. Performing a group flush without group-set-addressing may only
flush part of the TLB and can lead to unexpected behavior.
When working in IOMMU mode, the core can be configured to not store a IOPTE in the TLB if the
IOPTE’s valid (V) bit is cleared. This behavior is controller via the SIV field in the core’s Control register.
The core will not propagate any transfers while a flush operation is in progress.
18.6
Fault-tolerance
The Access Protection Vector cache and IOMMU TLB are implemented with use byte-parity to protect entries in the cache/TLB. If an error is detected it will be processed as a cache/TLB miss and the
data will be re-read from main memory. A detected error will also be reported via the core’s status
register and the core also signals errors via its statistic output..
Errors can be injected in the Access Protection Vector cache and IOMMU TLB via the Data and Tag
RAM Error Injection registers.
GR740-UM-DS, Jun 2016, Version 1.4
236
www.cobham.com/gaisler
GR740
18.7
Statistics
The bridge has outputs connected to the LEON4 Statistics Unit. The core has the following statistics
outputs:
Table 262.IOMMU Statistics
Output
Description
hit
High for one cycle during TLB/cache hit.
miss
High for one cycle during TLB/cache miss
pass
High for one cycle during passthrough access
accok
High for one cycle during access allowed
accerr
High for one cycle during access denied
walk
High while core is busy performing a table walk or accessing the access protection vector
lookup
High while core is performing cache lookup/table walk
perr
High for one cycle when core detects a parity error in the APV cache
See section 31 for more information.
18.8
ASMP support
In some systems there may be a need to have separated instances of software each controlling a group
of masters. In this case, sharing of the IOMMU register interface may not be wanted as it would allow
software to modify the protection settings for a group of masters that belongs to another software
instance. To prevent this, the core’s register interface is mirrored on different 4 KiB pages. Different
write protection settings can be set for each mirrored block of registers. This allows use of a memory
management unit to control that software running can write to one, and only one, subset of registers.
Four ASMP register blocks are available. Each ASMP register block mirrors the standard register set
described in section 18.9 with the addition that some registers may be write protected. Table 263 contains a column that shows if a register is writable when accessed from an ASMP register block. The
core’s Control register, Master configuration register(s), Diagnostic cache registers, the ASMP access
control register(s) can never be written via ASMP register block. These registers are only available in
the first register set starting at the core register set base address. ASMP register block n is mapped at
an offset n*0x1000 from the core’s register base address.
Software should first set up the IOMMU and assign the masters into groups. Then the ASMP control
registers should be configured to constrain which registers that can be written from each ASMP block.
After this initialization is done, other parts of the software environment can be brought up.
As an example, consider the case where OS A will control masters 0, 1 and 4 while OS B will control
masters 2 and 3. In this case it may be appropriate to map masters 0, 1 and 4 to group 0 and master 2
and 3 to group 1. The ASMP access control registers can then be configured to only allow accesses to
the Group control register for group 0 from ASMP register block 1 and likewise only allow accesses
to the Group control register for group 1 from ASMP register block 2.
OS A will then map in ASMP register block 1 (registers within page located at core base offset +
0x1000) and OS B will then map in ASMP register block 2 (registers within page located at core base
offset + 0x2000). This way OS a will be able to change the base address and the properties of group 0,
containing its masters, without being able to change the protection mechanisms of group 1 belonging
to OS B. Note that since an OS is able to flush the TLB/cache it is able to impact the I/O performance
of masters assigned to other OS instances. Also note that care must be taken when clearing status bits
and setting the mask register that controls interrupt generation.
GR740-UM-DS, Jun 2016, Version 1.4
237
www.cobham.com/gaisler
GR740
18.9
Registers
The core is programmed through registers mapped into AHB I/O address space. All accesses to register address space must be made with word (32-bit) accesses.
Table 263.GRIOMMU registers
AHB address offset
Register
Writable in ASMP block
0x00
Capability register 0
No
0x04
Capability register 1
No
0x08
Capability register 2
No
0x0C
Reserved
-
0x10
Control register
No
0x14
TLB/cache flush register
Yes, protected*
0x18
Status register
Yes, protected*
0x1C
Interrupt mask register
Yes, protected*
0x20
AHB Failing Access register
No
0x24 - 0x3C
Reserved, must not be accessed
-
0x40 - 0x7C
Master configuration registers. 
Master n configuration register is located at offset 0x40 + n*0x4.
No
0x80-0xBC
Group control registers. 
Group n’s control register is located at offset 0x80 + n*0x4.
Yes, protected*
0xC0
Diagnostic cache access register
No
0xC4 - 0xE0
Diagnostic cache access data registers 0 - 7
No
0xE4
Diagnostic cache access tag register
No
0xE8
Data RAM error injection register
No
0xEC
Tag RAM error injection register
No
0xF0 - 0xFF
Reserved, must not be accessed
No
0x100 - 0x10C
ASMP access control registers.
The control register for ASMP block n is located at offset
0x100+n*0x4.
No
* Register is duplicated in ASMP register block at offset 0x1000 + register offset. The number of ASMP register blocks is
four. ASMP register block n starts at offset n*0x1000. Register is only writable if allowed by the corresponding ASMP
access control register field.
GR740-UM-DS, Jun 2016, Version 1.4
238
www.cobham.com/gaisler
GR740
18.9.1 Capability register 0
Table 264.0x00 - CAP0 - Capability register 0
31 30 29 28 27
A AC CA CP
24 23
20 19 18 17 16 15 14 13 12 11
9
8
RESERVED
NARB
CS
FT
ST
I
IT
IA
IP RESERVED MB
7
4
3
0
GRPS
MSTS
1
1
1
0
0
0x4
1
0b01
1
1
1
1
0
0
1
7
9
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
31
Access Protection Vector (A) - Read-only ‘1’, the core has support for Access Protection Vector
30
Access Protection Vector Cache (AC) - Read-only ‘1’, the core has a internal cache for Access Protection vector lookups.
29
Access Protection Vector Cache Addressing (CA): Read only ‘1’: Core supports using group ID as
part of cache set address
28
Access Protection Vector Cache Pipeline (CP) - Read-only ‘0’, core does not have a pipeline stage
added on the APV cache’s address.
27:24
RESERVED
23:20
ASMP Register Blocks (NARB) - Read-only 4. This field contains the number of ASMP register
blocks that the core implements. The core has 4 ASMP register blocks with the first block starting at
offset 0x1000 and the last block starting at offset 4*0x1000.
19
Configurable Page Size (CS) - Read-only ‘1’, the core supports several page sizes and the size is set
via the Control register field PGSZ.
18:17
Fault Tolerance (FT) - Read-only “01” - APV cache and/or IOMMU TLB is protected by parity
16
Statistics (ST) - Read-only ‘1’, the core collects statistics
15
IOMMU functionality enable (I) - Read-only ‘1’, the core has support for IOMMU functionality.
14
IOMMU TLB (IT) - Read-only ‘1’, the core has an IOMMU Translation Lookaside Buffer (TLB)
13
IOMMU Addressing (IA): Read-only ‘1’: Core supports using group ID as part of TLB set address
12
IOMMU TLB Address Pipeline (IP) - Read-only ‘0’, the core does not have a pipeline stage added
on the TLB’s address.
11:9
RESERVED
8
Multi-bus (MB) - Read-only ‘1’, the core is connected to two system buses (bus 0 is Processor AHB
and bus 1 is Memory AHB).
7:4
Number of groups (GRPS) - Number of groups that the core has been implemented to support - 1.
Value of GRPS is 7, the core supports eight groups.
3:0
Numbers of masters (MSTS) - Number of masters that the core has been implemented to support - 1.
Value of MSTS is 9, the core supports ten masters.
GR740-UM-DS, Jun 2016, Version 1.4
239
www.cobham.com/gaisler
GR740
18.9.2 Capability register 1
Table 265.0x04 - CAP1 - Capability register 1
31
20 19
16 15
8
7
5
4
0
CADDR
CMASK
CTAGBITS
CISIZE
CLINES
0
1
11
2
5
r
r
r
r
r
31:20
Access Protection Vector Cacheable Address (CADDR) - Read-only 0
19:16
Access Protection Vector Cacheable Mask (CMASK) - Read-only 1. Number of ‘1’s in the Access
Protection Vector Cachable mask. The CMASK field together with the CADDR field specify a
memory area protected by a part of the bit vector that can be cached by the core. The CMASK value
corresponds to the number of most significant bits of the CADDR field that are matched against the
incoming AMBA address when determining if the protection bits for the memory area should be
cached. As CMASK is 1 and CADDR is 0x000, the core will cache protection information for the
address range 0x00000000 - 0x7FFFFFFF.
15:8
Access Protection Vector Cache Tag bits (CTAGBITS) - Read-only 11. The width in bits of the
Access Protection Vector cache’s tags.
7:5
Access Protection Vector Access size (CSIZE) - Read-only 2. 128-bit (16 byte). This field indicates
the AMBA access size used when accessing the Access Protection Vector in main memory. This is
also the cache line size for the APV cache.
4:0
Access Protection Vector Cache Lines (CLINES) - Read-only 5. Number of lines in the Access Protection Vector cache. The number of lines in the cache is 2CLINES.= 32.
18.9.3 Capability register 2
Table 266.0x08 - CAP2 - Capability register 2
31
24 23
20 19 18 17 16 15
TMASK
RESERVED
MTYPE TTYPE
0xFF
0
0
r
r
r
8
7
5
4
0
TTAGBITS
ISIZE
TLBENT
0
0xF
0b100
0x5
r
r
r
r
31:24
Translation Mask (TMASK) - Read-only 0xFF. The incoming IO address bits IOADDR[31:24] must
match this field, depending on the setting of the ITR field in the core’s Control register, for an
address translation operation to be performed.
23:20
RESERVED
19:18
IOMMU Type (MTYPE) - Read-only 0, shows IOMMU implementation type.
17:16
TLB Type (TTYPE) - Read-only 0, shows implementation type of Translation Lookaside Buffer.
15:8
TLB Tag bits (TTAGBITS) - Read-only 16. The width in bits of the TLB tag.
7:5
IOMMU Access size (ISIZE) - Read only 0b100, 128-bit (16 byte). This field indicates the AMBA
access size used when accessing page tables in main memory. This is also the line size for the TLB.
4:0
TLB entries (TLBENT) - Read-only 5. Number of entries in the TLB. The number of entries is 2TLBENT
= 32.
GR740-UM-DS, Jun 2016, Version 1.4
240
www.cobham.com/gaisler
GR740
18.9.4 Control register
Table 267.0x10 - CTRL - Control register
31
21 20
RESERVED
18 17 16 15
PGSZ
LB SP
12 11 10
ITR
9
8
7
6
5
4
3
DP SIV HPROT AU WP DM GS CE
0
0
0
2
1
PM
0
EN
0
0
0
0
0
0
0
0
0
0
0
0
r
rw
rw rw
rw
rw rw
rw
rw rw rw rw rw
rw
rw
31:21
RESERVED
20:18
Page Size (PGSZ) - The value in this field determines the page size mapped by page table entries and
bit vector positions. Valid values are:
000: 4 KiB, 001: 8 KiB, 010: 16 KiB, 011: 32 KiB, 100: 64 KiB, 101: 128 KiB, 110: 256 KiB, 111:
512 KiB
17
Lookup bus (LB) - The value of this bit controls AHB master interface to use for fetching bit vector
and/or page table entries from memory when the core has been implemented with support for multiple buses (multiple AHB master interfaces). If this field is ‘0’, the first master interface will be used
for vector/table lookups. If this field is ‘1’, the second master interface will be used for lookups.
16
SPLIT support (SP) - This implementation of the bridge does not support use of AMBA SPLIT
responses. This bit is read-only with a value of ‘0’.
15:12
IOMMU Translation Range (ITR) - This field defines the size of the address range translated by the
core’s IOMMU functionality. The size of the decoded address range is 16 MiB * 2ITR and the
decoded memory area is located on an address with the most significant bits specified by the
TMASK field in Capability register 2, unless ITR = 8 in which case the whole address space is covered by the translated range.
11
Disable Prefetch (DP) - When this bit is ‘1’ the core will not perform any prefetch operations. This
bit is read only if the core has been implemented without support for prefetching data. During normal operation prefetch of data improves performance and should be enabled (the value of this bit
should be ‘0’). Prefetching may need to be disabled in scenarios where IOMMU protection is
enabled, which leads to a prefetch operation on every incoming burst access, and when the core is
used in bi-directional bridge configurations where dead locks may be resolved by the core dropping
prefetch data.
10
Save Invalid IOPTE (SIV) - If this field is ‘1’ the core will save IOPTEs that have their valid (V) bit
set to ‘0’ if the core has been implemented with a TLB. If this field is ‘0’ the core will not buffer an
IOPTE with valid (V) set to ‘0’ and perform an page table lookup every time the page covered by the
IOPTE is accessed. If the value of this field is changed, a TLB flush must be made to remove any
existing IOPTEs from the core’s internal buffer. Also if this field is set to ‘0’, any diagnostic accesses
to the TLB should not set the IOPTE valid bit to ‘0’ unless the Tag valid bit is also set to ‘0’.
9:8
HPROT encoding (HPROT) - The value of this field will be assigned to the AMBA AHB HPROT
signal bits 3:2 when the core is fetching protection data from main memory. HPROT(3) signals if the
access is cacheable and HPROT(2) signals if the access is bufferable.
7
Always Update (AU) - If this bit is set to ‘0’ the AHB failing access register will only be updated if
the Access Denied (AD) bit in the Status register is ‘0’ when the access is denied. Otherwise the
AHB failing access register will be updated each time an access is denied, regardless of the Access
Denied (AD) bit’s value.
6
Write Protection only (WP) - If this bit is set to ‘1’ the core will only used the Access Protection Vector to protect against write accesses. Read accesses will be propagated over the core without any
access restriction checks. This will improve the latency for read operations.
This field has no effect when the core is using IOMMU protection (PM field = “01”). When using
IOMMU protection all accesses to the range determined by TMASK and ITR will be checked
against the page table, unless the access is from a master that is assigned to an inactive group or a
group in pass-through mode.
5
Diagnostic Mode (DM) - If this bit is set to ‘1’ the core’s internal buffers can be accessed via the
Diagnostic interface (see Diagnostic cache access register) when the DE field of the Status register
has been set by the core. Set this bit to ‘0’ to leave Diagnostic mode. While in this mode the core will
not forward any incoming AMBA accesses.
4
Group-Set-addressing (GS) - When this bit is set to ‘1’, the core will use the group number as part of
the Access Protection Vector cache set address.
3
Cache/TLB Enable (CE) - When this bit is set to ‘1’, the core’s internal cache/TLB is enabled.
GR740-UM-DS, Jun 2016, Version 1.4
241
www.cobham.com/gaisler
GR740
Table 267.0x10 - CTRL - Control register
2:1
Protection Mode (PM) - This value selects the protection mode to use. “00” selects Group Mode and/
or Access Protection Vector mode. “01” selects IOMMU mode.
0
Enable (EN) - Core enable. If this bit is set to 1 the core is enabled. If this bit is set to 0 the core is
disabled and in pass-through mode. After writing this bit software should read back the value. The
change has not taken effect before the value of this bit has changed. The bit transition may be
blocked if the core is in diagnostic access mode or otherwise occupied.
18.9.5 TLB/cache flush register
Table 268.0x14 - FLUSH - TLB/cache flush register
31
8
RESERVED
7
4
3
2
1
0
GF
F
0
0
0
r
rw rw
FGRP
RES
0
0
r
rw
31:1
RESERVED
7:4
Flush Group (FGRP) - This field specifies the group to be used for a Group Flush, see GF field
below.
3:2
RESERVED
1
Group Flush (GF) - When this bit is written to ‘1’ the cache entries for the group selected by the
FGRP field will be flushed. More precisely the core will use the FGRP field as (part of the) set
address when performing the flush. This flush option is only available if the core has support for
group set addressing (CA field of Capability register 1 is non-zero). This flush option must only be
used if the GS bit in the Control register is set to ‘1’, otherwise old data may still be marked as valid
in the Access Protection Vector cache or IOMMU TLB. This bit will be reset to ‘0’ when a flush
operation has completed. A flush operation also affects the FL and FC fields in the Status register.
0
Flush (F) - When this bit is written to ‘1’ the core’s internal cache will be flushed. This bit will be
reset to ‘0’ when a flush operation has completed. A flush operation also affects the FL and FC fields
in the Status register.
GR740-UM-DS, Jun 2016, Version 1.4
242
www.cobham.com/gaisler
GR740
18.9.6 Status register
Table 269.0x18 - STATUS - Status register
31
6
RESERVED
5
4
3
2
1
0
PE DE FC FL AD TE
0
0
r
0
0
0
0
0
wc wc wc wc wc wc
31:6
RESERVED
5
Parity Error (PE) - The core sets this bit to ‘1’ when it detects a parity error in the tag or data RAM
of the APV cache. This field is cleared by writing ‘1’ to this position, writes of ‘0’ have no effect.
4
Diagnostic Mode Enabled (DE) - If this bit is set to ‘1’ the core is in Diagnostic Mode where the
core’s internal buffers can be accessed via the Diagnostic access registers. While in this mode the
core will not forward any incoming AMBA accesses.
3
Flush Completed (FC) - The core sets this bit to ‘1’ when a flush operation completes. This field is
cleared by writing ‘1’ to this position, writes of ‘0’ have no effect.
2
Flush started (FL) - The core sets this bit to ‘1’ when a Flush operation has started. This field is
cleared by writing ‘1’ to this position, writes of ‘0’ have no effect.
1
Access Denied (AD) - The core denied an AMBA access. This field is cleared by writing ‘1’ to this
position, writes of ‘0’ have no effect.
0
Translation Error (TE) - The core received an AMBA ERROR response while accessing the bit vector or page tables in memory. This also leads to the incoming AMBA access being inhibited.
Depending on the status of the Control register’s AU field and this register’s AD field this may also
lead to an update of the AHB Failing Access register.
18.9.7 Interrupt mask register
Table 270.0x1c - IMASK - Interrupt mask register
31
6
RESERVED
5
4
3
2
1
0
PEI R FCI FLI ADI TEI
0
0
r
rw rw rw rw rw rw
0
0
0
0
0
31:6
RESERVED
5
Parity Error Interrupt (PEI) - If this bit is set to ‘1’ an interrupt will be generated when the PE bit in
the Status register transitions from ‘0’ to ‘1’.
4
RESERVED
3
Flush Completed Interrupt (FCI) - If this bit is set to ‘1’ an interrupt will be generated when the FC
bit in the Status register transitions from ‘0’ to ‘1’.
2
Flush Started Interrupt (FLI) - If this bit is set to ‘1’ an interrupt will be generated when the FL bit in
the Status register transitions from ‘0’ to ‘1’..
1
Access Denied Interrupt (ADI) - If this bit is set to ‘1’ an interrupt will be generated when the AD bit
in the Status register transitions from ‘0’ to ‘1’.
0
Translation Error Interrupt (TEI) - If this bit is set to ‘1’ an interrupt will be generated when the TE
bit in the Status register transitions from ‘0’ to ‘1’.
GR740-UM-DS, Jun 2016, Version 1.4
243
www.cobham.com/gaisler
GR740
18.9.8 AHB failing access register
Table 271.0x20 - AHBFAS - AHB failing access register
31
5
4
3
2
1
FADDR[31:5]
FW
FMASTER
0
0
0
r
r
r
0
31:5
Failing Address (FADDR[31:5]) - Bits 31:5 of IO address in access that was inhibited by the core.
This field is updated depending on the value of the Control register AU field and the Status register
AD field.
4
Failing Write (FW) - If this bit is set to ‘1’ the failed access was a write access, otherwise the failed
access was a read access. This field is updated depending on the value of the Control register AU
field and the Status register AD field.
3:0
Failing Master (FMASTER) - Index of the master that initiated the failed access. This field is
updated depending on the value of the Control register AU field and the Status register AD field.
18.9.9 Master configuration registers
Table 272.0x40 - 0x64 - MSTCFG0-9 - Master configuration register 0 - 9
31
24 23
12 11
5
4
BS
3
0
VENDOR
DEVICE
RESERVED
GROUP
*
*
0
0
0
r
r
r
rw
rw
31: 24
Vendor ID (VENDOR) - GRLIB Plug’n’play Vendor ID of master
23: 12
Device ID (DEVICE) - GRLIB Plug’n’play Device ID of master
11: 5
RESERVED
4
Bus select for master (BS) - Master n’s bus select register is located at register address offset 0x40 +
n*0x4. This field specifies the bus to use for accesses initiated by AHB master n. A ‘0’ in this field
routes master accesses to the Processor AHB bus. A ‘1’ in this field routes master accesses to the
Memory AHB bus.
3:0
Group assignment for master - Master n’s group assignment field is located at register address offset
0x40 + n*0x4. This field specifies the group to which a master is assigned.
GR740-UM-DS, Jun 2016, Version 1.4
244
www.cobham.com/gaisler
GR740
18.9.10 Group control registers
Table 273.0x80 - 0x9C - GRPCTRL - Group control register 0 - 7
31
4
3
2
1
0
BASE[31:4]
R
P AG
0
0
0
rw
r
rw rw
0
31: 4
Base address (BASE) - Group n’s control register is located at offset 0x80 + n*0x4. This field contains the base address of the data structure for the group. The data structure must start on a 16-byte
address boundary.
3: 2
RESERVED
1
Pass-through (P) - If this bit is set to ‘1’ and the group is active (see bit 0 below) the core will passthrough all accesses made by master in this group and not use the address specified by BASE to perform look-ups in main memory. Note that this also means that the access will pass through untranslated when the core is using IOMMU protection (even if the access is outside the translated range
defined by TMASK in Capability register 2).
If this bit is set to ‘0’, the core will use the contents in its cache, or in main memory, to perform
checks and possibly address translation on incoming accesses.
0
Active Group (AG) - Indicates if the group is active. If this bit is set to ‘0’, all accesses made by masters assigned to this group will be blocked.
If this bit is set to ‘1’, the core will check the P field of this register and possibly also the in-memory
data structure before allowing or blocking the access.
18.9.11 Diagnostic cache access register
Table 274.0xC0 - DIAGCTRL - Diagnostic cache access register
31 30 29
DA RW
0
22 21 20 19 18
RESERVED
0
0
rw* rw*
r
DP TP
0
0
rw* rw*
0
R
SETADDR
0
NR
r
rw*
31
Diagnostic Access (DA) - When this bit is set to ‘1’ the core will perform a diagnostic operation to
the cache address specified by the SETADDR field. When the operation has finished this bit will be
reset to ‘0’.
30
Read/Write (RW) - If this bit is ‘1’ and the A field is set to ‘1’ the core will perform a read operation
to the cache. The result will be available in the Diagnostic cache access tag and data register(s). If
this bit is set to ‘0’ and the A field is set to ‘1’, the core will write the contents of the Diagnostic
cache access tag and data registers to the internal cache.
29:22
RESERVED
21
Data Parity error (DP) - This bit is set to ‘1’ if a parity error has been detected in the word read from
the cache’s data RAM. This bit can be set even if no diagnostic cache access has been made and it
can also be set after a cache write operation. This bit is read-only.
20
Tag Parity error (TP) - This bit is set to ‘1’ if a parity error has been detected in the word read from
the cache’s tag RAM. This bit can be set even if no diagnostic cache access has been made and it can
also be set after a cache write operation. This bit is read-only.
19
RESERVED
18:0
Cache Set Address (SETADDR) - Set address to use for diagnostic cache access. When a read operation has been performed, this field should not be changed until all wanted data has been read from
the Diagnostic cache access data and tag registers. Changing this field invalidates the contents of the
data and tag registers.
* This register can only be accessed if STATUS.DE bit in is set to 1
GR740-UM-DS, Jun 2016, Version 1.4
245
www.cobham.com/gaisler
GR740
18.9.12 Diagnostic cache access data registers
Table 275.0xC4 - 0xE0 - DIAGD - Diagnostic cache access data register 0 - 7
31
0
CDATAn
NR
rw*
31:0
Cache data word n (CDATAn) - The core has 8 Diagnostic cache access data registers. Diagnostic
cache access data register n holds data bits [31+32*n:32*n] in the cache line.
When using APV protection then the fetched vector is reversed before it is written in the cache. This
means that bit 127 of the fetched vector is located in bit 0 of Diagnostic data access register 0.
* This register can only be accessed if the the STATUS.DE bit is set to 1
18.9.13 Diagnostic cache access tag register
Table 276.0xE4 - DIAGT - Diagnostic cache access tag register
31
0
TAG
V
NR
rw*
31:1
Cache tag (TAG) - The size of the tag depends on cache size. The contents of the tag depends on
cache size and addressing settings.
0
Valid (V) - Valid bit of tag
* This register can only be accessed if the the STATUS.DE bit is set to 1
18.9.14 Data RAM error injection register
Table 277.0xE8 - DERRI - Data RAM error injection register
31
0
DPERRINJ
0
rw
31:0
Data RAM Parity Error Injection (DPERRINJ) - Bit DPERRINJ[n] in this register is XOR:ed with
the parity bit for data bits [7+8*n:8*n] in the data RAM.
18.9.15 Tag RAM error injection register
Table 278.0xEC - TERRI - Tag RAM error injection register
31
0
TPERRINJ
0
rw
31:0
Tag RAM Parity Error Injection (TPERRINJ) - Bit TPERRINJ[n] in this register is XOR:ed with the
parity bit for tag bits [7+8*n:8*n] in the tag RAM.
GR740-UM-DS, Jun 2016, Version 1.4
246
www.cobham.com/gaisler
GR740
18.9.16 ASMP access control registers
Table 279.0x100 - 0x10C - ASMPCTRL - ASMP access control registers 0 - 3
31
19 18 17 16 15
RESERVED
FC SC MC
0
0
GRPACCSZCTRL
0
0
0
0
r
rw rw rw
rw
31: 19
RESERVED
18
Flush register access control (FC) - If this bit is set to ‘1’ in the ASMP control register at offset
0x100 + n*0x4 then the TLB/cache flush register in ASMP register block n is writable. Otherwise
writes to the TLB/cache flush register in ASMP register block n will be inhibited.
17
Status register access control (SC) - If this bit is set to ‘1’ in the ASMP control register at offset
0x100 + n*0x4 then the Status register in ASMP register block n is writable. Otherwise writes to the
Status register in ASMP register block n will be inhibited.
16
Mask register access control (MC) - If this bit is set to ‘1’ in the ASMP control register at offset
0x100 + n*0x4 then the Master register in ASMP register block n is writable. Otherwise writes to the
Mask register in ASMP register block n will be inhibited.
15:0
Group control register access control (GRPACCSZCTRL) - ASMP register block n’s group access
control field is located at register address offset 0x100 + n*0x4. This field specifies which of the
Group control registers that are writable from an ASMP register block. If GRPACCSZCTRL[i] in
the ASMP access control register at offset 0x100 + n*0x4 is set to ‘1’ then Group control register i is
writable from ASMP register block n.
GR740-UM-DS, Jun 2016, Version 1.4
247
www.cobham.com/gaisler
GR740
19
Gigabit Ethernet Media Access Controller (MAC)
19.1
Overview
Cobham Gaisler’s Gigabit Ethernet Media Access Controller (GRETH_GBIT) provides an interface
between an AMBA-AHB bus and an Ethernet network. It supports 10/100/1000 Mbit speed in both
full- and half-duplex. The AMBA interface consists of an APB interface for configuration and control
and an AHB master interface which handles the dataflow. The dataflow is handled through DMA
channels. There is one DMA engine for the transmitter and one for the receiver. Both share the same
AHB master interface.
The ethernet interface supports the MII and GMII interfaces which should be connected to an external
PHY. The GRETH also provides access to the MII Management interface which is used to configure
the PHY. Hardware support for the Ethernet Debug Communication Link (EDCL) protocol is also
provided. This is an UDP/IP based protocol used for remote debugging.
Some of the supported features for the DMA channels are Scatter Gather I/O and TCP/UDP over IPv4
checksum offloading for both receiver and transmitter.
The system contains two GRETH_GBIT cores. The AHB master interfaces are connected to the Master I/O AHB bus. The cores also gave dedicated EDCL interfaces connected to the Debug AHB bus.
The selection of which master interface to use for EDCL traffic is made via bootstrap signals.
APB
AHB
Ethernet MAC
ETH*_MDIO
Registers
MDIO
ETH*_MDC
Transmitter
DMA Engine
AHB Master
Interface
RAM
Transmitter
EDCL
Transmitter
EDCL
Receiver
Receiver
DMA Engine
ETH*_TXEN
ETH*_TXER
ETH*_TXD[7:0]
ETH*_TXCLK
ETH*_CRS
ETH*_COL
ETH*_GTXCLK
ETH*_RXDV
ETH*_RXER
ETH*_RXD[7:0]
ETH*_RXCLK
Receiver
RAM
Figure 25. Block diagram of the internal structure of the GRETH_GBIT
19.2
Operation
19.2.1 System overview
The GRETH_GBIT consists of 3 functional units: The DMA channels, MDIO interface and the
optional Ethernet Debug Communication Link (EDCL).
The main functionality consists of the DMA channels which are used for transferring data between an
AHB bus and an Ethernet network. There is one transmitter DMA channel and one Receiver DMA
channel. The operation of the DMA channels is controlled through registers accessible through the
APB interface.
GR740-UM-DS, Jun 2016, Version 1.4
248
www.cobham.com/gaisler
GR740
The MDIO interface is used for accessing configuration and status registers in one or more PHYs connected to the MAC. The operation of this interface is also controlled through the APB interface.
The EDCL provides read and write access to an AHB bus through Ethernet. It uses the UDP, IP and
ARP protocols together with a custom application layer protocol to accomplish this. The EDCL contains no user accessible registers and always runs in parallel with the DMA channels.
The Media Independent Interface (MII) and Gigabit Media Independent Interface (GMII) are used for
communicating with the PHY. More information can be found in section 19.7.
The EDCL and the DMA channels share the Ethernet receiver and transmitter. More information on
these functional units is provided in sections 19.3 - 19.6.
19.2.2 Protocol support
The GRETH_GBIT is implemented according to IEEE standard 802.3-2002. There is no support for
the optional control sublayer. This means that packets with type 0x8808 (the only currently defined
ctrl packets) are discarded.
19.2.3 Dedicated EDCL AHB master interface
The core has an additional master interface connected to the Debug AHB bus that can be used for the
EDCL. This master interface is enabled with the external signals GPIO[8] and GPIO[9]. These signals
are only sampled at reset and changes have no effect until the next reset. Note that the core can be
reset via the clock gating unit and that this will lead to the value of GPIO[9:8] being sampled. See section 3.1 for further information on bootstrap signals.
19.3
Tx DMA interface
The transmitter DMA interface is used for transmitting data on an Ethernet network. The transmission
is done using descriptors located in memory.
19.3.1 Setting up a descriptor.
A single descriptor is shown in table 280 and 281. The number of bytes to be sent should be set in the
length field and the address field should point to the data. There are no alignment restrictions on the
address field. If the interrupt enable (IE) bit is set, an interrupt will be generated when the packet has
been sent (this requires that the transmitter interrupt bit in the control register is also set). The interrupt will be generated regardless of whether the packet was transmitted successfully or not.
Table 280. Address offset 0x0 - GRETH_GBIT transmit descriptor word 0
31
21 20 19 18 17 16 15 14 13 12 11 10
RESERVED
UC TC IC MO LC AL UE IE WR EN
0
LENGTH
31: 21
RESERVED
20
UDP checksum (UC) - Calculate and insert the UDP checksum for this packet. The checksum is only
inserted if an UDP packet is detected.
19
TCP checksum (TC) - Calculate and insert the TCP checksum for this packet. The checksum is only
inserted if an TCP packet is detected.
18
IP checksum (IC) - Calculate and insert the IP header checksum for this packet. The checksum is
only inserted if an IP packet is detected.
17
More (MO) - More descriptors should be fetched for this packet (Scatter Gather I/O).
16
Late collision (LC) - A late collision occurred during the transmission (1000 Mbit mode only).
15
Attempt limit error (AL) - The packet was not transmitted because the maximum number of
attempts was reached.
14
Underrun error (UE) - The packet was incorrectly transmitted due to a FIFO underrun error.
GR740-UM-DS, Jun 2016, Version 1.4
249
www.cobham.com/gaisler
GR740
13
Table 280. Address offset 0x0 - GRETH_GBIT transmit descriptor word 0
Interrupt enable (IE) - Enable Interrupts. An interrupt will be generated when the packet from this
descriptor has been sent provided that the transmitter interrupt enable bit in the control register is set.
The interrupt is generated regardless if the packet was transmitted successfully or if it terminated
with an error.
12
Wrap (WR) - Set to one to make the descriptor pointer wrap to zero after this descriptor has been
used. If this bit is not set the pointer will increment by 8. The pointer automatically wraps to zero
when the 1 kB boundary of the descriptor table is reached.
11
Enable (EN) - Set to one to enable the descriptor. Should always be set last of all the descriptor
fields.
10: 0
LENGTH - The number of bytes to be transmitted.
Table 281. Address offset 0x4 - GRETH_GBIT transmit descriptor word 1
31
0
ADDRESS
31: 0
Address (ADDRESS) - Pointer to the buffer area from where the packet data will be loaded.
To enable a descriptor the enable (EN) bit should be set and after this is done, the descriptor should
not be touched until the enable bit has been cleared by the GRETH_GBIT. The rest of the fields in the
descriptor are explained later in this section.
19.3.2 Starting transmissions
Enabling a descriptor is not enough to start a transmission. A pointer to the memory area holding the
descriptors must first be set in the GRETH_GBIT. This is done in the transmitter descriptor pointer
register. The address must be aligned to a 1 kB boundary. Bits 31 to 10 hold the base address of
descriptor area while bits 9 to 3 form a pointer to an individual descriptor. The first descriptor should
be located at the base address and when it has been used by the GRETH_GBIT the pointer field is
incremented by 8 to point at the next descriptor. The pointer will automatically wrap back to zero
when the next 1 kB boundary has been reached (the descriptor at address offset 0x3F8 has been used).
The WR bit in the descriptors can be set to make the pointer wrap back to zero before the 1 kB boundary.
The pointer field has also been made writable for maximum flexibility but care should be taken when
writing to the descriptor pointer register. It should never be touched when a transmission is active.
The final step to activate the transmission is to set the transmit enable bit in the control register. This
tells the GRETH_GBIT that there are more active descriptors in the descriptor table. This bit should
always be set when new descriptors are enabled, even if transmissions are already active. The descriptors must always be enabled before the transmit enable bit is set.
19.3.3 Descriptor handling after transmission
When a transmission of a packet has finished, status is written to the first word in the corresponding
descriptor. The Underrun Error bit is set if the transmitter RAM was not able to provide data at a sufficient rate. This indicates a synchronization problem most probably caused by a low clock rate on the
AHB clock. The whole packet is buffered in the transmitter RAM before transmission so underruns
cannot be caused by bus congestion. The Attempt Limit Error bit is set if more collisions occurred
than allowed. When running in 1000 Mbit mode the Late Collision bit indicates that a collision
occurred after the slottime boundary was passed.
The packet was successfully transmitted only if these three bits are zero. The other bits in the first
descriptor word are set to zero after transmission while the second word is left untouched.
The enable bit should be used as the indicator when a descriptor can be used again, which is when it
has been cleared by the GRETH_GBIT. There are three bits in the GRETH_GBIT status register that
GR740-UM-DS, Jun 2016, Version 1.4
250
www.cobham.com/gaisler
GR740
hold transmission status. The Transmit Error (TE) bit is set each time an transmission ended with an
error (when at least one of the three status bits in the transmit descriptor has been set). The Transmit
Successful (TI) is set each time a transmission ended successfully.
The Transmit AHB Error (TA) bit is set when an AHB error was encountered either when reading a
descriptor, reading packet data or writing status to the descriptor. Any active transmissions are aborted
and the transmitter is disabled. The transmitter can be activated again by setting the transmit enable
register.
19.3.4 Setting up the data for transmission
The data to be transmitted should be placed beginning at the address pointed by the descriptor address
field. The GRETH_GBIT does not add the Ethernet address and type fields so they must also be
stored in the data buffer. The 4 B Ethernet CRC is automatically appended at the end of each packet.
Each descriptor will be sent as a single Ethernet packet. If the size field in a descriptor is greater than
1514 B, the packet will not be sent.
19.3.5 Scatter Gather I/O
A packet can be generated from data fetched from several descriptors. This is called Scatter Gather I/
O. The More (MO) bit should be set to 1 to indicate that more descriptors should be used to generate
the current packet. When data from the current descriptor has been read to the RAM the next descriptor is fetched and the new data is appended to the previous data. This continues until a descriptor with
the MO bit set to 0 is encountered. The packet will then be transmitted.
Status is written immediately when data has been read to RAM for descriptors with MO set to 1. The
status bits are always set to 0 since no transmission has occurred. The status bits will be written to the
last descriptor for the packet (which had MO set to 0) when the transmission has finished.
No interrupts are generated for descriptors with MO set to 1 so the IE bit is don’t care in this case.
The checksum offload control bits (explained in section 19.3.6) must be set to the same values for all
descriptors used for a single packet.
19.3.6 Checksum offloading
Support is provided for checksum calculations in hardware for TCP and UDP over IPv4. The checksum calculations are enabled in each descriptor and applies only to that packet (when the MO bit is set
all descriptors used for a single packet must have the checksum control bits set in the same way).
The IP Checksum bit (IC) enables IP header checksum calculations. If an IPv4 packet is detected
when transmitting the packet associated with the descriptor the header checksum is calculated and
inserted. If TCP Checksum (TC) is set the TCP checksum is calculated and inserted if an TCP/IPv4
packet is detected. Finally, if the UDP Checksum bit is set the UDP checksum is calculated and
inserted if a UDP/IPv4 packet is detected. In the case of fragmented IP packets, checksums for TCP
and UDP are only inserted for the first fragment (which contains the TCP or UDP header).
19.4
Rx DMA interface
The receiver DMA interface is used for receiving data from an Ethernet network. The reception is
done using descriptors located in memory.
19.4.1 Setting up descriptors
A single descriptor is shown in table 282 and 283. The address field points at the location where the
received data should be stored. There are no restrictions on alignment. The GRETH_GBIT will never
store more than 1518 B to the buffer (the tagged maximum frame size excluding CRC). The CRC
field (4 B) is never stored to memory so it is not included in this number. If the interrupt enable (IE)
bit is set, an interrupt will be generated when a packet has been received to this buffer (this requires
GR740-UM-DS, Jun 2016, Version 1.4
251
www.cobham.com/gaisler
GR740
that the receiver interrupt bit in the control register is also set). The interrupt will be generated regardless of whether the packet was received successfully or not.
The enable bit is set to indicate that the descriptor is valid which means it can be used by the to store a
packet. After it is set the descriptor should not be touched until the EN bit has been cleared by the
GRETH_GBIT.
The rest of the fields in the descriptor are explained later in this section..
Table 282. Address offset 0x0 - GRETH_GBIT receive descriptor word 0
31
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
RESERVED
0
MC IF TR TD UR UD IR ID LE OE CE FT AE IE WR EN
LENGTH
31: 27
RESERVED
26
Multicast address (MC) - The destination address of the packet was a multicast address (not broadcast).
25
IP fragment (IF) - Fragmented IP packet detected.
24
TCP error (TR) - TCP checksum error detected.
23
TCP detected (TD) - TCP packet detected.
22
UDP error (UR) - UDP checksum error detected.
21
UDP detected (UD) - UDP packet detected.
20
IP error (IR) - IP checksum error detected.
19
IP detected (ID) - IP packet detected.
18
Length error (LE) - The length/type field of the packet did not match the actual number of received
bytes.
17
Overrun error (OE) - The frame was incorrectly received due to a FIFO overrun.
16
CRC error (CE) - A CRC error was detected in this frame.
15
Frame too long (FT) - A frame larger than the maximum size was received. The excessive part
was truncated.
14
Alignment error (AE) - An odd number of nibbles were received.
13
Interrupt Enable (IE) - Enable Interrupts. An interrupt will be generated when a packet has been
received to this descriptor provided that the receiver interrupt enable bit in the control register is set.
The interrupt is generated regardless if the packet was received successfully or if it terminated with
an error.
12
Wrap (WR) - Set to one to make the descriptor pointer wrap to zero after this descriptor has been
used. If this bit is not set the pointer will increment by 8. The pointer automatically wraps to zero
when the 1 kB boundary of the descriptor table is reached.
11
Enable (EN) - Set to one to enable the descriptor. Should always be set last of all the descriptor
fields.
10: 0
LENGTH - The number of bytes received to this descriptor.
Table 283. Address offset 0x4 - GRETH_GBIT receive descriptor word 1
31
0
ADDRESS
31: 0
Address (ADDRESS) - Pointer to the buffer area from where the packet data will be loaded.
19.4.2 Starting reception
Enabling a descriptor is not enough to start reception. A pointer to the memory area holding the
descriptors must first be set in the GRETH_GBIT. This is done in the receiver descriptor pointer register. The address must be aligned to a 1 kB boundary. Bits 31 to 10 hold the base address of descriptor area while bits 9 to 3 form a pointer to an individual descriptor. The first descriptor should be
located at the base address and when it has been used by the GRETH_GBIT the pointer field is incremented by 8 to point at the next descriptor. The pointer will automatically wrap back to zero when the
GR740-UM-DS, Jun 2016, Version 1.4
252
www.cobham.com/gaisler
GR740
next 1 kB boundary has been reached (the descriptor at address offset 0x3F8 has been used). The WR
bit in the descriptors can be set to make the pointer wrap back to zero before the 1 kB boundary.
The pointer field has also been made writable for maximum flexibility but care should be taken when
writing to the descriptor pointer register. It should never be touched when reception is active.
The final step to activate reception is to set the receiver enable bit in the control register. This will
make the GRETH_GBIT read the first descriptor and wait for an incoming packet.
19.4.3 Descriptor handling after reception
The GRETH indicates a completed reception by clearing the descriptor enable bit. The other control
bits (WR, IE) are also cleared. The number of received bytes is shown in the length field. The parts of
the Ethernet frame stored are the destination address, source address, type and data fields. Bits 24-14
in the first descriptor word are status bits indicating different receive errors. Bits 18 - 14 are zero after
a reception without link layer errors. The status bits are described in table 282 (except the checksum
offload bits which are also described in section 19.4.6).
Packets arriving that are smaller than the minimum Ethernet size of 64 B are not considered as a
reception and are discarded. The current receive descriptor will be left untouched an used for the first
packet arriving with an accepted size. The TS bit in the status register is set each time this event
occurs.
If a packet is received with an address not accepted by the MAC, the IA status register bit will be set.
Packets larger than maximum size cause the FT bit in the receive descriptor to be set. The length field
is not guaranteed to hold the correct value of received bytes. The counting stops after the word containing the last byte up to the maximum size limit has been written to memory.
The address word of the descriptor is never touched by the GRETH.
19.4.4 Reception with AHB errors
If an AHB error occurs during a descriptor read or data store, the Receiver AHB Error (RA) bit in the
status register will be set and the receiver is disabled. The current reception is aborted. The receiver
can be enabled again by setting the Receive Enable bit in the control register.
19.4.5 Accepted MAC addresses
In the default configuration the core receives packets with either the unicast address set in the MAC
address register or the broadcast address. Multicast support can also be enabled and in that case a hash
function is used to filter received multicast packets. A 64-bit register, which is accessible through the
APB interface, determines which addresses should be received. Each address is mapped to one of the
64 bits using the hash function and if the bit is set to one the packet will be received. The address is
mapped to the table by taking the 6 least significant bits of the 32-bit Ethernet CRC calculated over
the destination address of the MAC frame. A bit in the receive descriptor is set if a packet with a multicast address has been received to it.
19.4.6 Checksum offload
Support is provided for checksum calculations in hardware for TCP/UDP over IPv4. The checksum
logic is always active and detects IPv4 packets with TCP or UDP payloads. If IPv4 is detected the ID
bit is set, UD is set if an UDP payload is detected in the IP packet and TD is set if a TCP payload is
detected in the IP packet (TD and UD are never set if an IPv4 packet is not detected). When one or
more of these packet types is detected its corresponding checksum is calculated and if an error is
detected the checksum error bit for that packet type is set. The error bits are never set if the corresponding packet type is not detected. The core does not support checksum calculations for TCP and
UDP when the IP packet has been fragmented. This condition is indicated by the IF bit in the receiver
descriptor and when set neither the TCP nor the UDP checksum error indications are valid.
GR740-UM-DS, Jun 2016, Version 1.4
253
www.cobham.com/gaisler
GR740
19.5
MDIO Interface
The MDIO interface provides access to PHY configuration and status registers through a two-wire
interface which is included in the MII interface. The GRETH_GBIT provides full support for the
MDIO interface.
The MDIO interface can be used to access from 1 to 32 PHY containing 1 to 32 16-bit registers. A
read transfer i set up by writing the PHY and register addresses to the MDIO Control register and setting the read bit. This caused the Busy bit to be set and the operation is finished when the Busy bit is
cleared. If the operation was successful the Linkfail bit is zero and the data field contains the read
data. An unsuccessful operation is indicated by the Linkfail bit being set. The data field is undefined
in this case.
A write operation is started by writing the 16-bit data, PHY address and register address to the MDIO
Control register and setting the write bit. The operation is finished when the busy bit is cleared and it
was successful if the Linkfail bit is zero.
19.5.1 PHY interrupts
The core also supports status change interrupts from the PHY. A level sensitive, active low, interrupt
signal can be connected on the eth{0,1}_mdint input. The PHY status change bit in the status register
is set each time an event is detected on this signal. If the PHY status interrupt enable bit is set at the
time of the event the core will also generate an interrupt on the AHB bus.
19.6
Ethernet Debug Communication Link (EDCL)
The EDCL provides access to an on-chip AHB bus through Ethernet. It uses the UDP, IP and ARP
protocols together with a custom application layer protocol. The application layer protocol uses an
ARQ algorithm to provide reliable AHB instruction transfers. Through this link, a read or write transfer can be generated to any address on the AHB bus.
19.6.1 Operation
The EDCL receives packets in parallel with the MAC receive DMA channel. It uses a separate MAC
address which is used for distinguishing EDCL packets from packets destined to the MAC DMA
channel. The EDCL also has an IP address. Since ARP packets use the Ethernet broadcast address, the
IP-address must be used in this case to distinguish between EDCL ARP packets and those that should
go to the DMA-channel. Packets that are determined to be EDCL packets are not processed by the
receive DMA channel.
When the packets are checked to be correct, the AHB operation is performed. The operation is performed with the same AHB master interface that the DMA-engines use. The replies are automatically
sent by the EDCL transmitter when the operation is finished. It shares the Ethernet transmitter with
the transmitter DMA-engine but has higher priority.
19.6.2 EDCL protocols
The EDCL accepts Ethernet frames containing IP or ARP data. ARP is handled according to the protocol specification with no exceptions.
IP packets carry the actual AHB commands. The EDCL expects an Ethernet frame containing IP,
UDP and the EDCL specific application layer parts. Table 284 shows the IP packet required by the
EDCL. The contents of the different protocol headers can be found in TCP/IP literature.
Table 284.The IP packet expected by the EDCL.
Ethernet
IP
UDP
2B
4B
4B
Data 0 - 242
Ethernet
Header
Header
Header
Offset
Control word
Address
4B Words
CRC
GR740-UM-DS, Jun 2016, Version 1.4
254
www.cobham.com/gaisler
GR740
The following is required for successful communication with the EDCL: A correct destination MAC
address, an Ethernet type field containing 0x0806 (ARP) or 0x0800 (IP). The IP-address is then compared for a match. The IP-header checksum and identification fields are not checked. There are a few
restrictions on the IP-header fields. The version must be four and the header size must be 5 B (no
options). The protocol field must always be 0x11 indicating a UDP packet. The length and checksum
are the only IP fields changed for the reply.
The EDCL only provides one service at the moment and it is therefore not required to check the UDP
port number. The reply will have the original source port number in both the source and destination
fields. UDP checksum are not used and the checksum field is set to zero in the replies.
The UDP data field contains the EDCL application protocol fields. Table 285 shows the application
protocol fields (data field excluded) in packets received by the EDCL. The 16-bit offset is used to
align the rest of the application layer data to word boundaries in memory and can thus be set to any
value. The R/W field determines whether a read (0) or a write(1) should be performed. The length
Table 285.The EDCL application layer fields in received frames.
16-bit Offset
14-bit Sequence number
1-bit R/W
10-bit Length
7-bit Unused
field contains the number of bytes to be read or written. If R/W is one the data field shown in Table
284 contains the data to be written. If R/W is zero the data field is empty in the received packets.
Table 286 shows the application layer fields of the replies from the EDCL. The length field is always
zero for replies to write requests. For read requests it contains the number of bytes of data contained in
the data field.
Table 286.The EDCL application layer fields in transmitted frames.
16-bit Offset
14-bit sequence number
1-bit ACK/NAK
10-bit Length
7-bit Unused
The EDCL implements a Go-Back-N algorithm providing reliable transfers. The 14-bit sequence
number in received packets are checked against an internal counter for a match. If they do not match,
no operation is performed and the ACK/NAK field is set to 1 in the reply frame. The reply frame contains the internal counter value in the sequence number field. If the sequence number matches, the
operation is performed, the internal counter is incremented, the internal counter value is stored in the
sequence number field and the ACK/NAK field is set to 0 in the reply. The length field is always set to
0 for ACK/NAK=1 frames. The unused field is not checked and is copied to the reply. It can thus be
set to hold for example some extra id bits if needed.
19.6.3 EDCL IP and Ethernet address settings
The default value of the EDCL IP and MAC addresses are shown in the table below. The addresses
can be changed by software:
Table 287.EDCL addresses
Core
MAC address
IP address
GRETH_GBIT 0
00:50:C2:75:A3:30 to 3F
192.168.0.16 to 31
GRETH_GBIT 1
00:50:C2:75:A3:40 to 4F
192.168.0.32 to 47
In order to allow several EDCL enabled GRETH controllers on the same sub-net without the need for
configuring the cores, the four least significant bits of the IP and MAC addresses are set via general
purpose I/O lines at reset. See the description of bootstrap signals in section 3.1. Note that the four
least significant bits of the IP and MAC addresses also will be reset if the Ethernet controllers are
reset via the clock gating unit.
GR740-UM-DS, Jun 2016, Version 1.4
255
www.cobham.com/gaisler
GR740
19.6.4 EDCL buffer size
The EDCL has a dedicated internal 2 KiB buffer memory which stores the received packets during
processing. Table 288 shows how many concurrent packets the EDCL can handle, the maximum size
of each packet including headers and the maximum size of the data payload. Sending more packets
before receiving a reply than specified for the selected buffer size will lead to dropped packets. The
behavior is unspecified if sending packets exceeding the maximum allowed size.
Table 288.EDCL buffer size limitations
19.7
Total buffer size (KiB)
Number of packet buffers
Packet buffer size (B)
Maximum data payload (B)
2
4
512
456
Media Independent Interfaces
There are several interfaces defined between the MAC sublayer and the Physical layer. The
GRETH_GBIT supports the Media Independent Interface (MII) and the Gigabit Media Independent
Interface (GMII).
The GMII is used in 1000 Mbit mode and the MII in 10 and 100 Mbit. These interfaces are defined
separately in the 802.3-2002 standard but in practice they share most of the signals. The GMII has 9
additional signals compared to the MII. Four data signals are added to the receiver and transmitter
data interfaces respectively and a new transmit clock for the gigabit mode is also introduced.
Table 289.Signals in GMII and MII.
MII and GMII
GMII Only
txd[3:0]
txd[7:4]
tx_en
rxd[7:4]
tx_er
gtx_clk
rx_col
rx_crs
rxd[3:0]
rx_clk
rx_er
rx_dv
GR740-UM-DS, Jun 2016, Version 1.4
256
www.cobham.com/gaisler
GR740
19.8
Registers
The core is programmed through registers mapped into APB address space.
Table 290.GRETH_GBIT registers
APB address offset
Register
0x0
Control register
0x4
Status/Interrupt-source register
0x8
MAC Address MSB
0xC
MAC Address LSB
0x10
MDIO Control/Status
0x14
Transmit descriptor pointer
0x18
Receiver descriptor pointer
0x1C
EDCL IP
0x20
Hash table msb
0x24
Hash table lsb
0x28
EDCL MAC address MSB
0x2C
EDCL MAC address LSB
0x30 - 0xFF
RESERVED
Table 291. 0x0 - GRETH_GBIT control register
31 30
28 27 26 25 24
EA
BS
GA MA MC
1
0b010
1
1
1
r
r
r
r
r
15 14 13 12 11 10
RESERVED
9
8
7
6
5
4
3
ED RD DD ME PI BM GB SP RS PM FD RI
*
0
0
0
0
0
0
0
0
0
0
0
2
1
0
TI RE TE
0
0
0
rw rw rw rw rw rw rw rw wc rw rw rw rw rw rw
31
EDCL available (EA) - Set to one if the EDCL is available.
30: 28
EDCL buffer size (BS) - Shows the amount of memory used for EDCL buffers. 1 = 2 KiB
27
Gigabit MAC available (GA) - This bit always reads as a 1 and indicates that the MAC has 1000
Mbit capability.
26
Mdio interrupts enabled (MA) - Set to one when the core supports mdio interrupts..
25
Multicast available (MC) - Set to one when the core supports multicast address reception.
24: 15
RESERVED
14
EDCL Disable (ED) - Set to one to disablethe EDCL and zero to enable it. Reset value taken from
the external DSU_EN signal. If DSU_EN is high then this bit will be low, and the EDCL will be
enabled after reset. Otherwise the EDCL will be disabled after reset.
13
RAM debug enable (RD) - RAM debug access is not available in this design. This bit is always zero.
Writes have no effect.
12
Disable duplex detection (DD) - Disable the EDCL speed/duplex detection FSM. If the FSM cannot
complete the detection the MDIO interface will be locked in busy mode. If software needs to access
the MDIO the FSM can be disabled here and as soon as the MDIO busy bit is 0 the interface is available. Note that the FSM cannot be re-enabled again.
11
Multicast enable (ME) - Enable reception of multicast addresses.
10
PHY status change interrupt enable (PI) - Enables interrupts for detected PHY status changes.
9
Burstmode (BM) - When set to 1, transmissions use burstmode in 1000 Mbit Half-duplex mode
(GB=1, FD = 0). When 0 in this speed mode normal transmissions are always used with extension
inserted. Operation is undefined when set to 1 in other speed modes.
8
Gigabit (GB) - 1 sets the current speed mode to 1000 Mbit and when set to 0, the speed mode is
selected with bit 7 (SP).
GR740-UM-DS, Jun 2016, Version 1.4
257
www.cobham.com/gaisler
GR740
7
Table 291. 0x0 - GRETH_GBIT control register
Speed (SP) - Sets the current speed mode. 0 = 10 Mbit, 1 = 100 Mbit. Must not be set to 1 at the same
time as bit 8 (GB).
6
Reset (RS) - A one written to this bit resets the GRETH_GBIT core. Self clearing. No other accesses
should be done .to the slave interface other than polling this bit until it is cleared.
5
Promiscuous mode (PM) - If set, the GRETH_GBIT operates in promiscuous mode which means it
will receive all packets regardless of the destination address.
4
Full duplex (FD) - If set, the GRETH_GBIT operates in full-duplex mode otherwise it operates in
half-duplex.
3
Receiver interrupt (RI) - Enable Receiver Interrupts. An interrupt will be generated each time a
packet is received when this bit is set. The interrupt is generated regardless if the packet was
received successfully or if it terminated with an error.
2
Transmitter interrupt (TI) - Enable Transmitter Interrupts. An interrupt will be generated each time a
packet is transmitted when this bit is set. The interrupt is generated regardless if the packet was
transmitted successfully or if it terminated with an error.
1
Receive enable (RE) - Should be written with a one each time new descriptors are enabled. As long
as this bit is one the GRETH_GBIT will read new descriptors and as soon as it encounters a disabled
descriptor it will stop until RE is set again. This bit should be written with a one after the new
descriptors have been enabled.
0
Transmit enable (TE) - Should be written with a one each time new descriptors are enabled. As long
as this bit is one the GRETH_GBIT will read new descriptors and as soon as it encounters a disabled
descriptor it will stop until TE is set again. This bit should be written with a one after the new
descriptors have been enabled.
Table 292. 0x4 - GRETH_GBIT status register.
31
9
RESERVED
8
7
6
5
4
3
PS IA TS TA RA TI
0
0
0
2
1
0
RI TE RE
NR NR NR NR NR NR
wc wc wc wc wc wc wc wc wc
31: 9
RESERVED
8
PHY status changes (PS) - Set each time a PHY status change is detected.
7
Invalid address (IA) - A packet with an address not accepted by the MAC was received. Cleared
when written with a one.
6
Too small (TS) - A packet smaller than the minimum size was received. Cleared when written with a
one.
5
Transmitter AHB error (TA) - An AHB error was encountered in transmitter DMA engine. Cleared
when written with a one. Not Reset.
4
Receiver AHB error (RA) - An AHB error was encountered in receiver DMA engine. Cleared when
written with a one. Not Reset.
3
Transmit successful (TI) - A packet was transmitted without errors. Cleared when written with a one.
Not Reset.
2
Receive successful (RI) - A packet was received without errors. Cleared when written with a one.
Not Reset.
1
Transmitter error (TE) - A packet was transmitted which terminated with an error. Cleared when
written with a one. Not Reset.
0
Receiver error (RE) - A packet has been received which terminated with an error. Cleared when written with a one. Not Reset.
Table 293. 0x8 - GRETH_GBIT MAC address MSB.
31
16 15
RESERVED
0
Bit 47 downto 32 of the MAC Address
NR
GR740-UM-DS, Jun 2016, Version 1.4
258
www.cobham.com/gaisler
GR740
Table 293. 0x8 - GRETH_GBIT MAC address MSB.
rw
31: 16
RESERVED
15: 0
The two most significant bytes of the MAC Address. Not Reset.
Table 294. 0xC - GRETH_GBIT MAC address LSB.
31
0
Bit 31 downto 0 of the MAC Address
NR
rw
31: 0
The 4 least significant bytes of the MAC Address. Not Reset.
Table 295. 0x10 - GRETH_GBIT MDIO control/status register.
31
16 15
DATA
11 10
6
5
4
3
2
1
0
PHYADDR
REGADDR
NV BU LF RD WR
0x0000
*
0b00000
?
0
1
0
rw
rw
rw
?
r
r
rw rw
0
31: 16
Data (DATA) - Contains data read during a read operation and data that is transmitted is taken from
this field. Reset value: 0x0000.
15: 11
PHY address (PHYADDR) - This field contains the address of the PHY that should be accessed
during a write or read operation. Reset value: 
GRETH GBIT 0: “00001”.
GRETH GBIT 1: “00010”
10: 6
Register address (REGADDR) - This field contains the address of the register that should be
accessed during a write or read operation. Reset value: “00000”.
5
RESERVED
4
Not valid (NV) - When an operation is finished (BUSY = 0) this bit indicates whether valid data has
been received that is, the data field contains correct data.
3
Busy (BU) - When an operation is performed this bit is set to one. As soon as the operation is finished and the management link is idle this bit is cleared.
2
Linkfail (LF) - When an operation completes (BUSY = 0) this bit is set if a functional management
link was not detected.
1
Read (RD) - Start a read operation on the management interface. Data is stored in the data field.
0
Write (WR) - Start a write operation on the management interface. Data is taken from the Data field.
Table 296. 0x14 - GRETH_GBIT transmitter descriptor table base address register.
31
10
9
3
BASEADDR
DESCPNT
NR
0b0000000
rw
rw
2
0
RES
31: 10
Transmitter descriptor table base address (BASEADDR) - Base address to the transmitter descriptor
table.Not Reset.
9: 3
Descriptor pointer (DESCPNT) - Pointer to individual descriptors. Automatically incremented by
the Ethernet MAC.
2: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
259
www.cobham.com/gaisler
GR740
Table 297. 0x18 - GRETH_GBIT receiver descriptor table base address register.
31
10
9
3
BASEADDR
DESCPNT
NR
0b0000000
rw
rw
2
0
RES
31: 10
Receiver descriptor table base address (BASEADDR) - Base address to the receiver descriptor
table.Not Reset.
9: 3
Descriptor pointer (DESCPNT) - Pointer to individual descriptors. Automatically incremented by
the Ethernet MAC.
2: 0
RESERVED
Table 298. 0x1C - GRETH_GBIT EDCL IP register
31
0
EDCL IP ADDRESS
*
rw
31: 0
EDCL IP address. Reset value: 
GRETH GBIT 0: 0xC0A80010 (192.168.0.16).
GRETH GBIT 1: 0xC0A80020 (192.168.0.32)
The four lowest bits of the EDCL IP address might be taken from GPIO inputs.
Table 299. 0x20 - GRETH_GBIT Hash table msb register
31
0
Hash table (64:32)
NR
rw
31: 0
Hash table msb. Bits 64 downto 32 of the hash table.
Table 300. 0x24 - GRETH_GBIT Hash table lsb register
31
0
Hash table (64:32)
NR
rw
31: 0
Hash table lsb. Bits 31downto 0 of the hash table.
Table 301. 0x28 - GRETH_GBIT EDCL MAC address MSB.
31
16 15
RESERVED
0
Bit 47 downto 32 of the EDCL MAC Address
0x50C2
rw
31: 16
RESERVED
15: 0
The 2 most significant bytes of the EDCL MAC Address.
GR740-UM-DS, Jun 2016, Version 1.4
260
www.cobham.com/gaisler
GR740
Table 302. 0x2C - GRETH_GBIT EDCL MAC address LSB.
31
0
Bit 31 downto 0 of the EDCL MAC Address
*
rw
31: 0
The 4 least significant bytes of the EDCL MAC Address. Reset value: 
GRETH GBIT 0: 0x0075A330.
GRETH GBIT 1: 0x0075A340
The four lowest bits of the EDCL IP address might be taken from GPIO inputs.
GR740-UM-DS, Jun 2016, Version 1.4
261
www.cobham.com/gaisler
GR740
20
32-bit PCI/AHB bridge
20.1
Overview
The GRPCI2 core is a bridge between the PCI bus and the AMBA AHB bus. The core is capable of
connecting to the PCI bus via both a target and a initiator/master interface. The connection to the
AMBA bus is a AHB master interface for the PCI target functionality and a AHB slave interface for
the PCI initiator functionality. The core also contains a DMA controller. For the DMA functionality,
the core uses the PCI initiator to connect to the PCI bus and an AHB master to connect to the AMBA
bus. Configuration registers in the core are accessible via an AMBA APB slave interface.
The PCI and AMBA interfaces belong to two different clock domains. Synchronization is performed
inside the core through FIFOs.
The PCI interface is compliant with the 2.3 PCI Local Bus Specification.
PCI BUS
32-bit
PCI
Initiator
PCI
Target
DMA
Ctrl
FIFO
FIFO
FIFO
AHB
Master
AHB
Slave
AHB
Master
APB
Slave
AHB BUS
APB BUS
32-bit
32, 64 or 128-bit
Figure 26. Block diagram
20.2
Configuration
The core has configuration registers located both in PCI Configuration Space (Compliant with the 2.3
PCI Local Bus Specification) and via an AMBA APB slave interface (for core function control and
DMA control). This section defines which configuration options that are implemented in the PCI configuration space together with a list of capabilities implemented in the core.
20.2.1 Configuration & Capabilities
The implemented configuration can be determined by reading the Status & Capability register accessible via the APB slave interface. The implementation described by this datasheet has the following
characteristics:
•
The PCI vendor 0x1AC8 and device ID 0x0740
•
The PCI class code 0x0B4000 and revision ID 0x00
•
32-bit PCI initiator interface.
•
32-bit PCI target interface
•
DMA controller
GR740-UM-DS, Jun 2016, Version 1.4
262
www.cobham.com/gaisler
GR740
•
Two FIFOs with a depth of eight words each
•
Two 128 MiB PCI BARs marked as prefetchable. One 8 MiB PCI BAR marked as non-prefetchable. The sizes given here are default sizes. The BAR sizes are configurable (down to a minimum
size of 8 MiB) and the BARs can also be disabled.
•
Device interrupt generation
•
PCI interrupt sampling and forwarding
20.2.2 PCI Configuration Space
The core implements the following registers in the PCI Configuration Space Header. For more
detailed information regarding each field in these registers please refer to the PCI Local Bus Specification.
Table 303.GRPCI2: Implemented registers in the PCI Configuration Space Header
PCI address offset
Register
0x00
Device ID, Vendor ID
0x04
Status, Command
0x08
Class Code, Revision ID
0x0C
BIST, Header Type, Latency Timer, Cache Line Size
0x10 - 0x24
Base Address Registers
0x34
Capabilities Pointer
0x3C
Max_Lat, Min_Gnt, Interrupt Pin, Interrupt Line
Table 304.0x00 - Device ID and Vendor ID register
31
16 16 15
0
Device ID
Vendor ID
0x0061
0x1AC8
r
r
31: 16
Device ID, 0x0740
15: 0
Vendor ID, 0x1AC8
GR740-UM-DS, Jun 2016, Version 1.4
263
www.cobham.com/gaisler
GR740
Table 305.0x04 - Status and Command register
31 30 29 28 27 26 25 24 23 22 21 20 19 18
11 10
RESERVED
8
7
6
5
4
3
2
1
0
S
S
E
R
M
A
R
T
A
S
T
A
DEV
SEL
timing
M
D
P
E
F
B
B
C
R
E
S
0
0
0
0
0
0b01
0
0
0
*
1
0
0
0
0
0
0
0
0
0
0
0
0
0
r
r
r
r
r
r
r
r
rw
r
rw
r
rw
r
rw
r
rw rw
r
wc wc wc wc wc
66 CL IS
M
H
z
9
D
P
E
31
Detected Parity Error (DPE)
30
Signaled System Error (SSE)
29
Received Master Abort (RMA)
28
Received Target Abort (RTA)
27
Signaled Target Abort (STA)
26: 25
DEVSEL timing - Returns “01“ indicating medium
24
Master Data Parity Error (MDPE)
23
Fast Back-to-Back Capable (FBBC) - Returns zero.
22
RESERVED
21
66 MHz Capable (66MHZ)
ID Not SE Not P Not M Not BM MS Not
Imp
Imp E Imp W Imp
Imp
R
I
NOTE: In this implementation this bit has been defined as the status of the PCI_M66EN signal
rather than the capability of the core. For a 33 MHz design, this signal should be connected to
ground and this status bit will have the correct value of ‘0’. For a 66 MHz design, this signal is
pulled-up by the backplane and this status bit will have the correct value of ‘1’. For a 66 MHz capable design inserted in a 33 MHz system, this bit will then indicate a 33 MHz capable device.
20
Capabilities List (CL) - Returns one
19
Interrupt Status (IS)
18: 11
RESERVED
10
Interrupt Disable (DI)
9
NOT IMPLEMENTED, Returns zero.
8
SERR# Enable (SE)
7
NOT IMPLEMENTED, Returns zero.
6
Parity Error Response (PER)
5
NOT IMPLEMENTED, Returns zero.
4
Memory Write and Invalidate Enable (MWI)
3
NOT IMPLEMENTED, Returns zero.
2
Bus Master (BM)
1
Memory Space (MS)
0
NOT IMPLEMENTED, Returns zero.
GR740-UM-DS, Jun 2016, Version 1.4
264
www.cobham.com/gaisler
GR740
Table 306.0x08 - Class Code and Revision ID register
31
8
7
0
Class Code
Revision ID
0x0B4000
0x00
r
r
31: 8
Class Code, 0x0B4000
7: 0
Revision ID, 0x00
Table 307.0x0C - BIST, Header Type, Latency Timer, and Cache Line Size register
31
24 23
16 15
8
0
Header Type
0
0
0
0
r
r
rw
rw
31: 24
Latency Timer
7
BIST
Cache Line Size
BIST - NOT IMPLEMENTED, Returns zeros
23: 16
Header Type - Returns 00
15: 8
Latency Timer - All bits are writable
7: 0
Cache Line Size - NOT IMPLEMENTED, Returns zero
Table 308.0x10-0x24 - Base Address Registers
31
4
31: 4
3
2
1
0
Base Address
PF
Type
MS
0
*
0
0
rw
r
r
r
Base Address - The size of the BAR is determine by how many of the bits (starting from bit 31) are
implemented. Bits not implemented returns zero.
The first two BARs are 128 MiB in size by default. The third BAR is 8 MiB by default. The 8 MiB
BAR is suitable for mapping registers.
3
Prefetchable (PF) - zero indicating non-prefetchable. The two first BARs have the prefetchable bit
set. The third BAR is not prefetchable and is suitable for mapping system registers.
2: 1
Type - Returns zero.
0
Memory Space Indicator (MS) - Returns zero
GR740-UM-DS, Jun 2016, Version 1.4
265
www.cobham.com/gaisler
GR740
Table 309.0x34 - Capabilities Pointer Register
31
8
7
0
RESERVED
Capabilities Pointer
0
0x40
r
r
31: 8
RESERVED
7: 0
Capabilities Pointer - Indicates the first item in the list of capabilities of the Extended PCI Configuration Space. Value: 0x40
Table 310.0x3C - Max_Lat, Min_Gnt, Interrupt Pin and Interrupt Line register
31
24 23
16 15
8
7
0
Max_Lat
Min_Gnt
Interrupt Pin
0
0
INTA
0
r
r
r
rw
31: 24
Interrupt Line
(Max_Lat) - NOT IMPLEMENTED, Returns zero
23: 16
(Min_Gnt) - NOT IMPLEMENTED, Returns zero
15: 8
Interrupt Pin - Indicates INTA
7: 0
Interrupt Line
GR740-UM-DS, Jun 2016, Version 1.4
266
www.cobham.com/gaisler
GR740
20.2.3 Extended PCI Configuration Space
This section describes the first item in the list of capabilities implemented in the Extended PCI Configuration Space. This capability is core specific and contains the PCI to AMBA address mapping and
the option to change endianess of the PCI bus.
Table 311.GRPCI2: Internal capabilities of the Extended PCI Configuration Space
PCI address offset (with the Capabilities
pointer as base)
Register
0x00
Length, Next Pointer, ID
0x04 - 0x18
PCI BAR to AHB address mapping
0x1C
Extended PCI Configuration Space to AHB address mapping
0x20
AHB IO base address and PCI bus config (endianess switch)
0x24 - 0x38
PCI BAR size and prefetch
0x3C
AHB master prefetch burst limit
Table 312.0x00 - Length, Next pointer and ID
31
24 23
16 15
8
7
0
RESERVED
Length
Next Pointer
Capability ID
0
0x40
0x00
0x09
r
r
r
r
31: 24
RESERVED
23: 16
Length - Returns 0x40.
15: 8
Next Pointer - Pointer to the next item in the list of capabilites. Set to 0x00
7: 0
Capability ID - Returns 0x09 indicating Vendor Specific.
Table 313.0x04-0x18 - PCI BAR to AHB address mapping register
31
0
PCI BAR to AHB address mapping
0
rw
31: 0
PCI BAR to AHB address mapping - 32-bit mapping register for each PCI BAR. Translate an access
to a PCI BAR to a AHB base address. The size of the BAR determine how many bits (starting form
bit 31) are implemented. Bits non implemented returns zero
Table 314.0x1C - Extended PCI Configuration Space to AHB address mapping register
31
8
7
0
Extended PCI Configuration Space to AHB address mapping
RESERVED
0
0
rw
r
31: 8
Extended PCI Configuration Space to AHB address mapping - Translates
an access to the Extended PCI Configuration Space (excluding the address range for the internal register located in this configuration
space) to a AHB address.
7: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
267
www.cobham.com/gaisler
GR740
Table 315.0x20 - AHB IO base address and PCI bus config (endianess register)
31
20 19
2
1
0
AHB IO base address
RESERVED
DIS EN
EN DIA
N
*
0
0
r
r
rw rw
0
31: 20
Base address of the AHB IO area
19: 2
RESERVED
1
Target access discard time out enable (DISEN) - When set to ‘1’, the target will discard a pending
access if no retry of the access is detected during 2**15 PCI clock cycles.
0
PCI bus endianess switch (ENDIAN) - 1: defines the PCI bus to be little-endian, 0: defines the PCI
bus to be big-endian
Table 316.0x24-0x38 - PCI BAR size and prefetch register
31
4
PCI BAR size mask
3
2
0
PF
RES
*
*
0
rw
rw
r
31: 4
PCI BAR size mask - A size mask register for each PCI BAR. When bit[n] is set to ‘1’ bit[n] in the
PCI BAR register is implemented and can return a non-zero value. All bits from the lowest bit set to
‘1’ upto bit 31 need to be set to ‘1’. When bit 31 is ‘0’, this PCI BAR is disabled.
3
prefetch bit (PF) - Prefetch bit in PCI BAR register
2: 0
RESERVED
Table 317.0x3C - AHB master prefetch burst limit
31
20.3
16 15
0
RESERVED
Burst length
0
0xFFFF
r
rw
31: 16
RESERVED
15: 0
Burst length - Maximum
number of beats - 1 in the burst. (Maximum value is 0xFFFF => 0x10000
beats => 65 KiB address)
Operation
20.3.1 Access support
The core supports both single and burst accesses on the AMBA AHB bus and on the PCI bus. For
more information on which PCI commands that are supported, see the PCI target section and for burst
limitations see the Burst section.
20.3.2 FIFOs
The core has separate FIFOs for each data path: PCI target read, PCI target write, PCI master read,
PCI master write, DMA AHB-to-PCI, and DMA PCI-to-AHB.
GR740-UM-DS, Jun 2016, Version 1.4
268
www.cobham.com/gaisler
GR740
20.3.3 Byte enables and byte twisting (endianess)
The core has the capability of converting endianess between the two busses. This means that all byte
lanes can be swapped by the core as shown in figure below.
PCI bus
31-24
23-16
15-8
Address 3
7-0
Address 0
GRPCI2
Address 0
AHB bus
Address 3
31-24
23-16
15-8
7-0
Figure 27. GRPCI2 byte twisting
Table 318 defines the supported AHB address/size and PCI byte enable combinations.
Table 318.AHB address/size <=> PCI byte enable combinations.
AHB HSIZE
AHB ADDRESS[1:0]
Little-endian CBE[3:0]
Big-endian CBE[3:0]
00 (8-bit)
00
1110
0111
00 (8-bit)
01
1101
1011
00 (8-bit)
10
1011
1101
00 (8-bit)
11
0111
1110
01 (16-bit)
00
1100
0011
01 (16-bit)
10
0011
1100
10 (32-bit)
00
0000
0000
As the AHB bus in the design is as big-endian, the core is able to define the PCI bus as little-endian
(as defined by the PCI Local Bus Specification) with endianess conversion or define the PCI bus as
big-endian without endianess conversion.
The endianess of the PCI bus is configured via the core specific Extended PCI Configuration Space.
20.3.4 PCI configuration cycles
Accesses to PCI Configuration Space are not altered by the endianess settings. The PCI Configuration
Space is always defined as little-endian (as specified in the PCI Local Bus Specification). This means
GR740-UM-DS, Jun 2016, Version 1.4
269
www.cobham.com/gaisler
GR740
that the PCI target does not change the byte order even if the endianess conversion is enabled and the
PCI master always converts PCI Configuration Space accesses to little-endian.
Data stored in a register in the PCI Configuration Space as 0x12345678 (bit[31:0]) is transferred to
the AHB bus as 0x78563412 (bit[31:0]). This means that non-8-bit accesses to the PCI Configuration
Space must be converted in software to get the correct byte order.
20.3.5 Memory and I/O accesses
Memory and I/O accesses are always affected by the endianess conversion setting. The core should
define the PCI bus as little-endian in the following scenarios: When the core is the PCI host and littleendian peripherals issues DMA transfers to host memory. When the core is a peripheral device and
issues DMA transfers to a little-endian PCI host.
20.3.6 Bursts
PCI bus: The PCI target terminates a burst when no FIFO is available (the AMBA AHB master is not
able to fill or empty the FIFO fast enough) or for reads when the burst reached the length specified by
the “AHB master prefetch burst limit” register. This register defines a boundary which a burst can not
cross i.e. when set to 0x400 beats (address boundary at 4 KiB) the core only prefetches data up to this
boundary and then terminates the burst with a disconnect.
The PCI master stops the burst when the latency timer times out (see the PCI Local Bus Specification
for information on the latency timer) or for reads when the burst reaches the limit defined by “PCI
master prefetch burst limit” register (if AHB master performing the access is unmasked). If the master
is masked in this register, the limit is set to 1 KiB. The PCI master does not prefetch data across this
address boundary.
AHB bus: As long as a FIFOs are available for writes and data in a FIFO is available for read, the
AHB slave does not limit the burst length. The burst length for the AHB master is limited by the FIFO
depth (8 words). The AHB master only bursts up to the FIFO boundary. Only linear-incremental burst
mode is supported.
DMA: DMA accesses are not affected by the “AHB master prefetch burst limit“ register or the “PCI
master prefetch burst limit" register.
All FIFOs are filled starting at the same word offset as the bus access (i.e. with a FIFO of depth 8
words and the start address of a burst is 0x4, the first data word is stored in the second FIFO entry and
only 7 words can be stored in this FIFO).
20.3.7 Host operation
The core provides a system host input (pci_hostn) signal that must be asserted (active low) for PCI
system host operations. The status of this signal is available in the Status & Capability register accessible via the APB slave interface. The device is only allowed to generate PCI configuration cycles
when this signal is asserted (device is the system host).
For designs intended to be host or peripherals only, the PCI system host signal can be tied low (host)
or high (peripheral). For multi-purpose designs it should be connected to a pin on the PCI interface.
The PCI Industrial Computer Manufacturers Group (PCIMG) cPCI specification uses pin C2 on connector P2 for this purpose. The pin should have a pull-up resistor since peripheral slots leave it unconnected.
An asserted PCI system host signal makes the PCI target respond to configuration cycles when no
IDSEL signal is asserted (none of AD[31:11] are asserted). This is done for the PCI master to be able
to configure its own PCI target.
GR740-UM-DS, Jun 2016, Version 1.4
270
www.cobham.com/gaisler
GR740
20.4
PCI Initiator interface
The PCI master interface is accessible via the AMBA AHB slave interface. The AHB slave interface
occupies 1 GiB of the AHB memory address space and 256 KiB of AHB I/O address space. An access
to the AHB memory address area is translated to a PCI memory cycle. An access to the first 64 KiB of
the AHB IO area is translated to a PCI I/O cycle. The next 64 KiB are translated to PCI configuration
cycles. A PCI trace buffer is accessible via the last 128 KiB of the AHB I/O area.
20.4.1 Memory cycles
A single read access to the AHB memory area is translated into a PCI memory read access, while a
burst read translates into a PCI memory read multiple access. A write to this memory area is translated
into a PCI write access.
The address translation is determined by AHB master to PCI address mapping registers accessible via
the APB slave interface. Each AHB master on the AMBA AHB bus has its own mapping register.
These registers contain the MSbs of the PCI address.
When the PCI master is busy performing a transaction on the PCI bus and not able to accept new
requests, the AHB slave interface will respond with an AMBA RETRY response. This occurs on
reads when the PCI master is fetching the requested data to fill the read FIFO or on writes when no
write FIFO is available.
20.4.2 I/O cycles
Accesses to the low 64 KiB of the AHB I/O address area are translated into PCI I/O cycles. The
address translation is determined by the “AHB to PCI mapping register for PCI I/O”. This register sets
the 16 MSb of the PCI address. The “AHB to PCI mapping register for PCI I/O” is accessible via the
APB slave interface. When the “IB” (PCI IO burst) bit in the Control register (accessible via the APB
slave interface) is cleared, the PCI master does not perform burst I/O accesses.
20.4.3 Configuration cycles
Accesses to the second 64 KiB address block (address offset range 64 KiB to 128 KiB) of the AHB I/
O area are translated into PCI configuration cycles. The AHB address is translated into PCI configuration address differently for type 0 and type 1 PCI configuration cycles. When the “bus number” field
in the control register (accessible via the APB slave interface) is zero, type 0 PCI configuration cycles
are issued. When the “bus number“ field is non-zero, type 1 PCI configuration cycles are issued to the
PCI bus determined by this field. The AHB I/O address mapping to PCI configuration address for
type 0 and type 1 PCI configuration cycles is defined in table 319 and table 320.
Only the system host is allowed to generate PCI configuration cycles. The core provides a system host
input signal that must be asserted (active low) for PCI system host operations. The status of this signal
is available in the Status & Capability register accessible via the APB slave interface.When the “CB”
(PCI Configuration burst) bit in the Control register (accessible via the APB slave interface) is
cleared, the PCI master does not perform burst configuration accesses.
Table 319. GRPCI2 Mapping of AHB I/O address to PCI configuration cycle, type 0
31
16 15
AHB ADDRESS MSB
11 10
IDSEL
8
FUNC
7
2
REGISTER
1
0
BYTE
31: 16
AHB address MSbs: Not used for PCI configuration cycle address mapping.
15: 11
IDSEL: This field is decoded to drive PCI AD[IDSEL+10]. Each of the signals AD[31:11] are suppose to be connected (by the PCI back plane) to one corresponding IDSEL line.
10: 8
FUNC: Selects function on a multi-function device.
GR740-UM-DS, Jun 2016, Version 1.4
271
www.cobham.com/gaisler
GR740
7: 2
1: 0
Table 319. GRPCI2 Mapping of AHB I/O address to PCI configuration cycle, type 0
REGISTER: Used to index a PCI DWORD in configuration space.
BYTE: Used to set the CBE correctly for non PCI DWORD accesses.
Table 320. GRPCI2 Mapping of AHB I/O address to PCI configuration cycle, type 1
31
16 15
AHB ADDRESS MSB
11 10
DEVICE
8
7
FUNC
2
REGISTER
31: 16
AHB address MSbs: Not used for PCI configuration cycle address mapping.
15: 11
DEVICE: Selects which device on the bus to access.
10: 8
FUNC: Selects function on a multi-function device.
7: 2
REGISTER: Used to index a PCI DWORD in configuration space.
1: 0
BYTE: Used to set the CBE correctly for non PCI DWORD accesses.
1
0
BYTE
20.4.4 Error handling
When a read access issued by the PCI master is terminated with target-abort or master-abort, the AHB
slave generates an AMBA ERROR response when the “ER” bit in the control register is set. When the
“EI“ bit in the control register is set, an AMBA interrupt is generated for the error. The interrupt status
field in the control register indicates the cause of the error.
20.5
PCI Target interface
The PCI Target occupies memory areas in the PCI address space corresponding to the BAR registers
in the PCI Configuration Space. Each BAR register (BAR0 to BAR2) defines the address allocation in
the PCI address space. The size of each BAR is set by the “BAR size and prefetch” registers accessible via the core specific Extended PCI Configuration Space. The size of a BAR can be determined by
checking the number of implemented bits in the BAR register. Non-implemented bits returns zero and
are read only.
This implementation has three PCI BARs. BAR0 and BAR1 default to prefetchable 128 MiB BARs
and BAR2 defaults to a non-prefetchable 8 MiB BAR.
20.5.1 Supported PCI commands
These are the PCI commands that are supported by the PCI target.
•
PCI Configuration Read/Write: Burst and single access to the PCI Configuration Space. These
accesses are not transferred to the AMBA AHB bus except for the access of the user defined
capability list item in the Extended PCI Configuration Space.
•
Memory Read: A read command to the PCI memory BAR is transferred to a single read access
on the AMBA AHB bus.
•
Memory Read Multiple, Memory Read Line: A read multiple command to the PCI memory
BAR is transferred to a burst access on the AMBA AHB bus. This burst access prefetch data to
fill the maximum amount of data that can be stored in the FIFO.
•
Memory Write, Memory Write and Invalidate: These command are handled similarly and are
transferred to the AMBA AHB bus as a single or burst access depending on the length of the PCI
access (a single or burst access).
20.5.2 Implemented PCI responses
The PCI target can terminate a PCI access with the following responses.
GR740-UM-DS, Jun 2016, Version 1.4
272
www.cobham.com/gaisler
GR740
•
Retry: This response indicates the PCI target is busy by either fetching data for the AMBA AHB
bus on a PCI read or emptying the write FIFO for a PCI write. A new PCI read access will always
be terminated with a retry at least one time before the PCI target is ready to deliver data.
•
Disconnect with data: Terminate the transaction and transfer data in the current data phase. This
occurs when the PCI master request more data and the next FIFO is not yet available or for a PCI
burst access with the Memory Read command.
•
Disconnect without data: Terminate the transaction without transferring data in the current data
phase. This occurs if the CBE change within a PCI burst write.
•
Target Abort: Indicates that the current access caused an internal error and the target is unable to
finish the access. This occurs when the core receives a AMBA AHB error during a read operation.
20.5.3 PCI to AHB translation
Each PCI BAR has translation register (mapping register) to translate the PCI access to an AMBA
AHB address area. These mapping registers are accessible via the core specific Extended PCI Configuration Space. The number of implemented bits in these registers correspond to the size of (and number of implemented bits in) the BARs registers.
20.5.4 PCI system host signal
When the PCI system host signal is asserted the PCI target responds to configuration cycles when no
IDSEL signal is asserted (none of AD[31:11] are asserted). This is done for the PCI master, in a system host position, to be able to configure its own PCI target.
20.5.5 Error handling
The PCI target terminates the access with target-abort when the PCI target requests data from the
AHB bus which results in an error response on the AHB bus. Because writes to the PCI target is
posted, no error is reported on write AHB errors.
When a PCI master is terminated with a retry response it is mandatory for that master to retry the
access until the access is completed or terminated with target-abort. If the master never retries the
access, the PCI target interface would be locked on this access and never accept any new access. To
recover from this situation, the PCI target has a option to discard an access if it is not retried within
215 clock cycles. This discard time out can be enabled via the “AHB IO base address and PCI bus
config” register located in the core specific Extended PCI Configuration Space.
20.6
DMA Controller
The DMA engine is descriptor based and uses two levels of descriptors.
20.6.1 DMA channel
The first level is a linked list of DMA channel descriptors. Each descriptor has a pointer to its data
descriptor list and a pointer to the next DMA channel. The last DMA channel descriptor should
always point to the first DMA channel for the list to be a closed loop. The descriptor needs to be
aligned to 4 words (0x10) in memory and have the following structure:
GR740-UM-DS, Jun 2016, Version 1.4
273
www.cobham.com/gaisler
GR740
Table 321.GRPCI2: DMA channel descriptor structure
Descriptor address offset
Descriptor word
0x00
DMA channel control
0x04
Next DMA channel (32-bit address to next DMA channel descriptor).
0x08
Next data descriptor in this DMA channel (32-bit address to next
data descriptor).
0x0C
RESERVED
Table 322. GRPCI2 DMA channel control
31 30
EN
25 24
RESERVED
31
22 21 20 19
CID
Type
16 15
RESERVED
0
Data descriptor count
Channel descriptor enable.
30: 25
RESERVED
24: 22
Channel ID. Each DMA channel needs a ID to determine the source of a DMA interrupt.
21: 20
Descriptor type. 01 = DMA channel descriptor.
19: 16
RESERVED
15: 0
Maximum number of data descriptors to be executed before moving to the next DMA channel. 0
indicates that all data descriptors should be executed before moving to the next DMA channel.
The number of enabled DMA channels must be stored in the “Number of DMA channels“ field in the
DMA control register accessible via the APB slave interface.
20.6.2 Data descriptor
The second descriptor level is a linked list of data transfers. The last descriptor in this list needs to be
a disabled descriptor. To add a new data transfer, this disabled descriptor is updated to reflect the data
transfer and to point to a new disabled descriptor. The control word in the descriptor should be
updated last to enable the valid descriptor. To make sure the DMA engine reads this new descriptor,
the enable bit in the DMA control register should be updated. The descriptor needs to be aligned to 4
words (0x10) in memory and have the following structure:
Table 323.GRPCI2: DMA data descriptor structure
Descriptor address offset
Descriptor word
0x00
DMA data control
0x04
32-bit PCI start address
0x08
32-bit AHB start address
0x0C
Next data descriptor in this DMA channel (32-bit address to next
data descriptor).
Table 324. GRPCI2 DMA data control
31 30 29 28
EN IE DR BE
31
22 21 20 19 18
RESERVED
Type
16 15
ER RESERVED
0
LEN
Data descriptor enable.
GR740-UM-DS, Jun 2016, Version 1.4
274
www.cobham.com/gaisler
GR740
30
Table 324. GRPCI2 DMA data control
Interrupt generation enable.
29
Transfer direction. 0: PCI to AMBA, 1: AMBA to PCI.
28
PCI bus endianess switch. 1: defines the PCI bus to be little-endian for this transfer, 0: defines the
PCI bus to be big-endian for this transfer.
27: 22
RESERVED (Must be set to zero)
21: 20
Descriptor type. 00 = DMA data descriptor.
19
Error status
18: 16
RESERVED
15: 0
Transfer length. The number of word of the transfer is (this field)+1.
20.6.3 Data transfer
The DMA engine starts by reading the descriptor for the first DMA channel. If the DMA channel is
enabled the first data descriptor in this channel is read and executed. When the transfer is done the
data descriptor is disabled and status is written to the control word. If no error occurred during the
transfer, the error bit is not set and the transfer length field is unchanged. If the transfer was terminated because of an error, the error bit is set in the control word and the length field indicates where in
the transfer the error occurred. If no error has occurred, the next data descriptor is read and executed.
When a disabled data descriptor is read or the maximum number of data descriptors has been executed, the DMA channel descriptor is updated to point to the next data descriptor and the DMA engine
moves on to the next DMA channel.
The DMA engine will stop when an error is detected or when no enabled data descriptors is found.
The error type is indicated by bit 7 to bit 11 in the DMA control register. The error type bits must be
cleared (by writing ‘1’) before the DMA can be re-enabled.
20.6.4 Interrupt
The DMA controller has a interrupt enable bit in the DMA control register (accessible via the APB
slave interface) which enables interrupt generation.
Each data descriptor has an interrupt enable bit which determine if the core should generate a interrupt
when the descriptor has been executed.
The DMA engine asserts the same interrupt as the PCI core.
20.7
PCI trace buffer
20.7.1 Trace data
The data from the trace buffer is accessible in the last 128 KiB block of the core’s AHB I/O area. Each
32-bit word in the first 64 KiB of this block represents a sample of the AD PCI signal. The second 64
KiB of the block is the corresponding PCI control signal. Each 32-bit word is defined in table 325.
Table 325. GRPCI2 PCI control signal trace (32-bit word)
31
20 19
RESERVED
31: 20
16 15 14 13 12 11 10
9
8
7
6
5
4
3
P
A
R
P
E
R
R
S
E
R
R
I
D
S
E
L
R
E
Q
G
N
T
L
O
C
K
R
S
T
CBE[3:0]
F
R
A
M
E
I
R
D
Y
T
R
D
Y
S
T
O
P
D
E
V
S
E
L
0
RES
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
275
www.cobham.com/gaisler
GR740
19: 3
Table 325. GRPCI2 PCI control signal trace (32-bit word)
The state of the PCI control signals.
2: 0
RESERVED
20.7.2 Triggering function
The core can be programmed to trigger on any combination of the PCI AD and PCI Control signals by
setting up the desired pattern and mask in the PCI trace buffer registers accessible via the APB slave
interface. Each bit the PCI AD signal and any PCI control signal can be masked (mask bit equal to
zero) to always match the triggering condition.
The “Trig count” field in the “PCI trace buffer: counter & mode” register defines how many times the
trigger condition should occur before the trace buffer disarms and eventually stops sampling. The
number of samples stored after the triggering condition occurs is defined by the “Delayed stop“ field.
To start sampling, the trace buffer needs to be armed by writing one to the start bit in the “PCI trace
buffer: Control“ register. The state of the trace buffer can be determine by reading the Armed and
Enable/Running bit in the control register. When the Armed bit is set, the triggering condition has not
occurred. The Enable/Running bit indicates that the trace buffer still is storing new samples. When the
delayed stop field is set to a non zero value, the Enabled bit is not cleared until all samples are stored
in the buffer). The trace buffer can also be disarmed by writing the “stop” bit in the “PCI trace buffer:
control” register.
When the trace buffer has been disarmed, the “trig index” in the “PCI trace buffer: control” register is
updated with index of trace entry which match the triggering condition. The address offset of this
entry is the value of the “trig index“ field times 4.
20.7.3 Trace Buffer APB interface
A separate APB register is available on the Debug AHB bus for access of the PCI trace buffer. The
register layout is the same as for the core’s AHB interface but only registers related to the PCI trace
buffer are available. The trace buffer data is located at offset 0x20000 for PCI AD and offset 0x30000
for PCI control signals.
20.8
Interrupts
The core is capable of sampling the PCI INTA-D signals and forwarding the interrupt to the APB bus.
The “host INT mask” field in the control register is used to only sample the valid PCI INT signal(s).
The core is capable of driving the PCI INTA signal. The Interrupt Request Level (IRL) vector for processor 0 is or:ed into a signal that is sampled and forwarded to the PCI INTA signal. The core has a
mask bit (the “Device INT mask” field in the control register) for each bit in the core’s input vector.
The or:ed IRL vector is connected to the first position in this vector. The core also has a PCI interrupt
force bit in the control register to be able to force assertion of PCI INTA.
When the system error PCI signal (SERR) is asserted the core sets the system error bit in the “core
interrupt status” field in the Status & Capability register. If the system interrupt is enabled the core
will also generate a interrupt on the APB bus.
20.9
Reset
The deassertion of the PCI reset is synchronized to the PCI clock and delayed 3 clock cycles. Reset
for the PCI clock domain is generated from the system’s SYS_RESETN input, as described in section
4.3.
GR740-UM-DS, Jun 2016, Version 1.4
276
www.cobham.com/gaisler
GR740
20.10 Registers
The core is configured via registers mapped into APB memory address space.
Table 326.GRPCI2: APB registers
APB address offset
Register
0x00
Control register
0x04
Status & Capability
0x08
PCI master prefetch burst limit
0x0C
AHB to PCI mapping for PCI IO
0x10
DMA Control & Status
0x14
DMA descriptor base
0x18
DMA channel active (read only)
0x1C
RESERVED
0x20 - 0x34
PCI BAR to AHB address mapping (Read only)
0x38
RESERVED
0x3C
RESERVED
0x40 - 0x7C
AHB master to PCI memory address mapping
0x80
PCI trace buffer: control & status
0x84
PCI trace buffer: counter & mode
0x88
PCI trace buffer: AD pattern
0x8C
PCI trace buffer: AD mask
0x90
PCI trace buffer: Ctrl signal pattern
0x94
PCI trace buffer: Ctrl signal mask
0x98
PCI trace buffer: AD state
0x9C
PCI trace buffer: Ctrl signal state
0xA0 - 0xFF
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
277
www.cobham.com/gaisler
GR740
20.10.1 Control register
Table 327.0x00 - CTRL - Control register
31 30 29 28 27 26 25 24 23
RE MR TR R
0
11 10
RESERVED
9
8
7
4
3
0
IB CB DIF Device INT mask Host INT mask
0
0
0
0
0
0
0
0
0
0
r
rw rw rw rw
rw
r
rw rw rw
rw
rw
31
0
16 15
Bus Number
rw rw rw
0
0
SI PE ER EI
0
PCI reset (RE) - When set, the PCI reset signal is asserted. Needs to be cleared to deassert PCI reset.
30
PCI master reset (MR) - Set to reset the cores PCI master. This bit is self clearing.
29
PCI target reset (TR) - Set to reset the cores PCI target. This bit is self clearing.
28
RESERVED
27
System error interrupt (SI) - When set, Interrupt is enabled for System error (SERR)
26
Parity error response (PE) - When set, AHB error response is enabled for Parity error
25
Abort error response (ER) - When set, AHB error response is enabled for Master and Target abort.
24
Error interrupt (EI) - When set, Interrupt is enabled for Master and Target abort and Parity error.
23: 16
Bus Number - When not zero, type 1 configuration cycles is generated.This field is also used as the
Bus Number in type 1 configuration cycles.
15: 11
RESERVED
10
IO burst enable (IB) - When set, burst accesses may be generated by the PCI master for PCI IO
cycles
9
Configuration burst enable (CB) - When set, burst accesses may be generated by the PCI master for
PCI configuration cycles.
8
Device interrupt force (DIF) - When set, a PCI interrupt is forced.
7: 4
Device interrupt mask - When bit[n] is set dirq[n] is unmasked
3: 0
Host interrupt mask -
bit[3] = 1: unmask INTD.
bit[2] = 1: unmask INTC.
bit[1] = 1: unmask INTB.
bit[0] = 1: unmask INTA.
GR740-UM-DS, Jun 2016, Version 1.4
278
www.cobham.com/gaisler
GR740
20.10.2 Status and capability register
Table 328.0x04 - STATCAP - Status and Capability register
31 30 29 28 27 26 25 24 23 22 21 20 19 18
H
o
s
t
M
S
T
T
A
R
D
M
A
DI HI
*
1
1
1
1
r
r
r
r
r
IRQ
mode
T
r
a
c
e
RES
C
F
G
D
O
C
F
G
E
R
1
0
1
0
0
r
r
r
r
r
12 11
8
7
5
4
2
1
0
Core interrupt status
Host interrupt
status
RES
FDEPTH
FNUM
0
0
*
0
3
2
r
wc
r
r
r
r
31
System Host indicator (Host) - When zero, the core is inserted in the System slot and is allowed to
act as System Host.
30
Master implemented (MST)
29
Target implemented (TAR)
28
DMA implemented (DMA)
27
Device Interrupt (DI) - Device drives PCI INTA
26
Host Interrupt (HI) - Device samples PCI INTA..D (for host operations)
25: 24
23
22: 21
APB IRQ mode
00: PCI INTA..D, Error interrupt and DMA interrupt on the same IRQ signal
01: PCI INTA..D and Error interrupt on the same IRQ signal. DMA interrupt on IRQ+1
10: PCI INTA..D on IRQ..IRQ+3. Error interrupt and DMA interrupt on IRQ.
11: PCI INTA..D on IRQ..IRQ+3. Error interrupt on IRQ. DMA interrupt on IRQ+4
PCI trace buffer implemented (Trace)
RESERVED
20
PCI configuration access done (CFGDO) - PCI configuration error status valid.
19
PCI configuration error (CFGER) - Error during PCI configuration access
18: 12
Core Interrupt status:
bit[6]: PCI target access discarded due to time out (access not retried for 215 PCI clock cycles)
bit[5]: System error
bit[4]: DMA interrupt
bit[3]: DMA error
bit[2]: Master abort.
bit[1]: Target abort.
bit[0]: Parity error..
11: 8
Host interrupt status
bit[3] = 0: indicates that INTD is asserted.
bit[2] = 0: indicates that INTC is asserted.
bit[1] = 0: indicates that INTB is asserted.
bit[0] = 0: indicates that INTA is asserted.
7: 5
RESERVED
4: 2
FIFO depth (FDEPTH) - Words in each FIFO = 2(FIFO depth)
1: 0
Number of FIFOs (FNUM)
GR740-UM-DS, Jun 2016, Version 1.4
279
www.cobham.com/gaisler
GR740
20.10.3 PCI master prefetch burst limit register
Table 329.0x08 - BCIM - PCI master prefetch burst limit
31
16 15
8
7
0
AHB master unmask
RESERVED
Burst length
0
0
0xFF
rw
r
r
31: 16
AHB master unmask - When bit[n] is set, the prefetch burst of AHB master n is limited by the “Burst
length” field.
15: 8
RESERVED
7: 0
Burst length - Maximum number of beats - 1 in the burst. (Maximin value is 0xFF => 0x100 beats
=> 1kB address)
20.10.4 AHB to PCI mapping for PCI IO register
Table 330.0x0C - AHB2PCI - AHB to PCI mapping for PCI IO
31
16 15
0
AHB to PCI IO
RESERVED
0
0
rw
r
31: 16
AHB to PCI IO - Used as the MSBs of the base address for a PCI IO access.
15: 0
RESERVED
20.10.5 DMA control and status register
Table 331.0x10 - DMACTRL - DMA control and status register
31 31
20 19
12 11 10
SA
FE
RESERVED
1
0
0
rw
r
wc
31 :
CHIRQ
9
8
7
MA TA PE AE DE
0
0
0
0
6
4
NUMCH
3
2
1
0
AC DIS IE EN
TIV
E
0
0
0
0
wc wc wc wc wc
rw
r
rw rw rw
0
0
Safety guard (SAFE) - Needs to be set to ‘1’ for the control fields to be updated
30: 20
RESERVED
19: 12
Channel IRQ status (CHIRQ) - Set to ‘1’ when a descriptor is configured to signal interrupt. bit[0]
corresponds to the channel with ID 0, bit[1] corresponds to the channel with ID 1, ... Clear by writing
‘1’
11
Master abort (MA) - received master abort during PCI access. Clear by writing ‘1’
10 :
Target abort (TA) - received target abort during PCI access. Clear by writing ‘1’
9
Parity error (PE) - parity error during PCI access. Clear by writing ‘1’
8
:
AHB data error (AE) - Error during AHB data access. Clear by writing ‘1’
7
:
AHB descriptor error (DE) - Error during descriptor access. Clear by writing ‘1’
6: 4
Number of DMA channels (NUMCH) - Guarded by bit 31, safety guard
3
DMA active (ACTIVE)
2
:
1
0
DMA disable/stop (DIS) - Writing ‘1’ to this bit disables the DMA.
Interrupt enable (IE) - (Guarded by bit[31], safety guard).
:
DMA enable/start (EN) - Writing ‘1’ to this bit enables the DMA
GR740-UM-DS, Jun 2016, Version 1.4
280
www.cobham.com/gaisler
GR740
20.10.6 DMA descriptor base address register
Table 332.0x14 - DMABASE - DMA descriptor base address register
31
0
DMA descriptor base address
0
rw
31: 0
Base address of the DMA descriptor table. When running, this register points to the active descriptor.
20.10.7 DMA channel active register
Table 333.0x18 - DMACHAN - DMA channel active register
31
0
DMA channel descriptor base address
0
rw
31: 0
Base address of the active DMA channel
20.10.8 PCI BAR to AHB address mapping register
Table 334.0x20-0x34 - PCI2AHB - PCI BAR to AHB address mapping register
31
0
PCI BAR to AHB address mapping
0
rw
31: 0
32-bit mapping register for each PCI BAR. Translate an access to a PCI BAR to a AHB base
address.
20.10.9 AHB master to PCI memory address mapping register
Table 335.0x40-0x7C - AHBM2PCI - AHB master to PCI memory address mapping register
31
0
AHB master to PCI memory address mapping
0
rw
31: 0
32-bit mapping register for each AHB master. Translate an access from a specific AHB master to a
PCI base address. The size of the AHB slave address area determine how many bits (starting from bit
31) are implemented. Bits not implemented returns zero. The mapping register for AHB master 0 is
located at offset 0x40, AHB master 1 at offset 0x44, and so on up to AHB master 15 at offset 0x7C.
Mapping registers are only implemented for existing AHB masters.
GR740-UM-DS, Jun 2016, Version 1.4
281
www.cobham.com/gaisler
GR740
20.10.10PCI trace control and status register
Table 336.0x80 - TCTRC - PCI trace Control and Status register
31
16 15 14 13 12 11
TRIG INDEX
31: 16
AR EN
4
3
2
RES
DEPTH
RES
1
0
SO SA
N/R
0
0
0
8
0
0
0
r
r
r
r
r
r
w
w
TRIG INDEX - Index of the first entry of the trace.
15
Armed (AR) - Set when trace buffer is armed (started but the trig condition has not occurred).
14
Running (EN) - Set when trace buffer is running
13: 12
RESERVED
11: 4
Trace buffer depth (DEPTH) - Number of buffer entries = 2**DEPTH = 256
3: 2
RESERVED
1
Stop tracing (SO)
0
Start tracing (SA)
20.10.11PCI trace counter and mode register
Table 337.0x84 - TMODE - PCI trace counter and mode register
31
28 27
24 23
16 15
0
RES
Tracing mode
Trig count
Delayed stop
0
0
0
0
r
rw
rw
rw
31: 28
RESERVED
27: 24
Tracing mode
00: Continuous sampling
01: RESERVED
10: RESERVED
11: RESERVED
23: 16
Trig count - The number of times the trig condition should occur before the trace is disarmed.
15: 0
Delayed stop - The number of entries stored after the trace buffer has been disarmed. (Should not be
lager than number of buffer entries - 2).
20.10.12PCI trace AD pattern register
Table 338.0x88 - TADP - PCI trace AD pattern register
31
0
PCI AD pattern
N/R
rw
31: 0
PCI AD pattern to trig on
GR740-UM-DS, Jun 2016, Version 1.4
282
www.cobham.com/gaisler
GR740
20.10.13PCI trace AD mask register
Table 339.0x8C - TADM - PCI trace AD mask register
31
0
PCI AD mask
N/R
rw
31: 0
PCI AD mask - Mask for the AD pattern. When mask bit[n] = 0 pattern bit[n] will always be a
match.
20.10.14PCI trace ctrl signal pattern register
Table 340.0x90 - TCP - PCI trace Ctrl signal pattern register
31
20 19
31: 20
16 15 14 13 12 11 10
9
8
7
6
5
4
3
D
E
V
S
E
L
P
A
R
P
E
R
R
S
E
R
R
I
D
S
E
L
R
E
Q
G
N
T
L
O
C
K
R
S
T
RES
N/ N/
R R
N/
R
N/ N/
R R
N/ N/
R R
N/
R
0
rw rw rw rw rw rw rw rw rw rw rw rw rw
r
RESERVED
CBE[3:0]
F
R
A
M
E
I
R
D
Y
0
N/R
N/
R
N/ N/
R R
r
rw
T
R
D
Y
S
T
O
P
N/ N/
R R
0
RESERVED
19: 3
PCI Ctrl signal pattern to trig on
2: 0
RESERVED
20.10.15PCI trace ctrl signal mask register
Table 341.0x94 - TCM - PCI trace Ctrl signal mask register
31
20 19
16 15 14 13 12 11 10
9
8
7
6
5
4
3
D
E
V
S
E
L
P
A
R
P
E
R
R
S
E
R
R
I
D
S
E
L
R
E
Q
G
N
T
L
O
C
K
R
S
T
RES
N/ N/
R R
N/
R
N/ N/
R R
N/ N/
R R
N/
R
0
rw rw rw rw rw rw rw rw rw rw rw rw rw
r
RESERVED
CBE[3:0]
F
R
A
M
E
I
R
D
Y
0
N/R
N/
R
N/ N/
R R
r
rw
T
R
D
Y
S
T
O
P
N/ N/
R R
31: 20
RESERVED
19: 3
Mask for the Ctrl signal pattern. When mask bit[n] = 0 pattern bit[n] will always be a match.
2: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
283
0
www.cobham.com/gaisler
GR740
20.10.16PCI trace PCI AD state register
Table 342.0x98 - TADS - PCI trace PCI AD state register
31
0
Sampled PCI AD signal
N/R
r
31: 0
The state of the PCI AD signal.
20.10.17PCI trace PCI control signal state register
Table 343.0x9C - TCS - PCI trace PCI Ctrl signal state register
31
20 19
16 15 14 13 12 11 10
RESERVED
CBE[3:0]
F
R
A
M
E
I
R
D
Y
0
N/R
N/
R
N/ N/
R R
r
r
r
31: 20
RESERVED
19: 3
The state of the PCI Ctrl signals
2: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
284
r
T
R
D
Y
r
9
8
7
6
5
4
3
D
E
V
S
E
L
P
A
R
P
E
R
R
S
E
R
R
I
D
S
E
L
R
E
Q
G
N
T
L
O
C
K
R
S
T
RES
N/ N/
R R
N/
R
N/ N/
R R
N/ N/
R R
N/
R
0
r
r
S
T
O
P
r
r
r
r
r
N/ N/
R R
r
r
r
r
0
www.cobham.com/gaisler
GR740
21
MIL-STD-1553B / AS15531 Interface
21.1
Overview
This interface core connects the AMBA AHB/APB bus to a single- or dual redundant MIL-STD1553B bus, and can act as either Bus Controller, Remote Terminal or Bus Monitor.
MIL-STD-1553B (and derived standard SAE AS15531) is a bus standard for transferring data
between up to 32 devices over a shared (typically dual-redundant) differential wire. The bus is
designed for predictable real-time behavior and fault-tolerance. The raw bus data rate is fixed at 1
Mbit/s, giving a maximum of around 770 kbit/s payload data rate.
One of the terminals on the bus is the Bus Controller (BC), which controls all traffic on the bus. The
other terminals are Remote Terminals (RTs), which act on commands issued by the bus controller.
Each RT is assigned a unique address between 0-30. In addition, the bus may have passive Bus Monitors (BM:s) connected.
There are 5 possible data transfer types on the MIL-STD-1553 bus:
•
BC-to-RT transfer (“receive”)
•
RT-to-BC transfer (“transmit”)
•
RT-to-RT transfer
•
Broadcast BC-to-RTs
•
Broadcast RT-to-RTs
Each transfer can contain 1-32 data words of 16 bits each.
The bus controller can also send “mode codes” to the RTs to perform administrative tasks such as time
synchronization, and reading out terminal status.
21.2
Electrical interface
The core is connected to the MIL-STD-1553B bus wire through single or dual transceivers, isolation
transformers and transformer or stub couplers as shown in figure 28. If single-redundancy is used, the
unused bus receive P/N signals should be tied both-high or both-low. The transmit/receive enables
may be inverted on some transceivers. See the standard and the respective component’s data sheets for
more information on the electrical connection
.
Bus A
txenA
txA_P
txA_N
rxA_P
rxA_N
rxenA
GR1553B
Bus B
txenB
txB_P
txB_N
rxB_P
rxB_N
rxenB
Terminal boundary
Figure 28. Interface between core and MIL-STD-1553B bus (dual-redundant, transformer coupled)
GR740-UM-DS, Jun 2016, Version 1.4
285
www.cobham.com/gaisler
GR740
21.3
Operation
21.3.1 Operating modes
The core contains three separate control units for the Bus Controller, Remote Terminal and Bus Monitor handling, with a shared 1553 codec. The operating mode of the core is controlled by starting and
stopping of these units via register writes. At start-up, none of the parts are enabled, and the core is
completely passive on both the 1553 and AMBA bus.
The BC and RT parts of the core can not be active on the 1553 bus at the same time. While the BC is
running or suspended, only the BC (and possibly BM) has access to the 1553 bus, and the RT can only
receive and respond to commands when both the BC schedules are completely stopped (not running
or even suspended).
The Bus Monitor, however, is only listening on the codec receivers and can therefore operate regardless of the enabled/disabled state of the other two parts.
21.3.2 Register interface
The core is configured and controlled through control registers accessed over the APB bus. Each of
the BC, RT, BM parts has a separate set of registers, plus there is a small set of shared registers.
Some of the control register fields for the BC and RT are protected using a ‘key’, a field in the same
register that has to be written with a certain value for the write to take effect. The purpose of the keys
are to give RT/BM designers a way to ensure that the software can not interfere with the bus traffic by
enabling the BC or changing the RT address. If the software is built without knowledge of the key to a
certain register, it is very unlikely that it will accidentally perform a write with the correct key to that
control register.
21.3.3 Interrupting
The core has one interrupt output, which can be generated from several different source events. Which
events should cause an interrupt can be controlled through the IRQ Enable Mask register.
21.3.4 MIL-STD-1553 Codec
The core’s internal codec receives and transmits data words on the 1553 bus, and generates and
checks sync patterns and parity.
Loop-back checking logic checks that each transmitted word is also seen on the receive inputs. If the
transmitted word is not echoed back, the transmitter stops and signals an error condition, which is
then reported back to the user.
GR740-UM-DS, Jun 2016, Version 1.4
286
www.cobham.com/gaisler
GR740
21.4
Bus Controller Operation
21.4.1 Overview
When operating as Bus Controller, the core acts as master on the MIL-STD-1553 bus, initiates and
performs transfers.
This mode works based on a scheduled transfer list concept. The software sets up in memory a
sequence of transfer descriptors and branches, data buffers for sent and received data, and an IRQ
pointer ring buffer. When the schedule is started (through a BC action register write), the core processes the list, performs the transfers one after another and writes resulting status into the transfer list
and incoming data into the corresponding buffers.
21.4.2 Timing control
In each transfer descriptor in the schedule is a “slot time” field. If the scheduled transfer finishes
sooner than its slot time, the core will pause the remaining time before scheduling the next command.
This allows the user to accurately control the message timing during a communication frame.
If the transfer uses more than its slot time, the overshooting time will be subtracted from the following
command’s time slot. The following command may in turn borrow time from the following command
and so on. The core can keep track of up to one second of borrowed time, and will not insert pauses
again until the balance is positive, except for intermessage gaps and pauses that the standard requires.
If you wish to execute the schedule as fast as possible you can set all slot times in the schedule to zero.
If you want to group a number of transfers you can move all the slot time to the last transfer.
The schedule can be stopped or suspended by writing into the BC action register. When suspended,
the schedule’s time will still be accounted, so that the schedule timing will still be correct when the
schedule is resumed. When stopped, on the other hand, the schedule’s timers will be reset.
When the extsync bit is set in the schedule’s next transfer descriptor, the core will wait for a positive
edge on the external sync input before starting the command. The external sync input is connected to
the tick of general purpose timer unit 0’s timer 3. The schedule timer and the time slot balance will
then be reset and the command is started. If the sync pulse arrives before the transfer is reached, it is
stored so the command will begin immediately. The trigger memory is cleared when stopping (but not
when suspending) the schedule. Also, the trigger can be set/cleared by software through the BC action
register.
21.4.3 Bus selection
Each transfer descriptor has a bus selection bit that allows you to control on which one of the two
redundant buses (‘0’ for bus A, ‘1’ for bus B) the transfer will occur.
Another way to control the bus usage is through the per-RT bus swap register, which has one register
bit for each RT address. Writing a ‘1’ to a bit in the register inverts the meaning of the bus selection
bit for all transfers to the corresponding RT, so ‘0’ now means bus ‘B’ and ‘1’ means bus ‘A’. This
allows you to switch all transfers to one or a set of RT:s over to the other bus with a single register
write and without having to modify any descriptors.
The hardware determines which bus to use by taking the exclusive-or of the bus swap register bit and
the bus selection bit. Normally it only makes sense to use one of these two methods for each RT, either
the bus selection bit is always zero and the swap register is used, or the swap register bit is always
zero and the bus selection bit is used.
If the bus swap register is used for bus selection, the store-bus descriptor bit can be enabled to automatically update the register depending on transfer outcome. If the transfer succeeded on bus A, the
bus swap register bit is set to ‘0’, if it succeeds on bus B, the swap register bit is set to ‘1’. If the transfer fails, the bus swap register is set to the opposite value.
GR740-UM-DS, Jun 2016, Version 1.4
287
www.cobham.com/gaisler
GR740
21.4.4 Secondary transfer list
The core can be set up with a secondary “asynchronous” transfer list with the same format as the ordinary schedule. This transfer list can be commanded to start at any time during the ordinary schedule.
While the core is waiting for a scheduled command’s slot time to finish, it will check if the next asynchronous transfer’s slot time is lower than the remaining sleep time. In that case, the asynchronous
command will be scheduled.
If the asynchronous command doesn’t finish in time, time will be borrowed from the next command
in the ordinary schedule. In order to not disturb the ordinary schedule, the slot time for the asynchronous messages must therefore be set to pessimistic values.
The exclusive bit in the transfer descriptor can be set if one does not want an asynchronous command
scheduled during the sleep time following the transfer.
Asynchronous messages will not be scheduled while the schedule is waiting for a sync pulse or the
schedule is suspended and the current slot time has expired, since it is then not known when the next
scheduled command will start.
21.4.5 Interrupt generation
Each command in the transfer schedule can be set to generate an interrupt after certain transfers have
completed, with or without error. Invalid command descriptors always generate interrupts and stop
the schedule. Before a transfer-triggered interrupt is generated, the address to the corresponding
descriptor is written into the BC transfer-triggered IRQ ring buffer and the BC Transfer-triggered IRQ
Ring Position Register is incremented.
A separate error interrupt signals DMA errors. If a DMA error occurs when reading/writing descriptors, the executing schedule will be suspended. DMA errors in data buffers will cause the corresponding transfer to fail with an error code (see table 347).
Whether any of these interrupt events actually cause an interrupt request on the AMBA bus is controlled by the IRQ Mask Register setting.
21.4.6 Transfer list format
The BC:s transfer list is an array of transfer descriptors mixed with branches as shown in table 344.
Each entry has to be aligned to start on a 128-bit (16-byte) boundary. The two unused words in the
branch case are free to be used by software to store arbitrary data.
Table 344.GR1553B transfer descriptor format
Offset
Value for transfer descriptor
DMA R/W Value for branch
0x00
Transfer descriptor word 0 (see table 345)
R
Condition word (see table 349) R
DMA R/W
0x04
Transfer descriptor word 1 (see table 346)
R
Jump address, 128-bit aligned
R
0x08
Data buffer pointer, 16-bit aligned.
R
Unused
-
W
Unused
-
For write buffers, if bit 0 is set the received
data is discarded and the pointer is ignored.
This can be used for RT-to-RT transfers where
the BC is not interested in the data transferred.
0x0C
Result word, written by core (see table 347)
GR740-UM-DS, Jun 2016, Version 1.4
288
www.cobham.com/gaisler
GR740
The transfer descriptor words are structured as shown in tables 345-347 below.
Table 345. GR1553B BC transfer descriptor word 0 (offset 0x00)
31
30
29
28
27
26
25
0
WTRIG
EXCL
IRQE
IRQN
SUSE
SUSN
24
23
22
RETMD
20
NRET
19
18
STBUS
GAP
17
16
RESERVED
15
0
STIME
31
Must be 0 to identify as descriptor
30
Wait for external trigger (WTRIG)
29
Exclusive time slot (EXCL) - Do not schedule asynchronous messages
28
IRQ after transfer on Error (IRQE)
27
IRQ normally (IRQN) - Always interrupts after transfer
26
Suspend on Error (SUSE) - Suspends the schedule (or stops the async transfer list) on error
25
Suspend normally (SUSN) - Always suspends after transfer
24 : 23
Retry mode (RETMD). 00 - Retry on same bus only. 01 - Retry alternating on both buses
10: Retry first on same bus, then on alternating bus. 11 - Reserved, do not use
22 : 20
Number of retries (NRET) - Number of automatic retries per bus
The total number of tries (including the first attempt) is NRET+1 for RETMD=00, 2 x (NRET+1) for RETMD=01/
10
19
Store bus (STBUS) - If the transfer succeeds and this bit is set, store the bus on which the transfer succeeded (0
for bus A, 1 for bus B) into the per-RT bus swap register. If the transfer fails and this bit is set, store the opposite
bus instead. (only if the per-RT bus mask is supported in the core)
See section 21.4.3 for more information.
18
Extended intermessage gap (GAP) - If set, adds an additional amount of gap time, corresponding to the RTTO
field, after the transfer
17 : 16
Reserved - Set to 0 for forward compatibility
15 : 0
Slot time (STIME) - Allocated time in 4 microsecond units, remaining time after transfer will insert delay
Table 346. GR1553B BC transfer descriptor word 1 (offset 0x04)
31
30
DUM
BUS
29
26
RTTO
25
21
RTAD2
20
16
RTSA2
15
11
RTAD1
10
TR
9
5
RTSA1
4
0
WCMC
31
Dummy transfer (DUM) - If set to ‘1’ no bus traffic is generated and transfer “succeeds” immediately
For dummy transfers, the EXCL,IRQN,SUSN,STBUS,GAP,STIME settings are still in effect, other bits and
the data buffer pointer are ignored.
30
Bus selection (BUS) - Bus to use for transfer, 0 - Bus A, 1 - Bus B
29:26
RT Timeout (RTTO) - Extra RT status word timeout above nominal in units of 4 us (0000 -14 us, 1111 -74
us). Note: This extra time is also used as extra intermessage gap time if the GAP bit is set.
25:21
Second RT Address for RT-to-RT transfer (RTAD2)
20:16
Second RT Subaddress for RT-to-RT transfer (RTSA2)
15:11
RT Address (RTAD1)
10
Transmit/receive (TR)
9:5
RT Subaddress (RTSA1)
4:0
Word count/Mode code (WCMC)
GR740-UM-DS, Jun 2016, Version 1.4
See table 348 for details on how to setup
RTAD1,RTSA1,RTAD2,RTSA2,WCMC,TR
for different transfer types.
Note that bits 15:0 correspond to the (first)
command word on the 1553 bus
289
www.cobham.com/gaisler
GR740
Table 347. GR1553B transfer descriptor result word (offset 0x0C)
31
0
30
24
23
Reserved
16
15
8
RT2ST
31
RTST
7
4
RETCNT
3
2
RES
0
TFRST
Always written as 0
30:24
Reserved - Mask away on read for forward compatibility
23:16
RT 2 Status Bits (RT2ST) - Status bits from receiving RT in RT-to-RT transfer, otherwise 0
Same bit pattern as for RTST below
15:8
RT Status Bits (RTST) - Status bits from RT (transmitting RT in RT-to-RT transfer)
15 - Message error, 14 - Instrumentation bit or reserved bit set, 13 - Service request, 
12 - Broadcast command received, 11 - Busy bit, 10 - Subsystem flag, 9 - Dynamic bus control acceptance, 8 - Terminal flag
7:4
Retry count (RETCNT) - Number of retries performed
3
Reserved - Mask away on read for forward compatibility
2:0
Transfer status (TFRST) - Outcome of last try
000 - Success (or dummy bit was set)
001 - RT did not respond (transmitting RT in RT-to-RT transfer)
010 - Receiving RT of RT-to-RT transfer did not respond
011 - A responding RT:s status word had message error, busy, instrumentation or reserved bit set (*)
100 - Protocol error (improperly timed data words, decoder error, wrong number of data words)
101 - The transfer descriptor was invalid
110 - Data buffer DMA timeout or error response
111 - Transfer aborted due to loop back check failure
* Error code 011 is issued only when the number of data words match the success case, otherwise code 100 is used. Error code 011 can be
issued for a correctly executed “transmit last command” or “transmit last status word” mode code since these commands do not reset the status
word.
Table 348.GR1553B BC Transfer configuration bits for different transfer types
RTAD1
(15:11)
RTSA1
(9:5)
RTAD2
(25:21)
RTSA2
(20:16)
WCMC
(4:0)
TR
(10)
Data buffer
direction
Data, BC-to-RT
RT address
(0-30)
RT subaddr
(1-30)
Don’t care
0
Word count
(0 for 32)
0
Read
(2-64 bytes)
Data, RT-to-BC
RT address
(0-30)
RT subaddr
(1-30)
Don’t care
0
Word count
(0 for 32)
1
Write
(2-64 bytes)
Data, RT-to-RT
Recv-RT
addr (0-30)
Recv-RT
subad. (1-30)
Xmit-RT
Xmit-RT
Word count
addr (0-30) subad. (1-30) (0 for 32)
0
Write 
(2-64 bytes)
Mode, no data
RT address
(0-30)
0 or 31 (*)
Don’t care
Don’t care
Mode code
(0-8)
1
Unused
Mode, RT-to-BC
RT address
(0-30)
0 or 31 (*)
Don’t care
Don’t care
Mode code
(16/18/19)
1
Write 
(2 bytes)
Mode, BC-to-RT RT address
(0-30)
0 or 31 (*)
Don’t care
Don’t care
Mode code
(17/20/21)
0
Read 
(2 bytes)
Broadcast
Data, BC-to-RTs
31
RTs subaddr
(1-30)
Don’t care
0
Word count
(0 for 32)
0
Read 
(2-64 bytes)
Broadcast
Data, RT-to-RTs
31
Recv-RTs
subad. (1-30)
Xmit-RT
Xmit-RT
Word count
addr (0-30) subad. (1-30) (0 for 32)
0
Write 
(2-64 bytes)
Broadcast 
Mode, no data
31
0 or 31 (*)
Don’t care
Don’t care
Mode code
(1, 3-8)
1
Unused
Broadcast
31
Mode, BC-to-RT
0 or 31 (*)
Don’t care
Don’t care
Mode code
(17/20/21)
0
Read 
(2 bytes)
Transfer type
(*) The standard allows using either of subaddress 0 or 31 for mode commands.
GR740-UM-DS, Jun 2016, Version 1.4
290
www.cobham.com/gaisler
GR740
The branch condition word is formed as shown in table 349.
Table 349. GR1553B branch condition word (offset 0x00)
31
1
30
27
Reserved (0)
26
25
24
IRQC
ACT
MODE
23
16
RT2CC
15
8
RTCC
7
0
STCC
31
Must be 1 to identify as branch
30 : 27
Reserved - Set to 0
26
Interrupt if condition met (IRQC)
25
Action (ACT) - What to do if condition is met, 0 - Suspend schedule, 1 - Jump
24
Logic mode (MODE):
0 = Or mode (any bit set in RT2CC, RTCC is set in RT2ST,RTST, or result is in STCC mask)
1 - And mode (all bits set in RT2CC,RTCC are set in RT2ST,RTST and result is in STCC mask)
23:16
RT 2 Condition Code (RT2CC) - Mask with bits corresponding to RT2ST in result word of last transfer
15:8
RT Condition Code (RTCC) - Mask with bits corresponding to RTST in result word of last transfer
7:0
Status Condition Code (STCC) - Mask with bits corresponding to status value of last transfer
Note that you can get a constant true condition by setting MODE=0 and STCC=0xFF, and a constant
false condition by setting STCC=0x00. 0x800000FF can thus be used as an end-of-list marker.
GR740-UM-DS, Jun 2016, Version 1.4
291
www.cobham.com/gaisler
GR740
21.5
Remote Terminal Operation
21.5.1 Overview
When operating as Remote Terminal, the core acts as a slave on the MIL-STD-1553B bus. It listens
for requests to its own RT address (or broadcast transfers), checks whether they are configured as
legal and, if legal, performs the corresponding transfer or, if illegal, sets the message error flag in the
status word. Legality is controlled by the subaddress control word for data transfers and by the mode
code control register for mode codes.
To start the RT, set up the subaddress table and log ring buffer, and then write the address and RT
enable bit into the RT Config Register.
21.5.2 Data transfer handling
The Remote Terminal mode uses a three-level structure to handle data transfer DMA. The top level is
a subaddress table, where each subaddress has a subaddress control word, and pointers to a transmit
descriptor and a receive descriptor. Each descriptor in turn contains a descriptor control/status word,
pointer to a data buffer, and a pointer to a next descriptor, forming a linked list or ring of descriptors.
Data buffers can reside anywhere in memory with 16-bit alignment.
When the RT receives a data transfer request, it checks in the subaddress table that the request is legal.
If it is legal, the transfer is then performed with DMA to or from the corresponding data buffer. After
a data transfer, the descriptor’s control/status word is updated with success or failure status and the
subaddress table pointer is changed to point to the next descriptor.
If logging is enabled, a log entry will be written into a log ring buffer area. A transfer-triggered IRQ
may also be enabled. To identify which transfer caused the interrupt, the RT Event Log IRQ Position
points to the corresponding log entry. For that reason, logging must be enabled in order to enable
interrupts.
If a request is legal but can not be fulfilled, either because there is no valid descriptor ready or because
the data cannot be accessed within the required response time, the core will signal a RT table access
error interrupt and not respond to the request. Optionally, the terminal flag status bit can be automatically set on these error conditions.
Descriptor ctrl/stat
SA N-1
Transmit data
Data buffer ptr.
Next pointer
SA ctrl word
SA N
Transmit descr. ptr
Descriptor ctrl/stat
Receive descr. ptr
Data buffer ptr.
Receive buffer
Next pointer
Descriptor ctrl/stat
SA N+1
Data buffer ptr.
Next pointer
Receive buffer
0x3
Subaddress table
Figure 29. RT subaddress data structure example diagram
GR740-UM-DS, Jun 2016, Version 1.4
292
www.cobham.com/gaisler
GR740
21.5.3 Mode Codes
Which of the MIL-STD-1553B mode codes that are legal and should be logged and interrupted are
controlled by the RT Mode Code Control register. As for data transfers, to enable interrupts you must
also enable logging. Inhibit mode codes are controlled by the same fields as their non-inhibit counterpart and mode codes that can be broadcast have two separate fields to control the broadcast and nonbroadcast variants.
The different mode codes and the corresponding action taken by the RT are tabulated below. Some
mode codes do not have a built-in action, so they will need to be implemented in software if desired.
The relation between each mode code to the fields in the RT Mode Code control register is also
shown.
Table 350.RT Mode Codes
Can
log/
IRQ
Enabled
after
reset
Ctrl.
reg
bits
Mode code
Description
Built-in action, if mode code is enabled
0
00000
Dynamic bus control
If the DBCA bit is set in the RT Bus Status register,
a Dynamic Bus Control Acceptance response is
sent.
Yes
No
17:16
1
00001
Synchronize
The time field in the RT sync register is updated.
Yes
Yes
3:0
2
00010
Transmit status word
No
Yes
-
The output rtsync is pulsed high one AMBA cycle.
Transmits the RT:s status word
Enabled always, can not be logged or disabled.
3
00011
Initiate self test
No built-in action
Yes
No
21:18
4
00100
Transmitter shutdown
The RT will stop responding to commands on the
other bus (not the bus on which this command was
given).
Yes
Yes
11:8
5
00101
Override transmitter
shutdown
Removes the effect of an earlier transmitter shutdown mode code received on the same bus
Yes
Yes
11:8
6
00110
Inhibit terminal flag
Masks the terminal flag of the sent RT status words Yes
No
25:22
7
00111
Override inhibit terminal flag
Removes the effect of an earlier inhibit terminal
flag mode code.
Yes
No
25:22
8
01000
Reset remote terminal
The fail-safe timers, transmitter shutdown and
inhibit terminal flag inhibit status are reset.
Yes
No
29:26
The Terminal Flag and Service Request bits in the
RT Bus Status register are cleared.
The extreset output is pulsed high one AMBA
cycle.
16
10000
Transmit vector word
Responds with vector word from RT Status Words
Register
Yes
No
13:12
17
10001
Synchronize with data
word
The time and data fields in the RT sync register are
updated. The rtsync output is pulsed high one
AMBA cycle
Yes
Yes
7:4
18
10010
Transmit last command Transmits the last command sent to the RT.
No
Yes
-
19
10011
Transmit BIT word
Responds with BIT word from RT Status Words
Register
Yes
No
15:14
20
10100
Selected transmitter
shutdown
No built-in action
No
No
-
21
10101
Override selected
transmitter shutdown
No built-in action
No
No
-
Enabled always, can not be logged or disabled.
GR740-UM-DS, Jun 2016, Version 1.4
293
www.cobham.com/gaisler
GR740
21.5.4 Event Log
The event log is a ring of 32-bit entries, each entry having the format given in table 351. Note that for
data transfers, bits 23-0 in the event log are identical to bits 23-0 in the descriptor status word.
Table 351. GR1553B RT Event Log entry format
31
30
IRQSR
29
28
TYPE
24
23
10
SAMC
9
TIMEL
8
3
BC
2
SZ
0
TRES
31
IRQ Source (IRQSRC) - Set to ‘1’ if this transfer caused an interrupt
30 : 29
Transfer type (TYPE) - 00 - Transmit data, 01 - Receive data, 10 - Mode code
28 : 24
Subaddress / Mode code (SAMC) - If TYPE=00/01 this is the transfer subaddress, If TYPE=10, this is the
mode code
23 : 10
TIMEL - Low 14 bits of time tag counter.
9
Broadcast (BC) - Set to 1 if request was to the broadcast address
8:3
Transfer size (SZ) - Count in 16-bit words (0-32)
2:0
Transfer result (TRES)
000 = Success
001 = Superseded (canceled because a new command was given on the other bus)
010 = DMA error or memory timeout occurred
011 = Protocol error (improperly timed data words or decoder error)
100 = The busy bit or message error bit was set in the transmitted status word and no data was sent
101 = Transfer aborted due to loop back checker error
21.5.5 Subaddress table format
Table 352.GR1553B RT Subaddress table entry for subaddress number N, 0<N<31
Offset
Value
DMA R/W
0x10*N + 0x00
Subaddress N control word (table 353)
R
0x10*N + 0x04
Transmit descriptor pointer, 16-byte aligned (0x3 to indicate invalid pointer)
R/W
0x10*N + 0x08
Receive descriptor pointer, 16-byte aligned (0x3 to indicate invalid pointer)
R/W
0x10*N + 0x0C
Unused
-
Note: The table entries for mode code subaddresses 0 and 31 are never accessed by the core.
Table 353. GR1553B RT Subaddress table control word (offset 0x00)
31
19
0 (reserved)
18
WRAP
17
16
IGNDV BCRXE
15
RXEN
14
13
RXLOG RXIRQ
12
8
RXSZ
7
6
5
TXEN
TXLOG
TXIRQ
4
0
TXSZ
31 : 19
Reserved - set to 0 for forward compatibility
18
Auto-wraparound enable (WRAP) - Enables a test mode for this subaddress, where transmit transfers send back the
last received data. This is done by copying the finished transfer’s descriptor pointer to the transmit descriptor pointer
address after each successful transfer.
Note: If WRAP=1, you should not set TXSZ > RXSZ as this might cause reading beyond buffer end
17
Ignore data valid bit (IGNDV) - If this is ‘1’ then receive transfers will proceed (and overwrite the buffer) if the receive
descriptor has the data valid bit set, instead of not responding to the request.
This can be used for descriptor rings where you don’t care if the oldest data is overwritten.
16
Broadcast receive enable (BCRXE) - Allow broadcast receive transfers to this subaddress
15
Receive enable (RXEN) - Allow receive transfers to this subaddress
14
Log receive transfers (RXLOG) - Log all receive transfers in event log ring (only used if RXEN=1)
13
Interrupt on receive transfers (RXIRQ) - Each receive transfer will cause an interrupt (only if also RXEN,RXLOG=1)
12 : 8
Maximum legal receive size (RXSZ) to this subaddress - in16-bit words, 0 means 32
7
Transmit enable (TXEN) - Allow transmit transfers from this subaddress
6
Log transmit transfers (TXLOG) - Log all transmit transfers in event log ring (only if also TXEN=1)
5
Interrupt on transmit transfers (TXIRQ) - Each transmit transfer will cause an interrupt (only if TXEN,TXLOG=1)
4:0
Maximum legal transmit size (TXSZ) from this subaddress - in 16-bit words, 0 means 32
GR740-UM-DS, Jun 2016, Version 1.4
294
www.cobham.com/gaisler
GR740
Table 354.GR1553B RT Descriptor format
Offset
Value
DMA R/W
0x00
Control and status word, see table 355
R/W
0x04
Data buffer pointer, 16-bit aligned
R
0x08
Pointer to next descriptor, 16-byte aligned
R
or 0x0000003 to indicate end of list
Table 355. GR1553B RT Descriptor control/status word (offset 0x00)
31
30
DV
IRQEN
29
26
Reserved (0)
25
10
TTIME
9
8
BC
3
SZ
2
0
TRES
31
Data valid (DV) - Should be set to 0 by software before and set to 1 by hardware after transfer.
If DV=1 in the current receive descriptor before the receive transfer begins then a descriptor table error will
be triggered. You can override this by setting the IGNDV bit in the subaddress table.
30
IRQ Enable override (IRQEN) - Log and IRQ after transfer regardless of SA control word settings
Can be used for getting an interrupt when nearing the end of a descriptor list.
29 : 26
Reserved - Write 0 and mask out on read for forward compatibility
25 : 10
Transmission time tag (TTIME) - Set by the core to the value of the RT timer when the transfer finished.
9
Broadcast (BC) - Set by the core if the transfer was a broadcast transfer
8:3
Transfer size (SZ) - Count in 16-bit words (0-32)
2:0
Transfer result (TRES)
000 = Success
001 = Superseded (canceled because a new command was given on the other bus)
010 = DMA error or memory timeout occurred
011 = Protocol error (improperly timed data words or decoder error)
100 = The busy bit or message error bit was set in the transmitted status word and no data was sent
101 = Transfer aborted due to loop back checker error
GR740-UM-DS, Jun 2016, Version 1.4
295
www.cobham.com/gaisler
GR740
21.6
Bus Monitor Operation
21.6.1 Overview
The Bus Monitor (BM) can be enabled by itself, or in parallel to the BC or RT. The BM acts as a passive logging device, writing received data with time stamps to a ring buffer.
Transfers can be filtered per RT address and per subaddress or mode code, and the filter conditions are
logically AND:ed. If all bits of the three filter registers and bits 2-3 of the control register are set to
’1’, the BM core will log all words that are received on the bus.
In order to filter on subaddress/mode code, the BM has logic to track 1553 words belonging to the
same message. All 10 message types are supported. If an unexpected word appears, the filter logic
will restart. Data words not appearing to belong to any message can be logged by setting a bit in the
control register.
The filter logic can be manually restarted by setting the BM enable bit low and then back to high. This
feature is mainly to improve testability of the BM itself.
21.6.2 No-response handling
Due to the nature of the MIL-STD-1553B protocol, ambiguity can arise when the subaddress or mode
code filters are used, an RT is not responding on a subaddress, and the BC then commands the same
RT again on subaddress 8 or mode code indicator 0 on the same bus. This can lead to the second command word being interpreted as a status word and filtered out.
The BM can use the instrumentation bit and reserved bits to disambiguate, which means that this case
will never occur when subaddresses 1-7, 16-30 and mode code indicator 31 are used. Also, this case
does not occur if only the RT address filter is used.
21.6.3 Log entry format
Each log entry is two 32-bit words.
Table 356. GR1553B BM Log entry word 0 (offset 0x00)
31
30
1
24
23
0
Reserved
TIME
31
Always written as 1
30 : 24
Reserved - Mask out on read for forward compatibility
23 : 0
Time tag (TIME)
Table 357. GR1553B BM Log entry word 1 (offset 0x04)
31
30
0
21.7
20
Reserved
19
BUS
18
17
WST
16
15
WTP
31
Always written as 0
30 : 20
Reserved - Mask out on read for forward compatibility
19
Receive data bus (BUS) - 0:A, 1:B
18 : 17
Word status (WST) - 00=word OK, 01=Manchester error, 10=Parity error
16
Word type (WTP) - 0:Data, 1:Command/status
15 : 0
Word data (WD)
0
WD
Clocking and reset
The core operates in two clock domains, the AMBA clock domain and the 1553 codec clock domain,
with synchronization and handshaking between the domains. For the codec clock domain, a 20 MHz
clock must be supplied. The AMBA clock can be at any frequency but must be at a minimum of 10
GR740-UM-DS, Jun 2016, Version 1.4
296
www.cobham.com/gaisler
GR740
MHz. A propagation delay of up to one codec clock cycle (50 ns) can be tolerated in each clockdomain crossing signal.
Both clock domains are reset via the SYS_RESETN input, as described in section 4.3. Care must be
taken so that the GR1553B_CLK is active and toggling before SYS_RESETN is deasserted.
21.8
Registers
The core is programmed through registers mapped into APB address space.
Table 358.MIL-STD-1553B interface registers
APB address offset
Register
0x00
IRQ Register
0x04
IRQ Enable
0x08...0x0F
(Reserved)
0x10
Hardware config register
0x14...0x3F
(Reserved)
0x40...0x7F
BC Register area (see table 359)
0x80...0xBF
RT Register area (see table 360)
0xC0...0xFF
BM Register area (see table 361)
Table 359.MIL-STD-1553B interface BC-specific registers
APB address offset
Register
0x40
BC Status and Config register
0x44
BC Action register
0x48
BC Transfer list next pointer
0x4C
BC Asynchronous list next pointer
0x50
BC Timer register
0x54
(Reserved)
0x58
BC Transfer-triggered IRQ ring position
0x5C
BC Per-RT bus swap register
0x60...0x67
(Reserved)
0x68
BC Transfer list current slot pointer
0x6C
BC Asynchronous list current slot pointer
0x70...0x7F
(Reserved)
GR740-UM-DS, Jun 2016, Version 1.4
297
www.cobham.com/gaisler
GR740
Table 360.MIL-STD-1553B interface RT-specific registers
APB address offset
Register
0x80
RT Status register
0x84
RT Config register
0x88
RT Bus status bits register
0x8C
RT Status words register
0x90
RT Sync register
0x94
RT Subaddress table base address
0x98
RT Mode code control register
0x9C...0xA3
(Reserved)
0xA4
RT Time tag control register
0xA8
(Reserved)
0xAC
RT Event log size mask
0xB0
RT Event log position
0xB4
RT Event log interrupt position
0xB8.. 0xBF
(Reserved)
Table 361.MIL-STD-1553B interface BM-specific registers
APB address offset
Register
0xC0
BM Status register
0xC4
BM Control register
0xC8
BM RT Address filter register
0xCC
BM RT Subaddress filter register
0xD0
BM RT Mode code filter register
0xD4
BM Log buffer start
0xD8
BM Log buffer end
0xDC
BM Log buffer position
0xE0
BM Time tag control register
0xE4...0xFF
(Reserved)
Table 362. 0x00 - IRQ - GR1553B IRQ Register
31
18
RESERVED
17
16
BMTOF
BMD
RESERVED
15
11
10
9
8
RTTE
RTD
RTEV
RESERVED
7
3
2
1
0
BCWK
BCD
BCEV
0
0
0
0
0
0
0
0
0
0
0
r
wc
wc
r
wc
wc
wc
r
wc
wc
wc
Bits read ‘1’ if interrupt occurred, write back ‘1’ to acknowledge
31: 18
RESERVED
17
BM Timer overflow (BMTOF)
16
BM DMA Error (BMD)
15: 11
RESERVED
10
RT Table access error (RTTE)
9
RT DMA Error (RTD)
8
RT transfer-triggered event interrupt (RTEV)
7: 3
RESERVED
2
BC Wake-up timer interrupt (BCWK)
1
BC DMA Error (BCD)
0
BC Transfer-triggered event interrupt (BCEV)
GR740-UM-DS, Jun 2016, Version 1.4
298
www.cobham.com/gaisler
GR740
Table 363. 0x04 - IRQE - GR1553B IRQ Enable Register
31
18
RESERVED
17
16
15
BMTOE BMDE
11
10
RESERVED
RTTEE
9
8
7
RTDE RTEVE
3
2
RESERVED
1
0
BCWKE BCDE BCEVE
0
0
0
0
0
0
0
0
0
0
0
r
rw
rw
r
rw
rw
rw
r
rw
rw
rw
31: 18
RESERVED
17
BM Timer overflow interrupt enable (BMTOE)
16
BM DMA error interrupt enable (BMDE)
15: 11
RESERVED
10
RT Table access error interrupt enable (RTTEE)
9
RT DMA error interrupt enable (RTDE)
8
RT Transfer-triggered event interrupt enable (RTEVE)
7: 3
RESERVED
2
BC Wake up timer interrupt enable (BCWKE)
1
BC DMA Error Enable (BCDE)
0
BC Transfer-triggered event interrupt enable (BCEVE)
Table 364. 0x10 - HC - GR1553B Hardware Configuration Register
31
30
12
11
10
9
8
7
0
MOD
RESERVED
XKEYS
ENDIAN
SCLK
CCFREQ
0
0
0
0
0
0
r
r
r
r
r
r
31
Modified (MOD) - Reserved to indicate that the core has been modified / customized in an unspecified manner
30: 12
RESERVED
11
Safety Key (XKEYS) - Set if safety keys are enabled for the BM Control Register and for all RT Control Register fields.
10 : 9
AHB Endianness (ENDIAN) - 00=Big-endian, 01=Little-endian, 10/11=Reserved
8
Same clock (SCLK) - Reserved for future versions to indicate that the core has been modified to run with a
single clock
7:0
Codec clock frequency (CCFREQ) - Reserved for future versions of the core to indicate that the core runs at
a different codec clock frequency. Frequency value in MHz, a value of 0 means 20 MHz.
Table 365. 0x40 - BCSC - GR1553B BC Status and Config Register
31
30
28
27
17
BCSUP
BCFEAT
RESERVED
1
0b101
r
r
16
15
11
10
9
8
7
3
2
0
BCCHK
ASADL
-
ASST
SCADL
SCST
0
0
0
0
0
0
0
r
rw
r
r
r
r
r
31
BC Supported (BCSUP) - Reads ‘1’ if core supports BC mode
30: 28
BC Features (BCFEAT) - Bit field describing supported optional features (‘1’=supported):
30
29
28
BC Schedule timer supported
BC Schedule time wake-up interrupt supported
BC per-RT bus swap register and STBUS descriptor bit supported
27: 17
RESERVED
16
Check broadcasts (BCCHK) - Writable bit, if set to ‘1’ enables waiting and checking for (unexpected)
responses to all broadcasts.
15: 11
Asynchronous list address low bits (ASADL) - Bit 8-4 of currently executing (if ASST=01) or next asynchronous command descriptor address
10
RESERVED
9: 8
Asynchronous list state (ASST) - 00=Stopped, 01=Executing command, 10=Waiting for time slot
7: 3
Schedule address low bits (SCADL) - Bit 8-4 of currently executing (if SCST=001) or next schedule descriptor address
2: 0
Schedule state (SCST) - 000=Stopped, 001=Executing command, 010=Waiting for time slot, 011=Suspended, 100=Waiting for external trigger
GR740-UM-DS, Jun 2016, Version 1.4
299
www.cobham.com/gaisler
GR740
Table 366. 0x44 - BCA - GR1553B BC Action Register
31
16
15
10
9
8
7
ASSTP ASSRT
5
RESERVED
4
3
CLRT
SETT
2
1
0
BCKEY
RESERVED
SCSTP SCSUS SCSRT
-
-
-
-
-
-
-
-
-
-
w
-
w
w
-
w
w
w
w
w
31 : 16
Safety code (BCKEY) - Must be 0x1552 when writing, otherwise register write is ignored
15: 10
RESERVED
9
Asynchronous list stop (ASSTP) - Write ‘1’ to stop asynchronous list (after current transfer, if executing)
8
Asynchronous list start (ASSRT) - Write ‘1’ to start asynchronous list
7: 5
RESERVED
4
Clear external trigger (CLRT) - Write ‘1’ to clear trigger memory
3
Set external trigger (SETT) - Write ‘1’ to force the trigger memory to set
2
Schedule stop (SCSTP) - Write ‘1’ to stop schedule (after current transfer, if executing)
1
Schedule suspend (SCSUS) - Write ‘1’ to suspend schedule (after current transfer, if executing)
0
Schedule start (SCSRT) - Write ‘1’ to start schedule
Table 367. 0x48 - BCTNP - GR1553B BC Transfer list next pointer register
31
0
SCHEDULE TRANSFER LIST POINTER
0x00000000
rw
31 : 0
Read: Currently executing (if SCST=001) or next transfer to be executed in regular schedule.
Write: Change address. If running, this will cause a jump after the current transfer has finished.
Table 368. 0x4C - BCANP- GR1553B BC Asynchronous list next pointer register
31
0
ASYNCHRONOUS LIST POINTER
0x00000000
rw
31 :0
Read: Currently executing (if ASST=01) or next transfer to be executed in asynchronous schedule.
Write: Change address. If running, this will cause a jump after the current transfer has finished.
Table 369. 0x50 - BCT - GR1553B BC Timer register
31
24
23
0
RESERVED
SCHEDULE TIME (SCTM)
0x00
0x000000
r
r
31: 24
23: 0
RESERVED
Schedule Time (SCTM) - Elapsed “transfer list” time in microseconds (read-only)
Set to zero when schedule is stopped or on external sync.
Table 370. 0x58 - BCRP - GR1553B BC Transfer-triggered IRQ ring position register
31
0
BC IRQ SOURCE POINTER RING POSITION
0x00000000
rw
31 : 0
The current write pointer into the transfer-tirggered IRQ descriptor pointer ring.
Bits 1:0 are constant zero (4-byte aligned)
The ring wraps at the 64-byte boundary, so bits 31:6 are only changed by user
Table 371. 0x5C - BCBS - GR1553B BC per-RT Bus swap register
31
0
BC PER-RT BUS SWAP
GR740-UM-DS, Jun 2016, Version 1.4
300
www.cobham.com/gaisler
GR740
Table 371. 0x5C - BCBS - GR1553B BC per-RT Bus swap register
0x00000000
rw
31 : 0
The bus selection value will be logically exclusive-or:ed with the bit in this mask corresponding to the
addressed RT (the receiving RT for RT-to-RT transfers). This register gets updated by the core if the STBUS
descriptor bit is used.
For more information on how to use this feature, see section 21.4.3.
Table 372. 0x68 - BCTCP - GR1553B BC Transfer list current slot pointer
31
0
BC TRANSFER SLOT POINTER
0x00000000
r
31 : 0
Points to the transfer descriptor corresponding to the current time slot (read-only, only valid while transfer list
is running).
Bits 3:0 are constant zero (128-bit/16-byte aligned)
Table 373. 0x6C - BCACP - GR1553B BC Asynchronous list current slot pointer
31
0
BC TRANSFER SLOT POINTER
0x00000000
r
31 : 0
Points to the transfer descriptor corresponding to the current asynchronous schedule time slot (read-only,
only valid while asynchronous list is running).
Bits 3:0 are constant zero (128-bit/16-byte aligned)
Table 374. 0x80 - RTS - GR1553B RT Status register
3
2
1
0
RTSUP
31
30
RESERVED
4
ACT
SHDA
SHDB
RUN
1
0
0
0
0
0
r
r
r
r
r
r
31
RT Supported (RTSUP) - Reads ‘1’ if core supports RT mode
30: 4
RESERVED
3
RT Active (ACT) - ‘1’ if RT is currently processing a transfer
2
Bus A shutdown (SHDA) - Reads ‘1’ if bus A has been shut down by the BC (using the transmitter shutdown
mode command on bus B)
1
Bus B shutdown (SHDB) - Reads ‘1’ if bus B has been shut down by the BC (using the transmitter shutdown
mode command on bus A)
0
RT Running (RUN) - ‘1’ if the RT is listening to commands.
GR740-UM-DS, Jun 2016, Version 1.4
301
www.cobham.com/gaisler
GR740
Table 375. 0x84 - RTC - GR1553B RT Config register
31
16
RTKEY
15
14
13
SYS
SYDS
BRS
RESERVED
12
7
RTEIS
6
5
RTADDR
1
0
RTEN
0
1
1
1
0
0
0b11111
0
w
rw
rw
rw
r
r
rw
rw
31 : 16
Safety code (RTKEY) - Must be written as 0x1553 when changing the RT address, otherwise the address
field is unaffected by the write. When reading the register, this field reads 0x0000.
If extra safety keys are enabled (see Hardware Config Register), the lower half of the key is used to also protect the other fields in this register.
15
Sync signal enable (SYS) - Set to ‘1’ to pulse the rtsync output when a synchronize mode code (without
data) has been received
14
Sync with data signal enable (SYDS) - Set to ‘1’ to pulse the rtsync output when a synchronize with data
word mode code has been received
13
Bus reset signal enable (BRS) - Set to ‘1’ to pulse the busreset output when a reset remote terminal mode
code has been received.
12: 7
RESERVED
6
(RTEIS) - Reads ‘1’ if current address was set through external inputs.
After setting the address from software this field is set to ‘0’
5:1
RT Address (RTADDR) - This RT:s address (0-30)
0
RT Enable (RTEN) - Set to ‘1’ to enable listening for requests
Table 376. 0x88 - RTBS - GR1553B RT Bus status register
31
9
RESERVED
8
7
5
4
3
2
1
0
SREQ
BUSY
SSF
DBCA
TFLG
TFDE
RESERVED
0
0
0
0
0
0
0
0
r
rw
r
rw
rw
rw
rw
rw
31: 9
RESERVED
8
Set Terminal flag automatically on DMA and descriptor table errors (TFDE)
7: 5
RESERVED
4:0
These bits will be sent in the RT:s status responses over the 1553 bus.
4
Service request (SREQ)
3
Busy bit (BUSY)
Note: If the busy bit is set, the RT will respond with only the status word and the transfer “fails”
2
Subsystem Flag (SSF)
1
Dynamic Bus Control Acceptance (DBCA)
Note: This bit is only sent in response to the Dynamic Bus Control mode code
0
Terminal Flag (TFLG)
The BC can mask this flag using the “inhibit terminal flag” mode command, if legal
Table 377. 0x8C - RTSW - GR1553B RT Status words register
31
16
15
0
BIT WORD (BITW)
VECTOR WORD (VECW)
0x0000
0x0000
rw
rw
31 : 16
BIT Word (BITW) - Transmitted in response to the “Transmit BIT Word” mode command, if legal
15 : 0
Vector word (VECW) - Transmitted in response to the “Transmit vector word” mode command, if legal.
Table 378. 0x90 - RTSY - GR1553B RT Sync register
31
16
15
0
SYNC TIME (SYTM)
SYNC DATA (SYD)
0x0000
0x0000
r
r
31 : 16
Sync Data (SYD) - The value of the RT timer at the last sync or sync with data word mode command, if legal.
15 : 0
Sync Time (SYTM) - The data received with the last synchronize with data word mode command, if legal
GR740-UM-DS, Jun 2016, Version 1.4
302
www.cobham.com/gaisler
GR740
Table 379. 0x94 - RTSTBA - GR1553B RT Subaddress table base address register
31
9
8
0
SUBADDRESS TABLE BASE (SATB)
-
0x000000
0
rw
r
31 : 9
Base address, bits 31-9 for subaddress table
8:0
Always read ‘0’, writing has no effect
Table 380. 0x98 - RTMCC- GR1553B RT Mode code control register
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
RESERVED
RRTB
RRT
ITFB
ITF
ISTB
IST
DBC
0b00
0b00
0b00
0b00
0b00
0b00
0b00
0b00
r
rw
rw
rw
rw
rw
rw
rw
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
TBW
TVW
TSB
TS
SDB
SD
SB
S
0b00
0b00
0b01
0b01
0b01
0b01
0b01
0b01
rw
rw
rw
rw
rw
rw
rw
rw
For each mode code: “00” - Illegal, “01” - Legal, “10” - Legal, log enabled, “11” - Legal, log and interrupt
31: 30
RESERVED
29 : 28
Reset remote terminal broadcast (RRTB)
27 : 26
Reset remote terminal (RRT)
25 : 24
Inhibit & override inhibit terminal flag bit broadcast (ITFB)
23 : 22
Inhibit & override inhibit terminal flag (ITF)
21 : 20
Initiate self test broadcast (ISTB)
19 : 18
Initiate self test (IST)
17 : 16
Dynamic bus control (DBC)
15 : 14
Transmit BIT word (TBW)
13 : 12
Transmit vector word (TVW)
11 : 10
Transmitter shutdown & override transmitter shutdown broadcast (TSB)
9:8
Transmitter shutdown & override transmitter shutdown (TS)
7:6
Synchronize with data word broadcast (SDB)
5:4
Synchronize with data word (SD)
3:2
Synchronize broadcast (SB)
1:0
Synchronize (S)
Table 381. 0xA4 - RTTTC - GR1553B RT Time tag control register
31
16
15
0
TIME RESOLUTION (TRES)
TIME TAG VALUE (TVAL)
0x0000
0x0000
rw
rw
31 : 16
Time tag resolution (TRES) - Time unit of RT:s time tag counter in microseconds, minus 1
15 : 0
Time tag value (TVAL) - Current value of running time tag counter
Table 382. 0xAC - RTELM - GR1553B RT Event log size mask register
31
18
17
2
-
EVENT LOG SIZE MASK
1
0
0
0xFFFFFFFC
r
31 : 0
rw
0
Mask determining size and alignment of the RT event log ring buffer. All bits “above” the size should be set to
‘1’, all bits below should be set to ‘0’
GR740-UM-DS, Jun 2016, Version 1.4
303
www.cobham.com/gaisler
GR740
Table 383. 0xB0 - RTELP - GR1553B RT Event log position register
31
0
EVENT LOG WRITE POINTER
0x00000000
rw
31 : 0
Address to first unused/oldest entry of event log buffer, 32-bit aligned
Table 384. 0xB4 - RTELIP - GR1553B RT Event Log interrupt position register
31
0
EVENT LOG IRQ POINTER
0x00000000
r
31 : 0
Address to event log entry corresponding to interrupt, 32-bit aligned
The register is set for the first interrupt and not set again until the interrupt has been acknowledged.
Table 385. 0xC0 - BMS - GR1553B BM Status register
31
30
29
0
BMSUP KEYEN
RESERVED
1
0
0
r
r
r
31
BM Supported (BMSUP) - Reads ‘1’ if BM support is in the core.
30
Key Enabled (KEYEN) - Reads ‘1’ if the BM validates the BMKEY field when the control register is written.
29: 0
RESERVED
Table 386. 0xC4- BMC - GR1553B BM Control register
31
16
15
6
5
4
WRSTP EXST
3
2
1
0
IMCL
UDWL
MANL
BMEN
BMKEY
RESERVED
0x0000
0
0
0
0
0
0
0
rw
r
rw
rw
rw
rw
rw
rw
31 : 16
Safety key (BMKEY) - If extra safety keys are enabled (see KEYEN), this field must be 0x1543 for a write to
be accepted. Is 0x0000 when read.
15: 6
RESERVED
5
Wrap stop (WRSTP) - If set to ‘1’, BMEN will be set to ‘0’ and stop the BM when the BM log position wraps
around from buffer end to buffer start
4
External sync start (EXST) - If set to ‘1’,BMEN will be set to ‘1’ and the BM is started when an external BC
sync pulse is received
3
Invalid mode code log (IMCL) - Set to ‘1’ to log invalid or reserved mode codes.
2
Unexpected data word logging (UDWL) - Set to ‘1’ to log data words not seeming to be part of any command
1
Manchester/parity error logging (MANL) - Set to ‘1’ to log bit decoding errors
0
BM Enable (BMEN) - Must be set to ‘1’ to enable any BM logging
Table 387. 0xC8 - BMRTAF - GR1553B BM RT Address filter register
31
0
ADDRESS FILTER MASK
0xFFFFFFFF
rw
31
Enables logging of broadcast transfers
30 : 0
Each bit position set to ‘1’ enables logging of transfers with the corresponding RT address
Table 388. 0xCC - BMRTSF - GR1553B BM RT Subaddress filter register
31
0
SUBADDRESS FILTER MASK
0xFFFFFFFF
rw
GR740-UM-DS, Jun 2016, Version 1.4
304
www.cobham.com/gaisler
GR740
Table 388. 0xCC - BMRTSF - GR1553B BM RT Subaddress filter register
31
Enables logging of mode commands on subaddress 31
30 : 1
Each bit position set to ‘1’ enables logging of transfers with the corresponding RT subaddress
0
Enables logging of mode commands on subaddress 0
Table 389. 0xD0 - BMRTMC - GR1553B BM RT Mode code filter register
31
20
19
RESERVED
18
17
16
STSB
STS
TLC
0x1fff
1
1
1
r
rw
rw
rw
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
TSW
RRTB
RRT
ITFB
ITF
ISTB
IST
DBC
TBW
TVW
TSB
TS
SDB
SD
SB
S
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
Each bit set to ‘1’ enables logging of a mode code:
31: 19
RESERVED
18
Selected transmitter shutdown broadcast & override selected transmitter shutdown broadcast (STSB)
17
Selected transmitter shutdown & override selected transmitter shutdown (STS)
16
Transmit last command (TLC)
15
Transmit status word (TSW)
14
Reset remote terminal broadcast (RRTB)
13
Reset remote terminal (RRT)
12
Inhibit & override inhibit terminal flag bit broadcast (ITFB)
11
Inhibit & override inhibit terminal flag (ITF)
10
Initiate self test broadcast (ISTB)
9
Initiate self test (IST)
8
Dynamic bus control (DBC)
7
Transmit BIT word (TBW)
6
Transmit vector word (TVW)
5
Transmitter shutdown & override transmitter shutdown broadcast (TSB)
4
Transmitter shutdown & override transmitter shutdown (TS)
3
Synchronize with data word broadcast (SDB)
2
Synchronize with data word (SD)
1
Synchronize broadcast (SB)
0
Synchronize (S)
Table 390. 0xD4 -BMLBS - GR1553B BM Log buffer start
31
0
BM LOG BUFFER START
0x00000000
rw
31 : 0
Pointer to the lowest address of the BM log buffer (8-byte aligned)
Due to alignment, bits 2:0 are always 0.
Table 391. 0xD8 - BMLBE - GR1553B BM Log buffer end
31
22
21
-
3
BM LOG BUFFER END
2
0
-
0x00000007
r
31 : 0
rw
r
Pointer to the highest address of the BM log buffer
Only bits 21:3 are settable, i.e. the buffer can not cross a 4 MB boundary Bits 31:22 read the same as the
buffer start address.Due to alignment, bits 2:0 are always equal to 1
GR740-UM-DS, Jun 2016, Version 1.4
305
www.cobham.com/gaisler
GR740
Table 392.0xDC - BMLBP - GR1553B BM Log buffer position
31
22
21
0
-
BM LOG BUFFER POSITION
0x00000000
r
rw
31 : 0
r
Pointer to the next position that will be written to in the BM log buffer
Only bits 21:3 are settable, i.e. the buffer can not cross a 4 MB boundary Bits 31:22 read the same as the
buffer start address.Due to alignment, bits 2:0 are always equal to 0
Table 393.0xE0 - BMTTC - GR1553B BM Time tag control register
31
24
23
0
TIME TAG RESOLUTION
TIME TAG VALUE
0x00
0x000000
rw
rw
31 : 24
Time tag resolution (TRES) - Time unit of BM:s time tag counter in microseconds, minus 1
23 : 0
Time tag value (TVAL) - Current value of running time tag counter
GR740-UM-DS, Jun 2016, Version 1.4
306
www.cobham.com/gaisler
GR740
22
CAN 2.0 Controllers with DMA
22.1
Overview
The CAN controller supports transmission and reception of sets of messages by use of circular buffers
located in memory external to the core. Separate transmit and receive buffers are assumed. Reception
and transmission of sets of messages can be ongoing simultaneously. This device has two CAN controllers, connected to the same two CAN busses.
After a set of message transfers has been set up via the AMBA APB interface the DMA controller initiates a burst of read accesses on the AMBA AHB bus to fetch messages from memory, which are performed by the AHB master. The messages are then transmitted by the CAN core. When a
programmable number of messages have been transmitted, the DMA controller issues an interrupt.
After the reception has been set up via the AMBA APB interface, messages are received by the CAN
core. To store messages to memory, the DMA controller initiates a burst of write accesses on the
AMBA AHB bus, which are performed by the AHB master. When a programmable number of messages have been received, the DMA controller issues an interrupt.
The CAN controller can detect a SYNC message and generate an interrupt. The SYNC message identifier is programmable via the AMBA APB interface. Separate synchronisation message interrupts are
provided.
The CAN controller can transmit and receive messages on either of two CAN busses, but only on one
at a time. The selection is programmable via the AMBA APB interface.
DMA
Controller
FIFO
CAN 2.0
Codec
Physical Layer
AMBA
APB
Slave
Nominal CAN bus
AMBA
AHB
Master
Coding Layer
Mux / DeMux
AMBA Layer
Redundant CAN bus
APB
Master IO bus
Note that it is not possible for the same controller to receive a CAN message while transmitting one.
GRCAN
Figure 30. Block diagram of one CAN controller
The controller implements the following functions:
•
CAN protocol
•
Message transmission, filtering and reception
•
SYNC message reception
•
Status and monitoring
•
Interrupt generation
•
Redundancy selection
GR740-UM-DS, Jun 2016, Version 1.4
307
www.cobham.com/gaisler
GR740
22.2
Interface
The external interface towards the CAN bus features two redundant pairs of transmit output and
receive input (i.e. 0 and 1).
The active pair (i.e. 0 or 1) is selectable by means of a configuration register bit. Note that all reception and transmission is made over the active pair.
For each pair, there is one enable output (i.e. 0 and 1), each being individually programmable. Note
that the enable outputs can be used for enabling an external physical driver. Note that both pairs can
be enabled simultaneously. Note that the polarity for the enable/inhibit inputs on physical interface
drivers differs, thus the meaning of the enable output is undefined.
Redundancy is implemented by means of Selective Bus Access. Note that the active pair selection
above provides means to meet this requirement.
22.3
Protocol
The CAN controller complies with CAN Specification Version 2.0 Part B, except for the overload
frame generation.
Note that there are three different CAN types generally defined:
•
2.0A, which considers 29 bit ID messages as an error
•
2.0B Passive, which ignores 29 bit ID messages
•
2.0B Active, which handles 11 and 29 bit ID messages
Only 2.0B Active is implemented.
22.4
Status and monitoring
The CAN interface incorporates status and monitoring functionalities. This includes:
•
Transmitter active indicator
•
Bus-Off condition indicator
•
Error-Passive condition indicator
•
Over-run indicator
•
8-bit Transmission error counter
•
8-bit Reception error counter
The status is available via a register and is also stored in a circular buffer for each received message.
22.5
Transmission
The transmit channel is defined by the following parameters:
•
base address
•
buffer size
•
write pointer
•
read pointer
The transmit channel can be enabled or disabled.
22.5.1 Circular buffer
The transmit channel operates on a circular buffer located in memory external to the CAN controller.
The circular buffer can also be used as a straight buffer. The buffer memory is accessed via the
AMBA AHB master interface.
GR740-UM-DS, Jun 2016, Version 1.4
308
www.cobham.com/gaisler
GR740
Each CAN message occupies 4 consecutive 32-bit words in memory. Each CAN message is aligned to
4 words address boundaries (i.e. the 4 least significant byte address bits are zero for the first word in a
CAN message).
The size of the buffer is defined by the CanTxSIZE.SIZE field. The number of CAN messages allocated are SIZE * 4.
E.g. CanTxSIZE.SIZE =2 means 8 CAN messages are allocated in the buffer.
Note however that it is not possible to fill the buffer completely, leaving at least one message position
in the buffer empty. This is to simplify wrap-around condition checking.
E.g. CanTxSIZE.SIZE = 2 means that a maximum of 7 CAN messages can be present in the buffer at
any given time.
22.5.2 Write and read pointers
The write pointer (CanTxWR.WRITE) indicates the position+1 of the last CAN message written to
the buffer. The write pointer operates on number of CAN messages, not on absolute or relative
addresses.
The read pointer (CanTxRD.READ) indicates the position+1 of the last CAN message read from the
buffer. The read pointer operates on number of CAN messages, not on absolute or relative addresses.
The difference between the write and the read pointers is the number of CAN messages available in
the buffer for transmission. The difference is calculated using the buffer size, specified by the CanTxSIZE.SIZE field, taking wrap around effects of the circular buffer into account.
Examples:
•
There are 2 CAN messages available
CanTxWR.WRITE=2 and CanTxRD.READ=0.
for
transmit
when
CanTxSIZE.SIZE=2,
•
There are 2 CAN messages available for transmit when CanTxSIZE.SIZE=2, CanTxWR.WRITE
=0 and CanTxRD.READ =6.
•
There are 2 CAN messages available for transmit when CanTxSIZE.SIZE=2, CanTxWR.WRITE
=1 and CanTxRD.READ =7.
•
There are 2 CAN messages available for transmit when CanTxSIZE.SIZE=2, CanTxWR.WRITE
=5 and CanTxRD.READ =3.
When a CAN message has been successfully transmitted, the read pointer (CanTxRD.READ) is automatically incremented, taking wrap around effects of the circular buffer into account. Whenever the
write pointer CanTxWR.WRITE and read pointer CanTxRD.READ are equal, there are no CAN messages available for transmission.
22.5.3 Location
The location of the circular buffer is defined by a base address (CanTxADDR.ADDR), which is an
absolute address. The location of a circular buffer is aligned on a 1 KiB address boundary.
22.5.4 Transmission procedure
When the channel is enabled (CanTxCTRL.ENABLE=1), as soon as there is a difference between the
write and read pointer, a message transmission will be started. Note that the channel should not be
enabled if a potential difference between the write and read pointers could be created, to avoid the
message transmission to start prematurely.
A message transmission will begin with a fetch of the complete CAN message from the circular buffer
to a local fetch-buffer in the CAN controller. After a successful data fetch, a transmission request will
be forwarded to the CAN core. If there is at least one additional CAN message available in the circular buffer, a prefetch of this CAN message from the circular buffer to a local prefetch-buffer in the
GR740-UM-DS, Jun 2016, Version 1.4
309
www.cobham.com/gaisler
GR740
CAN controller will be performed. The CAN controller can thus hold two CAN messages for transmission: one in the fetch buffer, which is fed to the CAN core, and one in the prefetch buffer.
After a message has been successfully transmitted, the prefetch-buffer contents are moved to the fetch
buffer (provided that there is message ready). The read pointer (CanTxRD.READ) is automatically
incremented after a successful transmission, i.e. after the fetch-buffer contents have been transmitted,
taking wrap around effects of the circular buffer into account. If there is at least an additional CAN
message available in the circular buffer, a new prefetch will be performed.
If the write and read pointers are equal, no more prefetches and fetches will be performed, and transmission will stop.
If the single shot mode is enabled for the transmit channel (CanTxCTRL.SINGLE=1), any message
for which the arbitration is lost, or failed for some other reason, will lead to the disabling of the channel (CanTxCTRL.ENABLE=0), and the message will not be put up for re-arbitration.
Interrupts are provided to aid the user during transmission, as described in detail later in this section.
The main interrupts are the Tx, TxEmpty and TxIrq which are issued on the successful transmission
of a message, when all messages have been transmitted successfully and when a predefined number of
messages have been transmitted successfully. The TxLoss interrupt is issued whenever transmission
arbitration has been lost, could also be caused by a communications error. The TxSync interrupt is
issued when a message matching the SYNC Code Filter Register.SYNC and SYNC Mask Filter Register.MASK registers is successfully transmitted. Additional interrupts are provided to signal error
conditions on the CAN bus and AMBA bus.
22.5.5 Straight buffer
It is possible to use the circular buffer as a straight buffer, with a higher granularity than the 1kbyte
address boundary limited by the base address (CanTxADDR.ADDR) field.
While the channel is disabled, the read pointer (CanTxRD.READ) can be changed to an arbitrary
value pointing to the first message to be transmitted, and the write pointer (CanTxWR.WRITE) can
be changed to an arbitrary value.
When the channel is enabled, the transmission will start from the read pointer and continue to the
write pointer.
22.5.6 AMBA AHB error
Definition:
•
a message fetch occurs when no other messages is being transmitted
•
a message prefetch occurs when a previously fetched message is being transmitted
•
the local fetch buffer holds the message being fetched
•
the local prefetch buffer holds the message being prefetched
•
the local fetch buffer holds the message being transmitted by the CAN core
•
a successfully prefetched message is copied from the local prefetch buffer to the local fetch buffer when that buffer is freed after a successful transmission.
An AHB error response occurring on the AMBA AHB bus while a CAN message is being fetched
will result in a TxAHBErr interrupt.
If the CanCONF.ABORT bit is set to 0b, the channel causing the AHB error will skip the message
being fetched from memory and will increment the read pointer. No message will be transmitted.
If the CanCONF.ABORT bit is set to 1b, the channel causing the AHB error will be disabled
(CanTxCTRL.ENABLE is cleared automatically to 0 b). The read pointer can be used to determine
which message caused the AHB error. Note that it could be any of the four word accesses required to
read a message that caused the AHB error.
GR740-UM-DS, Jun 2016, Version 1.4
310
www.cobham.com/gaisler
GR740
If the CanCONF.ABORT bit is set to 1b, all accesses to the AMBA AHB bus will be disabled after an
AMBA AHB error occurs, as indicated by the CanSTAT.AHBErr bit being 1b. The accesses will be
disabled until the CanSTAT register is read, and automatically clearing bit CanSTAT.AHBErr.
An AHB error response occurring on the AMBA AHB bus while a CAN message is being prefetched
will not cause an interrupt, but will stop the ongoing prefetch and further prefetch will be prevented
temporarily. The ongoing transmission of a CAN message from the fetch buffer will not be affected.
When the fetch buffer is freed after a successful transmission, a new fetch will be initiated, and if this
fetch results in an AHB error response occurring on the AMBA AHB bus, this will be handled as for
the case above. If no AHB error occurs, prefetch will be allowed again.
22.5.7 Enable and disable
When an enabled transmit channel is disabled (CanTxCTRL.ENABLE=0b), any ongoing CAN message transfer request will not be aborted until a CAN bus arbitration is lost or the message has been
sent successfully. If the message is sent successfully, the read pointer (CanTxRD.READ) is automatically incremented. Any associated interrupts will be generated.
The progress of the any ongoing access can be observed via the CanTxCTRL.ONGOING bit. The
CanTxCTRL.ONGOING must be 0b before the channel can be re-configured safely (i.e. changing
address, size or read pointer). It is also possible to wait for the Tx and TxLoss interrupts described
hereafter.
The channel can be re-enabled again without the need to re-configure the address, size and pointers.
Priority inversion is handled by disabling the transmitting channel, i.e. setting CanTxCTRL.ENABLE=0b as described above, and observing the progress, i.e. reading via the CanTxCTRL.ONGOING bit as described above. When the transmit channel is disabled, it can be reconfigured and a higher priority message can be transmitted. Note that the single shot mode does not
require the channel to be disabled, but the progress should still be observed as above.
No message transmission is started while the channel is not enabled.
22.5.8 Interrupts
During transmission several interrupts can be generated:
•
TxLoss:
Message arbitration lost for transmit (could be caused by
communications error, as indicated by other interrupts as well)
•
TxErrCntr: Error counter incremented for transmit
•
TxSync:
Synchronization message transmitted
•
Tx:
Successful transmission of one message
•
TxEmpty:
Successful transmission of all messages in buffer
•
TxIrq:
Successful transmission of a predefined number of messages
•
TxAHBErr: AHB access error during transmission
•
Off:
Bus-off condition
•
Pass:
Error-passive condition
The Tx, TxEmpty and TxIrq interrupts are only generated as the result of a successful message transmission, after the CanTxRD.READ pointer has been incremented.
22.6
Reception
The receive channel is defined by the following parameters:
•
base address
•
buffer size
GR740-UM-DS, Jun 2016, Version 1.4
311
www.cobham.com/gaisler
GR740
•
write pointer
•
read pointer
The receive channel can be enabled or disabled.
22.6.1 Circular buffer
The receive channel operates on a circular buffer located in memory external to the CAN controller.
The circular buffer can also be used as a straight buffer. The buffer memory is accessed via the
AMBA AHB master interface.
Each CAN message occupies 4 consecutive 32-bit words in memory. Each CAN message is aligned to
4 words address boundaries (i.e. the 4 least significant byte address bits are zero for the first word in a
CAN message).
The size of the buffer is defined by the CanRxSIZE.SIZE field, specifying the number of CAN messages * 4 that fit in the buffer.
E.g. CanRxSIZE.SIZE=2 means 8 CAN messages fit in the buffer.
Note however that it is not possible to fill the buffer completely, leaving at least one message position
in the buffer empty. This is to simplify wrap-around condition checking.
E.g. CanRxSIZE.SIZE=2 means that 7 CAN messages fit in the buffer at any given time.
22.6.2 Write and read pointers
The write pointer (CanRxWR.WRITE) indicates the position+1 of the last CAN message written to
the buffer. The write pointer operates on number of CAN messages, not on absolute or relative
addresses.
The read pointer (CanRxRD.READ) indicates the position+1 of the last CAN message read from the
buffer. The read pointer operates on number of CAN messages, not on absolute or relative addresses.
The difference between the write and the read pointers is the number of CAN message positions available in the buffer for reception. The difference is calculated using the buffer size, specified by the
CanRxSIZE.SIZE field, taking wrap around effects of the circular buffer into account.
Examples:
•
There are 2 CAN messages available for read-out when CanRxSIZE.SIZE=2, CanRxWR.WRITE=2 and CanRxRD.READ=0.
•
There are 2 CAN messages available for read-out when CanRxSIZE.SIZE=2, CanRxWR.WRITE =0 and CanRxRD.READ=6.
•
There are 2 CAN messages available for read-out when CanRxSIZE.SIZE=2, CanRxWR.WRITE =1 and CanRxRD.READ=7.
•
There are 2 CAN messages available for read-out when CanRxSIZE.SIZE=2, CanRxWR.WRITE =5 and CanRxRD.READ=3.
When a CAN message has been successfully received and stored, the write pointer (CanRxWR.WRITE) is automatically incremented, taking wrap around effects of the circular buffer into
account. Whenever the read pointer CanRxRD.READ equals (CanRxWR.WRITE+1) modulo (CanRxSIZE.SIZE*4), there is no space available for receiving another CAN message.
The error behavior of the CAN core is according to the CAN standard, which applies to the error
counter, bus-off condition and error-passive condition.
22.6.3 Location
The location of the circular buffer is defined by a base address (CanRxADDR.ADDR), which is an
absolute address. The location of a circular buffer is aligned on a 1 KiB address boundary.
GR740-UM-DS, Jun 2016, Version 1.4
312
www.cobham.com/gaisler
GR740
22.6.4 Reception procedure
When the channel is enabled (CanRxCTRL.ENABLE=1), and there is space available for a message
in the circular buffer (as defined by the write and read pointer), as soon as a message is received by
the CAN core, an AMBA AHB store access will be started. The received message will be temporarily
stored in a local store-buffer in the CAN controller. Note that the channel should not be enabled until
the write and read pointers are configured, to avoid the message reception to start prematurely
After a message has been successfully stored the CAN controller is ready to receive a new message.
The write pointer (CanRxWR.WRITE) is automatically incremented, taking wrap around effects of
the circular buffer into account.
Interrupts are provided to aid the user during reception, as described in detail later in this section. The
main interrupts are the Rx, RxFull and RxIrq which are issued on the successful reception of a message, when the message buffer has been successfully filled and when a predefined number of messages have been received successfully. The RxMiss interrupt is issued whenever a message has been
received but does not match a message filtering setting, i.e. neither for the receive channel nor for the
SYNC message described hereafter.
The RxSync interrupt is issued when a message matching the SYNC Code Filter Register.SYNC and
SYNC Mask Filter Register.MASK registers has been successfully received. Additional interrupts are
provided to signal error conditions on the CAN bus and AMBA bus.
22.6.5 Straight buffer
It is possible to use the circular buffer as a straight buffer, with a higher granularity than the 1kbyte
address boundary limited by the base address (CanRxADDR.ADDR) field.
While the channel is disabled, the write pointer (CanRxWR.WRITE) can be changed to an arbitrary
value pointing to the first message to be received, and the read pointer (CanRxRD.READ) can be
changed to an arbitrary value.
When the channel is enabled, the reception will start from the write pointer and continue to the read
pointer.
22.6.6 AMBA AHB error
An AHB error response occurring on the AMBA AHB bus while a CAN message is being stored will
result in an RxAHBErr interrupt.
If the CanCONF.ABORT bit is set to 0b, the channel causing the AHB error will skip the received
message, not storing it to memory. The write pointer will be incremented.
If the CanCONF.ABORT bit is set to 1b, the channel causing the AHB error will be disabled (CanRxCTRL.ENABLE is cleared automatically to 0b). The write pointer can be used to determine which
message caused the AHB error. Note that it could be any of the four word accesses required to writ a
message that caused the AHB error.
If the CanCONF.ABORT bit is set to 1b, all accesses to the AMBA AHB bus will be disabled after an
AMBA AHB error occurs, as indicated by the CanSTAT.AHBErr bit being 1b. The accesses will be
disabled until the CanSTAT register is read, and automatically clearing bit CanSTAT.AHBErr.
22.6.7 Enable and disable
When an enabled receive channel is disabled (CanRxCTRL.ENABLE=0b), any ongoing CAN message storage on the AHB bus will not be aborted, and no new message storage will be started. Note
that only complete messages can be received from the CAN core. If the message is stored successfully, the write pointer (CanRxWR.WRITE) is automatically incremented. Any associated interrupts
will be generated.
GR740-UM-DS, Jun 2016, Version 1.4
313
www.cobham.com/gaisler
GR740
The progress of the any ongoing access can be observed via the CanRxCTRL.ONGOING bit. The
CanRxCTRL.ONGOING must be 0b before the channel can be re-configured safely (i.e. changing
address, size or write pointer). It is also possible to wait for the Rx and RxMiss interrupts described
hereafter.
The channel can be re-enabled again without the need to re-configure the address, size and pointers.
No message reception is performed while the channel is not enabled
22.6.8 Interrupts
During reception several interrupts can be generated:
•
RxMiss:
Message filtered away for receive
•
RxErrCntr: Error counter incremented for receive
•
RxSync:
Synchronization message received
•
Rx:
Successful reception of one message
•
RxFull:
Successful reception of all messages possible to store in buffer
•
RxIrq:
Successful reception of a predefined number of messages
•
RxAHBErr: AHB access error during reception
•
OR:
Over-run during reception
•
OFF:
Bus-off condition
•
PASS:
Error-passive condition
The Rx, RxFull and RxIrq interrupts are only generated as the result of a successful message reception, after the CanRxWR.WRITE pointer has been incremented.
The OR interrupt is generated when a message is received while a previously received message is still
being stored. A full circular buffer will lead to OR interrupts for any subsequently received messages.
Note that the last message stored which fills the circular buffer will not generate an OR interrupt. The
overrun is also reported with the CanSTAT.OR bit, which is cleared when reading the register.
The error behavior of the CAN core is according to the CAN standard, which applies to the error
counter, buss-off condition and error-passive condition.
22.7
Global reset and enable
When the CanCTRL.RESET bit is set to 1b, a reset of the core is performed. The reset clears all the
register fields to their default values. Any ongoing CAN message transfer request will be aborted,
potentially violating the CAN protocol.
When the CanCTRL.ENABLE bit is cleared to 0b, the CAN core is reset and the configuration bits
CanCONF.SCALER, CanCONF.PS1, CanCONF.PS2, CanCONF.RSJ and CanCONF.BPR may be
modified. When disabled, the CAN controller will be in sleep mode not affecting the CAN bus by
only sending recessive bits. Note that the CAN core requires that 10 recessive bits are received before
any reception or transmission can be initiated. This can be caused either by no unit sending on the
CAN bus, or by random bits in message transfers.
22.8
Interrupt
Three interrupts are implemented by the CAN interface:
Index:Name:Description:
16
IRQ Common output from interrupt handler
17
TxSYNCSynchronization message transmitted (optional)
GR740-UM-DS, Jun 2016, Version 1.4
314
www.cobham.com/gaisler
GR740
18
RxSYNCSynchronization message received (optional)
The interrupt lines are shared with the general purpose I/O ports, see interrupt assignments in section
2.4.
22.9
Registers
The core is programmed through registers mapped into APB address space.
Table 394.GRCAN registers
APB address offset
Register
0x000
Configuration Register
0x004
Status Register
0x008
Control Register
0x018
SYNC Mask Filter Register
0x01C
SYNC Code Filter Register
0x100
Pending Interrupt Masked Status Register
0x104
Pending Interrupt Masked Register
0x108
Pending Interrupt Status Register
0x10C
Pending Interrupt Register
0x110
Interrupt Mask Register
0x114
Pending Interrupt Clear Register
0x200
Transmit Channel Control Register
0x204
Transmit Channel Address Register
0x208
Transmit Channel Size Register
0x20C
Transmit Channel Write Register
0x210
Transmit Channel Read Register
0x214
Transmit Channel Interrupt Register
0x300
Receive Channel Control Register
0x304
Receive Channel Address Register
0x308
Receive Channel Size Register
0x30C
Receive Channel Write Register
0x310
Receive Channel Read Register
0x314
Receive Channel Interrupt Register
0x318
Receive Channel Mask Register
0x31C
Receive Channel Code Register
GR740-UM-DS, Jun 2016, Version 1.4
315
www.cobham.com/gaisler
GR740
22.9.1 Configuration Register [CanCONF]
Table 395.0x000 - CanCONF - Configuration Register
31
24 23
SCALER
20 19
PS1
16 15 14
PS2
12 11 10
RSJ
9
8
BPR
7
6
5
4
3
2
1
0
SA SIL SE EN EN AB
M EN LE AB AB OR
T CT LE LE T
1 0
0
0
0
0
0
0
rw
rw
rw
rw
rw
rw rw rw rw rw rw
31: 24
SCALER - Prescaler setting. System clock / (SCALER +1)
23: 20
Phase Segment 1 (PS1) - valid range 1 to 15
19: 16
Phase Segment 2 (PS2) - valid range 2 to 8
14: 12
ReSynchronization Jumps (RSJ) - valid range 1 to 4
9: 8
BPR - Baud rate:
0
0
0
0
0
0b00 = system clock / (SCALER +1) / 1
0b01 = system clock / (SCALER +1) / 2
0b10 = system clock / (SCALER +1) / 4
0b11 = system clock / (SCALER +1) / 8
5
SAM - Single sample when 0. Triple sample when 1
4
SILENT - Listen only to the CAN bus, send recessive bits
3
SELECT - Selection receiver input and transmitter output:
Select receive input 0 as active when 0,
Select receive input 1 as active when 1
Select transmit output 0 as active when 0
Select transmit output 1 as active when 1
2
ENABLE1 - Set value of output 1 enable
1
ENABLE0 - Set value of output 0 enable
0
ABORT - Abort transfer on AHB ERROR
Note that constraints on PS1, PS2 and RSJ are defined as:
•
PS1 +1 >= PS2
•
PS1 > PS2
•
PS2
>= RSJ
Note that CAN standard TSEG1 is defined by PS1+1.
Note that CAN standard TSEG2 is defined by PS2.
Note that the SCALER setting defines the CAN time quantum, together with the BPR setting:
system clock / ((SCALER+1) * BPR)
where SCALER is in range 0 to 255, and the resulting division factor due to BPR is 1, 2, 4 or 8.
For a quantum equal to one system clock period, an additional quantum is added to the node delay.
Note that for minimizing the node delay, then set either SCALER > 0 or BRP > 0.
Note that the resulting bit rate is:
system clock / ((SCALER+1) * BPR * (1+ PS1+1 + PS2))
where PS1 is in the range 1 to 15, and PS2 is in the range 2 to 8.
Note that RSJ defines the number of allowed re-synchronization jumps according to the CAN standard, being in the range 1 to 4.
For SAM = 0b (single), the bus is sampled once; recommended for high speed buses (SAE class C).
GR740-UM-DS, Jun 2016, Version 1.4
316
www.cobham.com/gaisler
GR740
For SAM = 1b (triple), the bus is sampled three times; recommended for low/medium speed buses
(SAE class A and B) where filtering spikes on the bus line is beneficial.
Note that the transmit or receive channel active during the AMBA AHB error is disabled if the
ABORT bit is set to 1b. Note that all accesses to the AMBA AHB bus will be disabled after an AMBA
AHB error occurs while the ABORT bit is set to 1b. The accesses will be disabled until the CanSTAT
register is read.
22.9.2 Status Register [CanSTAT]
Table 396.0x004 - CanSTAT - Status Register
31
28 27
24 23
16 15
8
7
5
4
3
2
1
0
TxChannels
RxChannels
TxErrCnt
RxErrCnt
0
0
0
0
0
0
0
0
0
r
r
r
r
r
r
r
r
r
31: 28
TxChannels - Number of TxChannels -1
27: 24
RxChannels - Number of RxChannels -1
Acti AH OR Off Pa
ve BEr
ss
r
23: 16
TxErrCntr - Transmission error counter
15: 8
RxErrCntr - Reception error counter
4
ACTIVE - Transmission ongoing
3
AHBErr - AMBA AHB master interface blocked due to previous AHB error
2
OR - Overrun during reception
1
OFF - Bus-off condition
0
PASS - Error-passive condition
The OR bit is set if a message with a matching ID is received and cannot be stored via the AMBA
AHB bus, this can be caused by bandwidth limitations or when the circular buffer for reception is
already full.
The OR and AHBErr status bits are cleared when the register has been read.
Note that TxErrCntr and RxErrCntr are defined according to CAN protocol.
Note that the AHBErr bit is only set to 1b if an AMBA AHB error occurs while the CanCONF.ABORT bit is set to 1b.
22.9.3 Control Register [CanCTRL]
Table 397.0x008 - CanCTRL - Control Register
31
2
1
0
Re En
set abl
e
0
0
rw rw
1
RESET - Reset complete core when 1
0
ENABLE - Enable CAN controller, when 1. Reset CAN controller, when 0
Note that RESET is read back as 0b.
Note that ENABLE should be cleared to 0b to while other settings are modified, ensuring that the
CAN core is properly synchronized.
Note that when ENABLE is cleared to 0b, the CAN interface is in sleep mode, only outputting recessive bits.
GR740-UM-DS, Jun 2016, Version 1.4
317
www.cobham.com/gaisler
GR740
Note that the CAN core requires that 10 recessive bits be received before receive and transmit operations can begin.
22.9.4 SYNC Mask Filter Register [CanMASK]
Table 398.0x018- CanMASK - SYNC Mask Filter Register
31
29 28
0
MASK
0x1FFFFFFF
rw
28: 0
MASK - Message Identifier
Note that Base ID is bits 28 to 18 and Extended ID is bits 17 to 0.
A RxSYNC message ID is matched when:
((Received-ID XOR CanCODE.SYNC) AND CanMASK.MASK) = 0
A TxSYNC message ID is matched when:
((Transmitted-ID XOR CanCODE.SYNC) AND CanMASK.MASK) = 0
22.9.5 SYNC Code Filter Register [CanCODE]
Table 399.0x01C- CanCODE - SYNC Code Filter Register
31
29 28
0
SYNC
0
rw
28: 0
SYNC - Message Identifier
Note that Base ID is bits 28 to 18 and Extended ID is bits 17 to 0.
22.9.6 Transmit Channel Control Register [CanTxCTRL]
Table 400.0x200 - CanTxCTRL - Transmit Channel Control Register
31
3
2
1
0
Sin On En
gle goi abl
ng e
0
0
0
rw rw rw
2
SINGLE - Single shot mode
1
ONGOING - Transmission ongoing
0
ENABLE - Enable channel
Note that if the SINGLE bit is 1b, the channel is disabled (i.e. the ENABLE bit is cleared to 0b) if the
arbitration on the CAN bus is lost.
Note that in the case an AHB bus error occurs during an access while fetching transmit data, and the
CanCONF.ABORT bit is 1b, then the ENABLE bit will be reset automatically.
At the time the ENABLE is cleared to 0b, any ongoing message transmission is not aborted, unless
the CAN arbitration is lost or communication has failed.
Note that the ONGOING bit being 1b indicates that message transmission is ongoing and that configuration of the channel is not safe.
GR740-UM-DS, Jun 2016, Version 1.4
318
www.cobham.com/gaisler
GR740
22.9.7 Transmit Channel Address Register [CanTxADDR]
Table 401.0x204 - CanTxADDR - Transmit Channel Address Register
31
10
9
0
ADDR
0
rw
31: 10
ADDR - Base address for circular buffer
22.9.8 Transmit Channel Size Register [CanTxSIZE]
Table 402.0x208 - CanTxSIZE - Transmit Channel Size Register
31
21 20
6
5
0
SIZE
0
rw
20: 6
SIZE - The size of the circular buffer is SIZE*4 messages
Valid SIZE values are between 0 and 16384.
Note that each message occupies four 32-bit words.
Note that the resulting behavior of invalid SIZE values is undefined.
Note that only (SIZE*4)-1 messages can be stored simultaneously in the buffer. This is to simplify
wrap-around condition checking.
22.9.9 Transmit Channel Write Register [CanTxWR]
Table 403.0x20C - CanTxWR - Transmit Channel Write Register
31
20 19
4
3
0
WRITE
0
rw
19: 4
WRITE - Pointer to last written message +1
The WRITE field is written to in order to initiate a transfer, indicating the position +1 of the last message to transmit.
Note that it is not possible to fill the buffer. There is always one message position in buffer unused.
Software is responsible for not over-writing the buffer on wrap around (i.e. setting WRITE=READ).
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
GR740-UM-DS, Jun 2016, Version 1.4
319
www.cobham.com/gaisler
GR740
22.9.10 Transmit Channel Read Register [CanTxRD]
Table 404.0x210- CanTxRD - Transmit Channel Read Register
31
20 19
4
3
0
READ
0
rw
19: 4
READ - Pointer to last read message +1
The READ field is written to automatically when a transfer has been completed successfully, indicating the position +1 of the last message transmitted.
Note that the READ field can be used to read out the progress of a transfer.
Note that the READ field can be written to in order to set up the starting point of a transfer. This
should only be done while the transmit channel is not enabled.
Note that the READ field can be automatically incremented even if the transmit channel has been disabled, since the last requested transfer is not aborted until CAN bus arbitration is lost.
When the Transmit Channel Read Pointer catches up with the Transmit Channel Write Register, an
interrupt is generated (TxEmpty). Note that this indicates that all messages in the buffer have been
transmitted.
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
22.9.11 Transmit Channel Interrupt Register [CanTxIRQ]
Table 405.0x214 - CanTxRD - Transmit Channel Read Register
31
20 19
4
3
0
IRQ
0
rw
19: 4
IRQ - Interrupt is generated when CanTxRD.READ=IRQ, as a consequence of a message transmission
Note that this indicates that a programmed number of messages have been transmitted.
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
22.9.12 Receive Channel Control Register [CanRxCTRL]
Table 406.0x300 - CanRxCTRL - Receive Channel Control Register
31
2
1
0
On En
goi abl
ng e
1
ONGOING - Reception ongoing
0
ENABLE - Enable channel
0
0
r
rw
Note that in the case an AHB bus error occurs during an access while fetching transmit data, and the
CanCONF.ABORT bit is 1b, then the ENALBE bit will be reset automatically.
At the time the ENABLE is cleared to 0b, any ongoing message reception is not aborted
GR740-UM-DS, Jun 2016, Version 1.4
320
www.cobham.com/gaisler
GR740
Note that the ONGOING bit being 1b indicates that message reception is ongoing and that configuration of the channel is not safe.
22.9.13 Receive Channel Address Register [CanRxADDR]
Table 407.0x304 - CanRxADDR - Receive Channel Address Register
31
10
9
0
ADDR
0
rw
31: 10
ADDR - Base address for circular buffer
22.9.14 Receive Channel Size Register [CanRxSIZE]
Table 408.0x308 - CanRxSIZE - Receive Channel Size Register
31
21 20
6
5
0
SIZE
0
rw
20: 6
SIZE - The size of the circular buffer is SIZE*4 messages
Valid SIZE values are between 0 and 16384.
Note that each message occupies four 32-bit words.
Note that the resulting behavior of invalid SIZE values is undefined.
Note that only (SIZE*4)-1 messages can be stored simultaneously in the buffer. This is to simplify
wrap-around condition checking.
22.9.15 Receive Channel Write Register [CanRxWR]
Table 409.0x30C - CanRxWR - Receive Channel Write Register
31
20 19
4
3
0
WRITE
0
rw
19: 4
WRITE - Pointer to last written message +1
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
The WRITE field is written to automatically when a transfer has been completed successfully, indicating the position +1 of the last message received.
Note that the WRITE field can be used to read out the progress of a transfer.
Note that the WRITE field can be written to in order to set up the starting point of a transfer. This
should only be done while the receive channel is not enabled.
GR740-UM-DS, Jun 2016, Version 1.4
321
www.cobham.com/gaisler
GR740
22.9.16 Receive Channel Read Register [CanRxRD]
Table 410.0x310 - CanRxRD - Receive Channel Read Register
31
20 19
4
3
0
READ
0
rw
19: 4
READ - Pointer to last read message +1
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
The READ field is written to in order to release the receive buffer, indicating the position +1 of the
last message that has been read out.
Note that it is not possible to fill the buffer. There is always one message position in buffer unused.
Software is responsible for not over-reading the buffer on wrap around (i.e. setting WRITE=READ).
22.9.17 Receive Channel Interrupt Register [CanRxIRQ]
Table 411.0x314 - CanRxIRQ - Receive Channel Interrupt Register
31
20 19
4
3
0
IRQ
0
rw
19: 4
IRQ - Interrupt is generated when CanRxWR.WRITE=IRQ, as a consequence of a message reception
Note that this indicates that a programmed number of messages have been received.
The field is implemented as relative to the buffer base address (scaled with the SIZE field).
22.9.18 Receive Channel Mask Register [CanRxMASK]
Table 412.0x318 - CanRxMASK - Receive Channel Mask Register
31
29 28
0
AM
0x1FFFFFFF
rw
28: 0
Acceptance Mask (AM) - Bits set to 1b are taken into account in the comparison between the
received message ID and the CanRxCODE.AC field
Note that Base ID is bits 28 to 18 and Extended ID is bits 17 to 0.
22.9.19 Receive Channel Code Register [CanRxCODE]
Table 413.0x31C - CanRxCODE - Receive Channel Code Register
31
29 28
0
AC
0
rw
28: 0
Acceptance Code (AC) - Used in comparison with the received message
Note that Base ID is bits 28 to 18 and Extended ID is bits 17 to 0.
GR740-UM-DS, Jun 2016, Version 1.4
322
www.cobham.com/gaisler
GR740
A message ID is matched when:
((Received-ID XOR CanRxCODE.AC) AND CanRxMASS.AM) = 0
22.9.20 Interrupt registers
The interrupt registers allow for various transmission and reception strategies, by providing means to
mask interrupts, clear interrupts, force interrupts and read interrupt status.
When an interrupt occurs the corresponding bit in the Pending Interrupt Register is set. The normal
sequence to initialize and handle a module interrupt is:
•
Set up the software interrupt-handler to accept an interrupt from the module.
•
Read the Pending Interrupt Register to clear any spurious interrupts.
•
Initialize the Interrupt Mask Register, unmasking each bit that should generate the module interrupt.
•
When an interrupt occurs, read the Pending Interrupt Status Register in the software interrupthandler to determine the causes of the interrupt.
•
Handle the interrupt, taking into account all causes of the interrupt.
•
Clear the handled interrupt using Pending Interrupt Clear Register.
Masking interrupts: After reset, all interrupt bits are masked, since the Interrupt Mask Register is zero.
To enable generation of a module interrupt for an interrupt bit, set the corresponding bit in the Interrupt Mask Register.
Clearing interrupts: All bits of the Pending Interrupt Register are cleared when it is read or when the
Pending Interrupt Masked Register is read. Reading the Pending Interrupt Masked Register yields the
contents of the Pending Interrupt Register masked with the contents of the Interrupt Mask Register.
Selected bits can be cleared by writing ones to the bits that shall be cleared to the Pending Interrupt
Clear Register.
Forcing interrupts: When the Pending Interrupt Register is written, the resulting value is the original
contents of the register logically OR-ed with the write data. This means that writing the register can
force (set) an interrupt bit, but never clear it.
Reading interrupt status: Reading the Pending Interrupt Status Register yields the same data as a read
of the Pending Interrupt Register, but without clearing the contents.
Reading interrupt status of unmasked bits: Reading the Pending Interrupt Masked Status Register
yields the contents of the Pending Interrupt Register masked with the contents of the Interrupt Mask
Register, but without clearing the contents.
The interrupt registers comprise the following:
•
Pending Interrupt Masked Status Register[CanPIMSR]R
•
Pending Interrupt Masked Register[CanPIMR]R
•
Pending Interrupt Status Register[CanPISR]R
•
Pending Interrupt Register[CanPIR]R/W
•
Interrupt Mask Register[CanIMR]R/W
•
Pending Interrupt Clear Register[CanPICR]W
Table 414.Interrupt Registers
31
GR740-UM-DS, Jun 2016, Version 1.4
17 16 15 14 13 12 11 10
5
4
TxL Rx Tx Rx Tx Rx Tx Rx Tx Rx Tx Rx
oss Mis Err Err Sy Sy
Em Full IR IR
s Cnt Cnt nc nc
pty
Q Q
r
r
Tx
AH
B
Err
323
9
8
7
6
3
2
1
0
Rx OR Off Pa
AH
ss
B
Err
www.cobham.com/gaisler
GR740
Table 414.Interrupt Registers
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
16
TxLoss - Message arbitration lost during transmission (could be caused by communications error, as
indicated by other interrupts as well)
15
RxMiss - Message filtered away during reception
14
TxErrCntr - Transmission error counter incremented
13
RxErrCntr - Reception error counter incremented
12
TxSync - Synchronization message transmitted
11
RxSync - Synchronization message received
10
Tx - Successful transmission of message
9
Rx - Successful reception of message
8
TxEmpty - Successful transmission of all messages in buffer
7
RxFull - Successful reception of all messages possible to store in buffer
6
TxIRQ - Successful transmission of a predefined number of messages
5
RxIRQ - Successful reception of a predefined number of messages
4
TxAHBErr - AHB error during transmission
3
RxAHBErr - AHB error during reception
2
OR - Over-run during reception
1
OFF - Bus-off condition
0
PASS - Error-passive condition
* Read-only or read-write depends on the register according to the previous description
Note that the TxAHBErr interrupt is generated in such way that the corresponding read and write
pointers are valid for failure analysis. The interrupt generation is independent of the CanCONF.ABORT field setting.
Note that the RxAHBErr interrupt is generated in such way that the corresponding read and write
pointers are valid for failure analysis. The interrupt generation is independent of the CanCONF.ABORT field setting.
GR740-UM-DS, Jun 2016, Version 1.4
324
www.cobham.com/gaisler
GR740
22.10 Memory mapping
The CAN message is represented in memory as shown in table 415.
Table 415.CAN message representation in memory.
AHB addr
0x0
31
30
29
28
IDE
RT
R
-
bID
15
14
13
12
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
eID
11
10
9
8
7
6
5
4
3
2
1
0
22
21
20
19
18
17
16
2
1
0
eID
0x4
31
DLC
15
14
13
12
27
26
25
24
23
-
-
-
-
TxErrCntr
11
10
9
8
7
6
5
4
3
-
-
-
-
Ahb OR
Err
Off
Pass
23
22
21
20
19
18
17
16
6
5
4
3
2
1
0
22
21
20
19
18
17
16
6
5
4
3
2
1
0
RxErrCntr
0x8
31
30
29
28
27
26
25
24
Byte 0 (first transmitted)
15
14
13
12
Byte 1
11
10
9
8
Byte 2
0xC
31
30
29
28
27
26
25
24
14
13
12
11
10
9
8
Byte 4
15
7
Byte 3
23
Byte 5
Byte 6
7
Byte 7 (last transmitted)
Values: Levels according to CAN standard: 1b is recessive,
0b is dominant
Legend: Naming and number in according to CAN standard
IDE
Identifier Extension:
1b for Extended Format,
0b for Standard Format
RTR
Remote Transmission Request: 1b for Remote Frame,
0b for Data Frame
bID
Base Identifier
eID
Extended Identifier
DLC
Data Length Code, according to CAN standard:
0000b 0 bytes
0001b 1 byte
0010b 2 bytes
0011b 3 bytes
0100b 4 bytes
0101b 5 bytes
0110b 6 bytes
0111b 7 bytes
1000b 8 bytes
OTHERS illegal
TxErrCntr Transmission Error Counter
RxErrCntr Reception Error Counter
AHBErr AHB interface blocked due to AHB Error when 1b
OR
Reception Over run when 1b
OFF
Bus Off mode when 1b
PASS
Error Passive mode when 1b
GR740-UM-DS, Jun 2016, Version 1.4
325
www.cobham.com/gaisler
GR740
Byte 00 to 07 Transmit/Receive data, Byte 00 first Byte 07 last
GR740-UM-DS, Jun 2016, Version 1.4
326
www.cobham.com/gaisler
GR740
23
Bridge connecting Slave I/O AHB bus to Processor AHB bus
23.1
Overview
A uni-directional AHB/AHB bridge is used to connect the Processor AHB bus to the Slave I/O bus.
The buses are connected through a pair consisting of an AHB slave and an AHB master interface.
AHB transfer forwarding is performed in one direction, where AHB transfers to the slave interface are
forwarded to the master interface.
Features offered by the uni-directional AHB to AHB bridge are:
23.2
•
Single and burst AHB transfers
•
Data buffering in internal FIFOs
•
Efficient bus utilization through use of AMBA SPLIT response and data prefetching
•
Posted writes
•
Read and write combining, improves bus utilization and allows connecting cores with differing
AMBA access size restrictions.
Operation
23.2.1 General
The bridge is capable of handling single and burst transfers of all burst types. Supported transfer sizes
(HSIZE) are BYTE, HALF-WORD, WORD, DWORD, 4WORD and 8WORD.
For AHB write transfers write data is always buffered in an internal FIFO implementing posted
writes. For AHB read transfers the bridge uses GRLIB’s AMBA Plug&Play information to determine
whether the read data will be prefetched and buffered in an internal FIFO. If the target address for an
AHB read burst transfer is a prefetchable location the read data will be prefetched and buffered.
An AHB master initiating a read transfer to the bridge always receives a SPLIT response on the first
transfer attempt to allow other masters to use the Processor AHB bus while the bridge performs the
read transfer on the Slave I/O AHB bus.
23.2.2 AHB read transfers
When a read transfer is registered on the slave interface the bridge (connected to the Processor AHB
bus) gives a SPLIT response. The master that initiated the transfer will be de-granted allowing other
bus masters to use the slave bus while the bridge performs a read transfer on the master side (on the
Slave I/O bus). The master interface requests the bus and starts the read transfer on the master side.
Single transfers on the Processor AHB bus are normally translated to single transfers with the same
AHB address and control signals on the master side, however read combining can translate one access
into several smaller accesses. Translation of burst transfers from the Processor AHB bus to the Slave
I/O bus side depends on the burst type, burst length and access size.
If the transfer is a burst transfer to a prefetchable location, the master interface will prefetch data in
the internal read FIFO. If the SPLIT burst on the slave side was an incremental burst of unspecified
length (INCR), the master interface performs an incremental burst up to a 32-byte address boundary.
When the burst transfer is completed on the Slave I/O AHB bus side, the SPLIT master that initiated
the transfer (on the Processor AHB bus) is allowed in bus arbitration by asserting the appropriate
HSPLIT signal to the AHB controller. The SPLIT master re-attempts the transfer and the bridge will
return data with zero wait states.
If the burst is to non-prefetchable area, the burst transfer on the master side is performed using
sequence of NONSEQ, BUSY and SEQ transfers. The first access in the burst on the master side is of
NONSEQ type. Since the master interface can not decide whether the splitted burst will continue on
the slave side or not, the master bus is held by performing BUSY transfers. On the slave side the split-
GR740-UM-DS, Jun 2016, Version 1.4
327
www.cobham.com/gaisler
GR740
ted master that initiated the transfer is allowed in bus arbitration by asserting the HSPLIT signal to the
AHB controller. The first access in the transfer is completed by returning read data. The next access in
the transfer on the slave side is extended by asserting HREADY low. On the master side the next
access is started by performing a SEQ transfer (and then holding the bus using BUSY transfers). This
sequence is repeated until the transfer is ended on the slave side.
In case of an ERROR response on the master side the ERROR response will be given for the same
access (address) on the slave side. SPLIT and RETRY responses on the master side are re-attempted
until an OKAY or ERROR response is received.
23.2.3 AHB write transfers
The AHB/AHB bridge implements posted writes. During the AHB write transfer on the slave side the
data is buffered in the internal write FIFO and the transfer is completed on the slave side by always
giving an OKAY response. The master interface requests the bus and performs the write transfer when
the master bus is granted.
Writes are accepted with zero wait states if the bridge is idle and the incoming access is not locked. If
the incoming access is locked, each access will have one wait state.
23.2.4 Locked transfers
The AHB/AHB bridge supports locked transfers. When a locked transfer is made from the Processor
AHB bus, the Slave I/O AHB bus will be locked when the bus is granted and remain locked until the
transfer completes on the Processor AHB side.
Locked transfers can lead to deadlock conditions when a locked transfer is made after a read access
that has received a SPLIT response from the bridge. The AMBA specification requires that the locked
transfer is handled before the previous transfer, which received a SPLIT response, is completed. The
bridge will avoid the deadlock condition by saving state for the read access that received a SPLIT
response, allow the locked access to complete, and then complete the first access. All non-locked
accesses from other masters will receive SPLIT responses until the saved data has been read out.
23.2.5 Read and write combining
Read and write combining allows the bridge to assemble or split AMBA accesses on the bridge’s
slave interface into one or several accesses on the master interface. The table below shows the effect
of read and write combining on incoming access from the Processor AHB bus.
Table 416.Read and write combining
Access on slave interface
Access size
Resulting access(es) on master interface
BYTE or HALF-WORD single
read access to any area
-
Single access of same size
BYTE or HALF-WORD read
burst to prefetchable area
-
Incremental read burst of same access size as on slave interface, the
length is the same as the number of 32-bit words in the read buffer, but
will not cross the read burst boundary.
BYTE or HALF-WORD read
burst to non-prefetchable area
-
Incremental read burst of same access size as on slave interface, the
length is the same as the length of the incoming burst. The master
interface will insert BUSY cycles between the sequential accesses.
BYTE or HALF-WORD single
write
-
Single access of same size
BYTE or HALF-WORD write
burst
-
Incremental write burst of same size and length, the maximum length
is the number of 32-bit words in the write FIFO.
Single read access to any area
Access size <=
32-bits
Single access of same size
Single read access to any area
Access size >
32-bits
Burst of 32-bit accesses. Length of burst: (access size)/(32 bits)
GR740-UM-DS, Jun 2016, Version 1.4
328
www.cobham.com/gaisler
GR740
Table 416.Read and write combining
Access on slave interface
Access size
Resulting access(es) on master interface
Read burst to prefetchable area -
Burst of 32-bit accesses up to 32-byte address boundary.
Read burst to non-prefetchable
area
Access size <=
32-bits
Incremental read burst of same access size as on slave interface, the
length is the same as the length of the incoming burst. The master
interface will insert BUSY cycles between the sequential accesses.
Read burst to non-prefetchable
area
Access size >
32-bits
Burst of 32-bit accesses. Length of burst: 
(incoming burst length)*(access size)/(32 bits)
Single write
Access size <=
32-bits
Single write access of same size
Single write
Access size >
32-bits
Burst of 32-bit accesses. Length of burst: (access size)/(32 bits).
Write burst
-
Burst of 32-bit accesses
23.2.6 Transaction ordering
The bridge implements first-come, first-served ordering and will keep track of the order of incoming
accesses. The accesses will then be served in the same order. For instance, if master 0 initiates an
access to the bridge, followed by master 3 and then master 5, the bridge will propagate the access
from master 0 (and respond with SPLIT on a read access) and then respond with SPLIT to the other
masters. When the bridge has a response for master 0, this master will be allowed in arbitration again
by the bridge asserting HSPLIT. When the bridge has finished serving master 0 it will allow the next
queued master in arbitration, in this case master 3. Other incoming masters will receive SPLIT
responses and will not be allowed in arbitration until all previous masters have been served.
An incoming locked access will always be given precedence over any other masters in the queue.
23.2.7 Core latency
The delay incurred when performing an access over the core depends on several parameters such as
the operating frequency of the AMBA buses and memory access patterns. Table 417 below shows one
example of core behavior.
Table 417.Example of single read
Clock cycle
Core slave side activity
Core master side activity
0
Discovers access and transitions from idle state
Idle
1
Slave side waits for master side, SPLIT response
is given to incoming access, any new incoming
accesses also receive SPLIT responses.
Discovers slave side transition. Master interface output
signals are assigned.
2
3
If bus access is granted, perform address phase. Otherwise wait for bus grant.
Register read data and transition to data ready state.
4
Discovers that read data is ready, assign read
data output and assign SPLIT complete
5
SPLIT complete output is HIGH
6
Typically a wait cycle for the SPLIT:ed master to
be allowed into arbitration. Core waits for master
to return. Other masters receive SPLIT
responses.
7
Master has been allowed into arbitration and performs address phase. Core keeps HREADY high
8
Access data phase. Core has returned to idle
state.
GR740-UM-DS, Jun 2016, Version 1.4
329
Idle
www.cobham.com/gaisler
GR740
While the transitions shown in table 417 are simplified they give an accurate view of the core delay. If
the read operation receives wait states, these cycles must be added to the cycle count in the table.
Table 418 below lists the delays incurred for single operations that traverse the bridge while the bridge
is in its idle state. The second column shows the number of cycles it takes the master side to perform
the requested access, this column assumes that the master slave gets access to the bus immediately
and that each access is completed with zero wait states. The table only includes the delay incurred by
traversing the core. For instance, when the access initiating master reads the core’s prefetch buffer,
each additional read will consume one clock cycle. However, this delay would also have been present
if the master accessed any other slave.
Write accesses are accepted with zero wait states if the bridge is idle, this means that performing a
write to the idle core does not incur any extra latency. However, the core must complete the write
operation on the master side before it can handle a new access on the slave side. If the core has not
transitioned into its idle state, pending the completion of an earlier access, the delay suffered by an
access be longer than what is shown in the tables in this section. Locked accesses that abort on-going
read operations will also mean additional delays.
With read and write combining, the number of cycles required for the master will change depending
on the access size and length of the incoming burst access.
Table 418.Access latencies
Access
Master acc. cycles Slave cycles
Delay incurred by performing access over core
Single read
3
2
5* clk
Burst read with prefetch
2 + (burst length)x
4
(6 + burst length)* clk
Single writexx
(2)
0
0
Burst writexx
(2 + (burst length))
0
0
x A prefetch
xx The
23.3
operation ends at the address boundary defined by the prefetch buffer’s size
core implements posted writes, the number of cycles taken by the master side can only affect the next access.
Registers
The bridge does not implement any registers.
GR740-UM-DS, Jun 2016, Version 1.4
330
www.cobham.com/gaisler
GR740
24
Fault-tolerant 8/16-bit PROM/IO Memory Interface
24.1
Overview
The combined 8/16-bit memory controller provides a bridge between external memory and the AHB
bus. The memory controller can handle two types of devices: PROM and memory mapped I/O
devices (IO). The PROM area can be EDAC-protected using a (39,7) BCH code. The BCH code provides single-error correction and double-error detection for each 32-bit memory word.
The memory controller is configured through three configuration registers accessible via an APB bus
interface. The external data bus can be configured in 8-, 16-bit mode, depending on application
requirements. The controller decodes two address spaces on the AHB bus (PROM, IO)
External chip-selects are provided for up to two PROM banks and one IO bank. Figure 31 below
shows how the connection to the different device types is made.
APB
A
AHB
APB
PROM_CEN[1 :0]
PROMIO_OEN
PROMIO_WEN
CS
OE
WE
PROM
A
D
IO_SN
CS
OE
WE
I/O
A
D
FTMCTRL
AHB
D
PROMIO_ADDR[27:0]
PROMIO_DATA[15:0]
Figure 31. FTMCTRL connected to different types of memory devices
24.2
PROM access
Up to two PROM chip-select signals are provided for the PROM area, PROM_CEN[1:0]. The size of
the banks can be set in binary steps from 16 KiB to 256 MiB.
A read access to PROM consists of two data cycles and between 0 and 240 waitstates. The read data
(and optional EDAC check-bits) are latched on the rising edge of the clock on the last data cycle. On
non-consecutive accesses, a idle cycle is placed between the read cycles to prevent bus contention due
to slow turn-off time of PROM devices. Figure 32 shows the basic read cycle waveform (zero waitstate) for non-consecutive PROM reads. Note that the address is undefined in the idle cycle. Figure 33
shows the timing for consecutive cycles (zero waitstate). Waitstates are added by extending the data2
phase. This is shown in figure 34 and applies to both consecutive and non-consecutive cycles. Only an
even number of waitstates can be assigned to the PROM area.
GR740-UM-DS, Jun 2016, Version 1.4
331
www.cobham.com/gaisler
GR740
data1
data2
data1
data2
clk
promio_addr
A2
A1
prom_cen
promio_oen
promio_data
D1
D2
Figure 32. Prom non-consecutive read cyclecs.
data1
data2
data1
data
data2
clk
promio_addr
A1
A2
prom_cen
promio_oen
promio_data
D1
D2
Figure 33. Prom consecutive read cyclecs.
data1
data2
data2
data
data2
clk
promio_addr
A1
prom_cen
promio_oen
promio_data
D1
Figure 34. Prom read access with two waitstates.
GR740-UM-DS, Jun 2016, Version 1.4
332
www.cobham.com/gaisler
GR740
lead-in
data
lead-out
clk
promio_addr
A1
prom_cen
promio_wen
promio_data
D1
Figure 35. Prom write cycle (0-waitstates)
lead-in
data
data
data lead-out
clk
promio_addr
A1
prom_cen
promio_wen
promio_data
D1
Figure 36. Prom write cycle (2-waitstates)
24.3
Memory mapped IO
Accesses to IO have similar timing as PROM accesses. The IO select (IO_SN) and output enable
(PROMIO_OEN) signals are delayed one clock to provide stable address before IO_SN is asserted.
All accesses are performed as non-consecutive accesses as shown in figure 37. The data2 phase is
extended when waitstates are added.
GR740-UM-DS, Jun 2016, Version 1.4
333
www.cobham.com/gaisler
GR740
lead-in
data1
data2
lead-out
clk
promio_addr
A1
ios_n
promio_oen
promio_data
D1
Figure 37. I/O read cycle (0-waitstates)
lead-in
data
lead-out
clk
promio_addr
A1
io_sn
promio_wen
promio_data
D1
Figure 38. I/O write cycle (0-waitstates)
24.4
8-bit and 16-bit PROM access
The PROM areas can be configured for 8- or 16-bit operation by programming the ROM width field
in the memory configuration register. Since reads to memory are always done on 32-bit word basis,
read access to 8-bit memory will be transformed in a burst of four read cycles while access to 16-bit
memory will generate a burst of two 16-bit reads. During writes, only the necessary bytes will be written. Figure 39 shows an interface example with 8-bit PROM. Figure 40 shows an example of a 16-bit
memory interface.
EDAC is not supported for 16-bit wide memories and therefore the EDAC enable bit corresponding to
a 16-bit wide area must not be set.
GR740-UM-DS, Jun 2016, Version 1.4
334
www.cobham.com/gaisler
GR740
8-bit PROM
PROM_CEN[0]
PROMIO_OEN
PROMIO_WEN
CS
OE
WE
A
D
A[25:0]
PROM
A
D
D[15:8]
MEMORY
CONTROLLER
PROMIO_ADDR[27:0]
PROMIO_DATA[15:8]
Figure 39. 8-bit memory interface example
16-bit PROM
PROM_CEN[0]
PROMIO_OEN
PROMIO_WEN
CS
OE
WE
A
D
A[26:1]
PROM
A
D
D[15:0]
MEMORY
CONTROLLER
PROMIO_ADDR[27:0]
PROMIO_DATA[15:0]
Figure 40. 16-bit memory interface example
In 8-bit mode, the PROM devices should be connected to the MSB byte of the data bus (PROMIO_DATA[15:8]). The LSB address bus should be used for addressing (PROMIO_ADDR[25:0]). In 16bit mode, PROMIO_DATA[15:0] should be used as data bus, and PROMIO_ADDR[26:1] as address
bus. EDAC protection is not available in 16-bit mode.
24.5
8- and 16-bit I/O access
Similar to the PROM area, the IO area can also be configured to 8- or 16-bits mode. However, the I/O
device will NOT be accessed by multiple 8/16 bits accesses as the memory areas, but only with one
single access just as in 32-bit mode. To access an IO device on an 8-bit bus, only byte accesses should
be used (LDUB/STB instructions for the CPU). To accesses an IO device on a 16-bit bus, only halfword accesses should be used (LDUH/STH instructions for the CPU).
24.6
Burst cycles
To improve the bandwidth of the memory bus, accesses to consecutive addresses can be performed in
burst mode. Burst transfers will be generated when the memory controller is accessed using an AHB
burst request. These includes instruction cache-line fills, double loads and double stores. The timing
of a burst cycle is identical to the programmed basic cycle with the exception that during read cycles,
the idle cycle will only occurs after the last transfer. Burst cycles will not be generated to the IO area.
Only word (32-bit) bursts of incremental type is supported. Note that the processors can access the
PROM area using larger accesses. The AHB/AHB bridge connecting the Processor AHB bus to the
Slave I/O AHB bus will split larger accesses into bursts of 32-bit accesses. Note that all accesses to
this memory controller traverses over the bridge connecting the Processor AHB bus to the Slave I/O
bus.
GR740-UM-DS, Jun 2016, Version 1.4
335
www.cobham.com/gaisler
GR740
24.7
Memory EDAC
24.7.1 BCH EDAC
The core provides BCH EDAC that can correct one error and detect two errors in a 32-bit word. For
each word, a 7-bit checksum is generated according to the equations below. A correctable error will be
handled transparently by the memory controller, but adding one waitstate to the access. If an un-correctable error (double-error) is detected, the current AHB cycle will end with an error response. The
EDAC can be used during access to the PROM area by setting the PROM EDAC enable bit in the
MCFG3 register. The equations below show how the EDAC checkbits are generated:
CB0
CB1
CB2
CB3
CB4
CB5
CB6
=
=
=
=
=
=
=
D0
D0
D0
D0
D2
D8
D0
^
^
^
^
^
^
^
D4
D1
D3
D1
D3
D9
D1
^
^
^
^
^
^
^
D6 ^ D7 ^
D2 ^ D4 ^
D4 ^ D7 ^
D5 ^ D6 ^
D4 ^ D5 ^
D10 ^ D11
D2 ^ D3 ^
D8 ^ D9 ^ D11 ^ D14 ^ D17 ^ D18 ^ D19 ^ D21 ^ D26 ^ D28 ^ D29 ^ D31
D6 ^ D8 ^ D10 ^ D12 ^ D16 ^ D17 ^ D18 ^ D20 ^ D22 ^ D24 ^ D26 ^ D28
D9 ^ D10 ^ D13 ^ D15 ^ D16 ^ D19 ^ D20 ^ D23 ^ D25 ^ D26 ^ D29 ^ D31
D7 ^ D11 ^ D12 ^ D13 ^ D16 ^ D17 ^ D21 ^ D22 ^ D23 ^ D27 ^ D28 ^ D29
D6 ^ D7 ^ D14 ^ D15 ^ D18 ^ D19 ^ D20 ^ D21 ^ D22 ^ D23 ^ D30 ^ D31
^ D12 ^ D13 ^ D14 ^ D15 ^ D24 ^ D25 ^ D26 ^ D27 ^ D28 ^ D29 ^ D30 ^ D31
D4 ^ D5 ^ D6 ^ D7 ^ D24 ^ D25 ^ D26 ^ D27 ^ D28 ^ D29 ^ D30 ^ D31
Data is always accessed as words (4 bytes at a time) and the corresponding checkbits are located at the
address acquired by inverting the word address (bits 2 to 27) and using it as a byte address. The same
chip-select is kept active. A word written as four bytes to addresses 0, 1, 2, 3 will have its checkbits at
address 0xFFFFFFF, addresses 4, 5, 6, 7 at 0xFFFFFFE and so on. All the bits up to the maximum
bank size will be inverted while the same chip-select is always asserted. This way all the bank sizes
can be supported and no memory will be unused (except for a maximum of 4 byte in the gap between
the data and checkbit area). A read access will automatically read the four data bytes individually
from the nominal addresses and the EDAC checkbit byte from the top part of the bank. Write accesses
must only be performed as individual byte accesses by the software, writing one byte at a time, and
the corresponding checkbit byte must be calculated and be written to the correct location by the software
NOTE: when the EDAC is enabled in 8-bit bus mode, only the first bank select (PROM_CEN[0]) can
be used.
24.7.2 EDAC Error reporting
As mentioned above an un-correctable error results in an AHB error response which can be monitored
on the bus. Correctable errors however are handled transparently and are not visible on the AHB bus.
A sideband signal is provided which is asserted during one clock cycle for each access for which a
correctable error is detected. This sideband signal is connected to the AHB status register monitoring
the Slave I/O AHB bus (see section 32).
Note that bit errors remain in external memory until a software-initiated re-write is performed at the
faulty memory location.
24.8
Bus Ready signalling
The PROMIO_BRDYN signal can be used to stretch all types of access cycles to the PROM and I/O
areas. The accesses will always have at least the pre-programmed number of waitstates as defined in
memory configuration registers 1 & 2, but will be further stretched until PROMIO_BRDYN is
asserted. PROMIO_BRDYN should be asserted in the cycle preceding the last one. If bit 29 in
MCFG1 is set, PROMIO_BRDYN can be asserted asynchronously with the system clock. In this
case, the read data must be kept stable until the de-assertion of PROMIO_OEN and PROMIO_BRDYN must be asserted for at least 1.5 clock cycle. The use of PROMIO_BRDYN can be
enabled separately for the PROM and I/O areas. It is recommended that PROMIO_BRDYN is
asserted until the corresponding chip select signal is de-asserted, to ensure that the access has been
properly completed and avoiding the system to stall.
GR740-UM-DS, Jun 2016, Version 1.4
336
www.cobham.com/gaisler
GR740
data1
data2
data2
lead-out
clk
promio_address
A1
prom_cen/io_sn
promio_oen
promio_data
D1
promio_brdyn
Figure 41. READ cycle with one extra data2 cycle added with BRDYN (synchronous sampling). Lead-out cycle is only
applicable for I/O accesses.
Figure 42 shows the use of BRDYN with asynchronous sampling. BRDYN is kept asserted for more
than 1.5 clock-cycle. Two synchronization registers are used so it will take at least one additional
cycle from when BRDYN is first asserted until it is visible internally. In figure 42 one cycle is added
to the data2 phase.
data1
data2
data2
lead-out
clk
promio_addr
A1
prom_cen/io_sn
promio_oen
promio_data
D1
promio_brdyn
Figure 42. BRDYN (asynchronous) sampling. Lead-out cycle is only applicable for I/O-accesses.
GR740-UM-DS, Jun 2016, Version 1.4
337
www.cobham.com/gaisler
GR740
data1
data2
data2
ws
data2
brdyn
lead-out
clk
promio_addr
A1
prom_cen/io_sn
promio_oen
promio_data
D1
promio_brdyn
Figure 43. Read cycle with one waitstate (configured) and one BRDYN generated waitstate (synchronous sampling).
The memory controller has been implemented with a bus ready timeout counter. The counter value
and counter reload value are available in MCFG7. The counter will be reloaded whenever the bus
ready signal is low (asserted). If the reload value is nonzero, then the counter will decrement with one
each clock cycle the core is waiting for bus ready to be asserted. If the counter reaches zero, the action
taken depends on the state of Bus Error Enable (BEXCN) in MCFG1. If BEXCN is ‘1’, then an
AMBA ERROR response will be generated and the counter will be reloaded. If BEXCN is ‘0’, then
the bus ready enable for the accessed memory area will be disabled and the core will ignore bus ready
for the accessed area.
Bus ready timeout functionality is disabled when the bus ready counter reload value is zero
(MCFG7.BRDYCNTRLD = 0).
24.9
Registers
The core is programmed through registers mapped into APB address space.
Table 419.FTMCTRL memory controller registers
APB Address offset
Register
0x00
Memory configuration register 1 (MCFG1)
0x04
RESERVED
0x08
Memory configuration register 3 (MCFG3).
0x0C
RESERVED
0x10
Memory configuration register 5 (MCFG5).
0x14
RESERVED
0x18
Memory configuration register 7 (MCFG7)
24.9.1 Memory configuration register 1 (MCFG1)
Memory configuration register 1 is used to program the timing of ROM and IO accesses.
Table 420. Memory configuration register 1
31
30
29
PBRDY ABRDY
28
27
IOBUSW
26
25
IBRDY BEXCN
24
RES
23
20
IO WAITSTATES
19
18
17
IOEN
ROMBANKSZ
0
0
N/R
0
0
0x0
0
0x0
rw
rw
rw
rw
rw
rw
rw
rw
GR740-UM-DS, Jun 2016, Version 1.4
338
www.cobham.com/gaisler
GR740
Table 420. Memory configuration register 1
14
13
12
11
10
9
8
7
4
3
0
RESERVED
PWEN
PROM WIDTH
PROM WRITE WS
PROM READ WS
0
(bootstrap)
0xF
0xF
rw
rw
rw
rw
rw
31
RESERVED
30
PROM area bus ready enable (PBRDY) - Enables bus ready (BRDYN) signalling for the PROM
area. Reset to ‘0’.
29
Asynchronous bus ready (ABRDY) - Enables asynchronous bus ready.
28 : 27
I/O bus width (IOBUSW) - Sets the data width of the I/O area (“00”=8, “01”=16, others=Illegal).
26
I/O bus ready enable (IBRDY) - Enables bus ready (BRDYN) signalling for the I/O area. Reset to
‘0’.
25
Bus Error Enable (BEXCN) - Generate AMBA error response if memory bus timeouts.
24
RESERVED
23 : 20
I/O waitstates (IO WAITSTATES) - Sets the number of waitstates during I/O accesses (“0000”=0,
“0001”=8, “0010”=16,..., “1111”=120). The number of waitstates is 8*(IO WAITSTATES).
19
I/O enable (IOEN) - Enables accesses to the memory bus I/O area.
18
RESERVED
17: 14
PROM bank size (ROMBANKSZ) - Returns current PROM bank size when read. “0000” is a special case and corresponds to a bank size of 256 MiB. All other values give the bank size in binary
steps: “0001”=16KiB, “0010”=32KiB, ... , “1111”=256 MiB.
Programmable bank sizes can be changed by writing to this register field. The written values correspond to the bank sizes and number of chip-selects as above. Reset to “0000” when programmable.
13:12
RESERVED
11
PROM write enable (PWEN) - Enables write cycles to the PROM area.
10
RESERVED
9:8
PROM width (PROM WIDTH) - Sets the data width of the PROM area (“00”=8, “01”=16, others=Illegal).
7:4
PROM write waitstates (PROM WRITE WS) - Sets the number of wait states for PROM write
cycles (“0000”=0, “0001”=16, “0010”=32,..., “1111”=240). The number of waitstates is 16*(PROM
WRITE WS).
3:0
PROM read waitstates (PROM READ WS) - Sets the number of wait states for PROM read cycles
(“0000”=0, “0001”=16, “0010”=32,...,”1111”=240). The number of waitstates is 16*(PROM READ
WS). Reset to “1111”.
During reset, the prom width (bits [9:8]) are set with value on general purpose I/O inputs, see section
3.1. The prom waitstates fields are set to 15 (maximum). External bus ready is disabled. All other
fields are undefined.
24.9.2 Memory configuration register 3 (MCFG3)
MCFG3 contains fields to control and monitor memory EDAC.
Table 421. Memory configuration register 3
31
28
RESERVED
27
26
ME
RESERVED
r
12
11
10
9
8
WB
RB
R
PE
7
TCB
0
0
0
(btstr)
N/R
rw
rw
rw
rw
31 : 28
RESERVED
27
Memory EDAC (ME) - Indicates if memory EDAC is present. (read-only)
GR740-UM-DS, Jun 2016, Version 1.4
339
www.cobham.com/gaisler
GR740
Table 421. Memory configuration register 3
26 : 12
RESERVED
11
EDAC diagnostic write bypass (WB) - RESERVED in this device, always write to 0.
10
EDAC diagnostic read bypass (RB) - RESERVED in this device, always write to 0.
9
RESERVED
8
PROM EDAC enable (PE) - Enable EDAC checking of the PROM area. At reset, this bit is initialized with the value of GPIO line 14 (see section 3.1)
7:0
Test checkbits (TCB) - RESERVED in this device. Always write to 0.
24.9.3 Memory configuration register 5 (MCFG5)
MCFG5 contains fields to control lead out cycles for the ROM and IO areas.
Table 422. Memory configuration register 5
31
30
29
23
RESERVED
22
IOHWS
16
RESERVED
0x00
rw
15
14
13
7
RESERVED
6
ROMHWS
0
RESERVED
0x00
rw
31 : 30
RESERVED
29:23
IO lead out (IOHWS) - Lead out cycles added to IO accesses are IOHWS(3:0)*2IOHWS(6:4)
22 : 14
RESERVED
13:7
ROM lead out (ROMHWS) - Lead out cycles added to ROM accesses are
ROMHWS(3:0)*2ROMHWS(6:4)
6:0
RESERVED
24.9.4 Memory configuration register 7 (MCFG7)
MCFG7 contains fields to control bus ready timeout.
Table 423. Memory configuration register 7
31
16
BRDYNCNT
0
rw
15
0
BRDYNRLD
0
rw
31 : 16
Bus ready count (BRDYNCOUNT) - Counter value. If this register is written then the counter shall
be written with the same value as BRDYNRLD.
15: 0
Bus ready reload value (BRDYNRLD) - Reload value for BRDYNCNT
GR740-UM-DS, Jun 2016, Version 1.4
340
www.cobham.com/gaisler
GR740
25
General Purpose Timer Units
25.1
Overview
A General Purpose Timer Unit acts a slave on AMBA APB bus and provides a common prescaler and
decrementing timers. The system has five general purpose timer units (GPTIMER). Each unit implements one 16-bit prescaler and four or five decrementing timers. The units are capable of asserting
interrupt on timer under flow and the first unit, GPTIMER 0, also provides system watchdog funtionality.
GPTIMER 0 has a separate interrupt line for each timer while GPTIMER units 1 - 4 each use a shared
interrupt for all timers. Several timer units are provided in order to support separated ASMP configurations with potentially shared access to the first timer unit that controls the watchdog system reset.
timer 1 reload
timer 2 reload
prescaler reload
timer n reload
prescaler value
timer 1 value
pirq
timer 2 value
pirq+1
timer n value
pirqn+(n-1)
-1
tick
-1
Figure 44. General Purpose Timer Unit block diagram
25.2
Operation
The prescaler is clocked by the system clock and decremented on each clock cycle. When the prescaler underflows, it is reloaded from the prescaler reload register and a timer tick is generated.
The operation of each timer within a timer unit is controlled through the timer’s control register. A
timer is enabled by setting the enable bit in the control register. The timer value is then decremented
on each prescaler tick. When a timer underflows, it will automatically be reloaded with the value of
the corresponding timer reload register if the restart bit in the control register is set, otherwise it will
stop at -1 and reset the enable bit.
If the interrupt enable bit for a timer is set, a timer unit will signal an interrupt on the appropriate interrupt line when the timer underflow. The interrupt pending bit in the control register of the underflown
timer will be set and remain set until cleared by writing ‘1’. The first timer unit has a separate interrupt line for each timer. The other timer units each use a shared interrupt line for all timers in a unit.
To minimize complexity, timers share the same decrementer. This means that the minimum allowed
prescaler division factor is ntimers+1 (reload register = ntimers) where ntimers is the number of
implemented timers (five for GPTIMER 0 and four for GPTIMER 1 - 4). By setting the chain bit in
the control register timer n can be chained with preceding timer n-1. Timer n will be decremented
each time when timer n-1 underflows.
Each timer can be reloaded with the value in its reload register at any time by writing a ‘one’ to the
load bit in the control register. The last timer on GPTIMER 0 acts as a watchdog, driving the watchdog output signal WDOGN when expired.
GR740-UM-DS, Jun 2016, Version 1.4
341
www.cobham.com/gaisler
GR740
Each timer can be configured to latch its value to a dedicated register when an event is detected on the
interrupt. All timers can be forced to reload when an event is detected on the interrupt bus. A dedicated mask register is provided to filter the interrupts. See also section 5.9.
At reset, all timers are disabled except the watchdog timer on GPTIMER 0. The prescaler value and
reload registers are set to all ones, while the watchdog timer is set to 0xFFFF.
25.3
Registers
The cores are programmed through registers mapped into APB address space. The number of implemented registers depend on the number of implemented timers.
Table 424.General Purpose Timer Unit registers
APB address offset
Register
0x00
Scaler value register
0x04
Scaler reload value register
0x08
Configuration register
0x0C
Timer latch configuration register
0x10
Timer 1 counter value register
0x14
Timer 1 reload value register
0x18
Timer 1 control register
0x1C
Timer 1 latch register
0xn0
Timer n counter value register
0xn4
Timer n reload value register
0xn8
Timer n control register
0xnC
Timer n latch register
Table 425.0x00 - SCALER - Scaler value register
31
16 15
0
RESERVED
SCALER
0
0xFFFF
r
rw
31: 16
RESERVED
15: 0
Scaler value (SCALER)
GR740-UM-DS, Jun 2016, Version 1.4
342
www.cobham.com/gaisler
GR740
Table 426.0x04 - SRELOAD- Scaler reload value register
31
16 15
0
RESERVED
SRELOAD
0
0xFFFF
r
rw
31: 16
RESERVED
15: 0
Scaler reload value (SRELOAD)
Table 427.0x08 - CONFIG- Configuration register
31
14 13 12 11 10
RESERVED
9
8
7
EV ES EL EE DF SI
0
0
0
3
2
0
IRQ
TIMERS
0
0
0
1
*
*
r
rw rw rw rw rw
r
r
r
31: 14
RESERVED
13
External Events (EV). If set then the latch events are taken from the secondary input. If this field is
zero then the source of the latch events is the interrupt bus.
12
Enable set (ES). If set, on the next matching interrupt, the timers will be loaded with the corresponding timer reload values. The bit is then automatically cleared, not to reload the timer values until set
again.
11
Enable latching (EL). If set, on the next matching interrupt, the latches will be loaded with the corresponding timer values. The bit is then automatically cleared, not to load a timer value until set again.
10
Enable external clock source (EE). If set the prescaler is clocked from the external clock source.
9
Disable timer freeze (DF). If set the timer unit can not be freezed, otherwise the debug support unit
can freeze the timer unit when processors enter debug mode.
8
Separate interrupts (SI). Reads ‘1’ if the timer unit generates separate interrupts for each timer, otherwise ‘0’.
7: 3
APB Interrupt: If configured to use common interrupt all timers will drive the same interrupt line,
otherwise timer n will drive the first interrupt line assigned to the core+n. GPTIMER 0 has one dedicated interrupt for each timer, GPTIMER unit 1 to 4 have one shared interrupt line for all timers.
2: 0
Number of implemented timers (TIMERS) - Timer unit 0 has five timers, timer unit 1 - 4 has four
timers.
Table 428.0x0C - LATCHCFG - Timer latch configuration register
31
14 13 12 11 10
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
9
8
7
4
3
2
0
0
r
rw
343
0
LATCHSEL
www.cobham.com/gaisler
GR740
Table 428.0x0C - LATCHCFG - Timer latch configuration register
31: 5
RESERVED
4: 0
Latch select (LATCHSEL) - Specifies what bits of the interrupt bus, or external latch vector, bus that
shall cause the Timer Latch Registers to latch the timer values. If the configuration register EV field
is zero then latching is done based on events on the interrupt bus. If the EV field is ‘1’ then the external latch vecor is used and the following events can be used:
Position 0: MIL-STD-1553B controller RTSYNC event. Time will be latched when a valid command is detected by the controller. Time latching will be disabled when a RTSYNC event is reported
by the MIL-STD-1553B controller.
Position 1: Connectef to SpaceWire router tick out 0
Position 2: Connectef to SpaceWire router tick out 1
Position 3: Connectef to SpaceWire router tick out 2
Position 4: Connectef to SpaceWire router tick out 3
Table 429.0xn0 where n selects the timer - TCNTVALn - Timer n counter value register
31
0
TCVAL
0
rw
31: 0
Timer Counter value (TCVAL) - Decremented by 1 for each prescaler tick.
Table 430.0xn4 where n selects the timer - TRLDVALn - Timer n counter reload value register
31
0
TRLDVAL
*
rw
31: 0
Timer Reload value (TRLDVAL) - This value is loaded into the timer counter value register when
‘1’ is written to the TCTRL.LD load bit or when the TCTRL.RS bit is set and the timer underflows.
This field is set to 0xFFFF for the watchdog timer. The reset value is undefined for the other timers.
Table 431.0xn8 where n selects the timer - TCTRLn - Timer n control register
31
9
RESERVED
8
7
6
5
4
WS WN DH CH IP
0
0
r
0
0
rw* rw
r
*
0
3
2
1
0
IE LD RS EN
*
*
*
*
rw wc rw rw rw rw
31: 9
RESERVED
8
Disable Watchdog Output (WS/WDOGDIS) - If this field is set to ’1’ then the watchdog output will
not be affected by the timer unit. This field is only available for the last timer of timer unit 0.
7
Enable Watchdog NMI (WN/WDOGNMI) - If this field is set to ’1’ then the watchdog timer will
also generate a non-maskable interrupt (interrupt 15) when an interrupt is signaled. This field is only
available for the last timer of timer unit 0.
GR740-UM-DS, Jun 2016, Version 1.4
344
www.cobham.com/gaisler
GR740
Table 431.0xn8 where n selects the timer - TCTRLn - Timer n control register
6
Debug Halt (DH): Value of internal signal that is used to freeze counters (e.g. when a system is in
debug mode).
5
Chain (CH): Chain with preceding timer. If set for timer n, timer n will be decremented each time
when timer (n-1) underflows.
This field is reset to ’0’ for the watchdog timer. It is not reset for the other timers.
4
Interrupt Pending (IP): The core sets this bit to ‘1’ when an interrupt is signalled. This bit remains ‘1’
until cleared by writing ‘1’ to this bit, writes of ‘0’ have no effect.
3
Interrupt Enable (IE): If set the timer signals interrupt when it underflows.
2
Load (LD): Load value from the timer reload register to the timer counter value register.
This field is reset to ’1’ for the watchdog timer. It is reset to ’0’ for the other timers.
This field is reset to ’1’ for the watchdog timer. It is not reset for the other timers.
1
Restart (RS): If set, the timer counter value register is reloaded with the value of the reload register
when the timer underflows.
This field is reset to ’0’ for the watchdog timer. It is not reset for the other timers.
0
Enable (EN): Enable the timer.
The watchdog timer (GPTIMER 0, timer 5) is disabled after reset if external signal BREAK = LOW
or if the DSU is enabled via external signal DSU_EN = HIGH.
Table 432.0xnC where n selects the timer - TLATCHn - Timer n latch register
31
0
LTCV
0
r
31: 0
Latched timer counter value (LTCV): Valued latched from corresponding timer.
GR740-UM-DS, Jun 2016, Version 1.4
345
www.cobham.com/gaisler
GR740
26
Multiprocessor Interrupt Controller with extended ASMP support
26.1
Overview
The system implements an interrupt scheme where interrupt lines are routed together with the remaining AHB/APB bus signals forming an interrupt bus. The multiprocessor interrupt controller core is
attached to the AMBA bus as an APB slave and monitors the combined interrupt signals.
The interrupts generated on the interrupt bus are all forwarded to the interrupt controller. The interrupt
controller prioritizes, masks and propagates the interrupt with the highest priority. In order to support
separated ASMP configurations, the controller implements four internal interrupt controllers. Each
processor in a system can be dynamically routed to one of the internal controllers. For Symmetric
Multiprocessor (SMP) operation, several processors can be routed to the same internal interrupt controller.
Interrupt level
Interrupt acknowledge
(A)MP IRQ
CTRL
Processor 0
Processor n
Processor 1
AMBA BUS
BUS
CONTROL
SLAVE 1
SLAVE 2
Figure 45. LEON multiprocessor system with Multiprocessor Interrupt controller
26.2
Operation
26.2.1 Support for Asymmetric Multiprocessing
Asymmetric Multiprocessing support means that parts of the interrupt controller are duplicated in
order to provide safe ASMP operation. The core’s register set is duplicated on 4 KiB address boundaries. In addition to the traditional LEON multiprocessor interrupt controller register interface, the
core’s register interface will also enable the use of three new registers, one Asymmetric Multiprocessing Control Register and two Interrupt Controller Select Registers.
Software can detect if the controller has been implemented with support for ASMP by reading the
Asymmetric Multiprocessing Control register. If the field NCTRL is 0, the core was not implemented
with ASMP extensions. If the value of NCTRL is non-zero, the core has NCTRL+1 sets of registers
with additional underlying functionality. From a software view this is equivalent to having NCTRL+1
interrupt controllers available and software can configure to which interrupt controller a processor
should connect.
After system reset, all processors are connected to the first interrupt controller accessible at the core’s
base address. Software can then use the Interrupt Controller Select Registers to assign processors to
other (internal) interrupt controllers. After assignments have been made, it is recommended to freeze
GR740-UM-DS, Jun 2016, Version 1.4
346
www.cobham.com/gaisler
GR740
the contents of the select registers by writing ‘1’ to the lock bit in the Asymmetric Multiprocessing
Control Register. The lock bit can be cleared by software by writing ‘0’ to the bit
When a software driver for the interrupt controller is loaded, the driver should check the Asymmetric
Multiprocessing Control Register and Interrupt Controller Select Registers to determine to which controller the current processor is connected. After software has determined that it has been assigned to
controller n, software should only access the controller with registers at offset 0x1000 * n. Note that
the controllers are enumerated with the first controller being n = 0.
The processor specific registers (mask, force, interrupt acknowledge) can be read from all interrupt
controllers. However the processor specific mask and interrupt acknowledge registers can only be
written from the interrupt controller to which the processor is assigned. This also applies to individual
bits in the Multiprocessor Status Register. Interrupt Force bits in a processor’s Interrupt Force Register can only be cleared through the controller to which the processor is assigned. If the ICF field in the
Asymmetric Multiprocessing Control Register is set to ‘1’, all bits in all Interrupt Force Registers can
be set, but not cleared, from all controllers. If the ICF field is ‘0’ the bits in a processor’s Interrupt
Force register can only be set from the controller to which the processor is assigned.
26.2.2 Interrupt prioritization
The interrupt controller monitors interrupt 1 - 15 of the interrupt bus. When any of these lines are
asserted high, the corresponding bit in the interrupt pending register is set. The pending bits will stay
set even if the PIRQ line is de-asserted, until cleared by software or by an interrupt acknowledge from
the processor. The default behaviour for peripherals is to use pulsed interrupts (an interrupt line is
asserted for one clock cycle to signal an interrupt).
Each interrupt can be assigned to one of two levels (0 or 1) as programmed in the interrupt level register. Level 1 has higher priority than level 0. The interrupts are prioritised within each level, with interrupt 15 having the highest priority and interrupt 1 the lowest. The highest interrupt from level 1 will
be forwarded to the processor. If no unmasked pending interrupt exists on level 1, then the highest
unmasked interrupt from level 0 will be forwarded.
Interrupts are prioritised at system level, while masking and forwarding of interrupts in done for each
processor separately. Each processor in an multiprocessor system has separate interrupt mask and
force registers. When an interrupt is signalled on the interrupt bus, the interrupt controller will prioritize interrupts, perform interrupt masking for each processor according to the mask in the corresponding mask register and forward the interrupts to the processors.
GR740-UM-DS, Jun 2016, Version 1.4
347
www.cobham.com/gaisler
GR740
Priority
select
IRQ
Pending
Priority
encoder
APBI.PIRQ[15:1]
4
15
IRQO[0].IRL[3:0]
IRQ
IRQ
Force[0] mask[0]
Priority
encoder
4
IRQO[n].IRL[3:0]
IRQ
IRQ
Force[n] mask[n]
Figure 46. Interrupt controller block diagram
When a processor acknowledges the interrupt, the corresponding pending bit will automatically be
cleared. Note that in a multiprocessor system, the bit in the pending register will be cleared as soon as
one of the processors acknowledges the interrupt and interrupt broadcast functionality should be used
for interrupts that need to be propagated to all processors. Interrupt can also be forced by setting a bit
in the interrupt force register. In this case, the processor acknowledgement will clear the force bit
rather than the pending bit. After reset, the interrupt mask register is set to all zeros while the remaining control registers are undefined. Note that interrupt 15 cannot be maskable by the LEON processor
and should be used with care - most operating systems do not safely handle this interrupt.
26.2.3 Extended interrupts
The AHB/APB interrupt consist of 32 signals ([31:0]), while the interrupt controller only uses lines 1
- 15 in the nominal mode. To use the additional 16 interrupt lines (16-31), extended interrupt handling
is enabled. The interrupt lines 16 - 31 are also handled by the interrupt controller, and the interrupt
pending and mask registers have been extended to 32 bits. Since the processor only has 15 interrupt
levels (1 - 15), the extended interrupts will generate one of the regular interrupts, in this system interrupt line 10. When the interrupt is taken and acknowledged by the processor, the regular interrupt (10)
and the extended interrupt pending bits are automatically cleared. The extended interrupt acknowledge register will identify which extended interrupt that was most recently acknowledged. This register can be used by software to invoke the appropriate interrupt handler for the extended interrupts.
26.2.4 Processor status monitoring
The processor status can be monitored through the Multiprocessor Status Register. The STATUS field
in this register indicates if a processor is halted (‘1’) or running (‘0’). A halted processor can be reset
and restarted by writing a ‘1’ to its status field. After reset, all processors except processor 0 are
halted. When the system is properly initialized, processor 0 can start the remaining processors by
writing to their STATUS bits.
GR740-UM-DS, Jun 2016, Version 1.4
348
www.cobham.com/gaisler
GR740
The core has support for specifying the processor reset start address dynamically. Please see section
26.2.10 for further information.
26.2.5 Interrupt broadcasting
An incoming interrupt request that has its bit set in the Broadcast Register is propagated to the force
register of all CPUs instead of to the Pending Register. This can be used to implement a timer that
fires to all CPUs with that same IRQ.
26.2.6 Interrupt timestamping description
Interrupt timestamping is controlled via the Interrupt Timestamp Control registers. Each Interrupt
Timestamp Control register contains a field (TSTAMP) that contains the number of timestamp registers sets that the core implements. A timestamp register sets consist of one Interrupt Timestamp
Counter register, one Interrupt Timestamp Control register, one Interrupt Assertion Timestamp register and one Interrupt Acknowledge Timestamp register.
Software enables timestamping for a specific interrupt via a Interrupt Timestamp Control Register.
When the selected interrupt line is asserted, software will save the current value of the interrupt timestamp counter into the Interrupt Assertion Timestamp register and set the S1 field in the Interrupt
Timestamp Control Register. When the processor acknowledges the interrupt, the S2 field of the Interrupt Timestamp Control register will be set and the current value of the timestamp counter will be
saved in the Interrupt Acknowledge Timestamp Register. The difference between the Interrupt Assertion timestamp and the Interrupt Acknowledge timestamp is the number of system clock cycles that
was required for the processor to react to the interrupt and divert execution to the trap handler.
The core can be configured to stamp only the first occurrence of an interrupt or to continuously stamp
interrupts. The behavior is controlled via the Keep Stamp (KS) field in the Interrupt Timestamp Control Register. If KS is set, only the first assertion and acknowledge of an interrupt is stamped. Software must then clear the S1 and S2 fields for a new timestamp to be taken. If Keep Stamp is disabled
(KS field not set), the controller will update the Interrupt Assertion Timestamp Register every time
the selected interrupt line is asserted. In this case the controller will also automatically clear the S2
field and also update the Interrupt Acknowledge Timestamp register with the current value when the
interrupt is acknowledged.
For controllers with extended ASMP support, each internal controller has a dedicated set of Interrupt
timestamp registers. This means that the Interrupt Acknowledge Timestamp Register(s) on a specific
controller will only be updated if and when the processor connected to the controller acknowledges
the selected interrupt. The Interrupt Timestamp Counter is shared by all controllers and the DSU timer
is used as the time source. The same timer source can also be read via the processors’ internal upcounter described in section 6.10.3.
26.2.7 Interrupt timestamping usage guidelines
Note that KS = ‘0’ and a high interrupt rate may cause the Interrupt Assertion Timestamp register to
be updated (and the S2 field reset) before the processor has acknowledged the first occurrence of the
interrupt. When the processor then acknowledges the first occurrence, the Interrupt Acknowledge
Timestamp register will be updated and the difference between the two Timestamp registers will not
show how long it took the processor to react to the first interrupt request. If the interrupt frequency is
expected to be high it is recommended to keep the first stamp (KS field set to ‘1’) in order to get reliable measurements. KS = ‘0’ should not be used in systems that include cores that use level interrupts,
the timestamp logic will register each cycle that the interrupt line is asserted as an interrupt.
In order to measure the full interrupt handling latency in a system, software should also read the current value of the Interrupt Timestamp Counter when entering the interrupt handler. In the typical case,
a software driver’s interrupt handler reads a status register and then determines the action to take.
Adding a read of the timestamp counter before this status register read can give an accurate view of
the latency during interrupt handling.
GR740-UM-DS, Jun 2016, Version 1.4
349
www.cobham.com/gaisler
GR740
The core listens to the system interrupt vector when reacting to interrupt line assertions. This means
that the Interrupt Assertion Timestamp Register(s) will not be updated if software writes directly to
the pending or force registers. To measure the time required to serve a forced interrupt, read the value
of the Interrupt Timestamp counter before forcing the interrupt and then read the Interrupt Acknowledge Timestamp and Interrupt Timestamp counter when the processor has reacted to the interrupt.
26.2.8 Watchdog
The core can be configured to assert a bit in the controller’s Interrupt Pending Register when an external watchdog signal is asserted. This functionality can be used to implement a sort of soft watchdog
for one or several processor cores. The controller’s Watchdog Control Register contains a field that
shows the number of external watchdog inputs supported and fields for configuring which watchdog
inputs that should be able to assert a bit in the Interrupt Pending Register.
The on-chip watchdog inputs are connected to the tick outputs from timer 4 on general purpose timer
units 1 - 4. This means that watchdog input n will be high for one cycle when timer 4 on general purpose timer unit n underflows.
Each internal controller has a dedicated Watchdog Control register. Assertion of a watchdog input will
only affect the pending register on the internal interrupt controllers that have enabled the watchdog
input in their Watchdog Control Register.
26.2.9 Interrupt (re)map functionality
The interrupt controller has functionality to allow dynamic remapping between bus interrupt lines and
interrupt controller interrupt lines. Switch-logic has been placed on the incoming interrupt vector
from the AMBA bus before the IRQ pending register. The Interrupt map registers are available starting at offset 0x300 from the interrupt controller's base address.
The interrupt map registers contain one field for each bus interrupt line in the system. The value
within this field determines to which interrupt controller line the bus interrupt line is connected. In
case several bus interrupt lines are mapped to the same controller interrupt line (several fields in the
Interrupt map registers have the same value) then the bus interrupt lines will be OR:ed together.
Note that if bus interrupt line X is remapped to controller interrupt line Y then bit Y of the pending
register will be set when a peripheral asserts interrupt X. Remapping interrupt lines via the Interrupt
map registers has the same effect as changing the interrupt assignments in the physical design.
26.2.10 Dynamic processor reset start address
The core has registers that are used to dynamically specify the reset start address for each CPU in the
system. The processor start address registers are available, one for each processor, starting at register
offset 0x200. The reset value for all Processor Reset Start Address registers is 0xC0000000 (system
PROM area). If software wishes to boot a processor from a different address, the processor’s start
address register should be written (start address must be aligned on a 4 KiB address boundary) and the
processor should then be enabled through Processor boot register.
The Processor Reset Start Address registers are visible and writable from the register space of all
internal controllers.
26.3
Registers
The core is controlled through registers mapped into APB address space. The register set for internal
controller n is accessed at offset 0x1000*n.
Table 433.Interrupt Controller registers
APB address offset
Register
0x000
Interrupt level register
GR740-UM-DS, Jun 2016, Version 1.4
350
www.cobham.com/gaisler
GR740
Table 433.Interrupt Controller registers
APB address offset
Register
0x004
Interrupt pending register
0x008
Interrupt force register (NCPU = 0)
0x00C
Interrupt clear register
0x010
Multiprocessor status register
0x014
Broadcast register
0x018
Reserved
0x01C
Watchdog control register
0x020
Asymmetric multiprocessing control register
0x024
Interrupt controller select register for processor 0 - 3
0x028 - 0x03C
Reserved
0x040
Processor 0 interrupt mask register
0x044
Processor 1 interrupt mask register
0x048
Processor 2 interrupt mask register
0x04C
Processor 3 interrupt mask register
0x050 - 0x07C
Reserved
0x080
Processor 0 interrupt force register
0x084
Processor 1 interrupt force register
0x088
Processor 2 interrupt force register
0x08C
Processor 3 interrupt force register
0x090 - 0xBC
Reserved
0x0C0
Processor 0 extended interrupt acknowledge register
0x0C4
Processor 1 extended interrupt acknowledge register
0x0C8
Processor 2 extended interrupt acknowledge register
0x0CC
Processor 3 extended interrupt acknowledge register
0x0D0 - 0x0FC
Reserved
0x100
Interrupt timestamp counter register
0x104
Interrupt timestamp 0 control register
0x108
Interrupt assertion timestamp 0 register
0x10C
Interrupt acknowledge timestamp 0 register
0x110
Interrupt timestamp counter register
0x114
Interrupt timestamp 1 control register
0x118
Interrupt assertion timestamp 1 register
0x11C
Interrupt acknowledge timestamp 1 register
0x120 - 0x1FC
Reserved
0x200
Processor 0 reset start address register
0x204
Processor 1 reset start address register
0x208
Processor 2 reset start address register
0x20C
Processor 3 reset start address register
0x210 - 0x23C
Reserved
0x240
Processor boot register
0x300
Interrupt map register 0
0x304
Interrupt map register 1
0x308
Interrupt map register 2
0x30C
Interrupt map register 3
GR740-UM-DS, Jun 2016, Version 1.4
351
www.cobham.com/gaisler
GR740
Table 433.Interrupt Controller registers
APB address offset
Register
0x310
Interrupt map register 4
0x314
Interrupt map register 5
0x318
Interrupt map register 6
0x31C
Interrupt map register 7
26.3.1 Interrupt level register
Table 434.0x000 - ILEVEL - Interrupt level register
31
16 15
31:16
1
0
RESERVED
IL[15:1]
R
0
NR
0
r
rw
0
Reserved
15:1
Interrupt Level n (IL[n]) - Interrupt level for interrupt n
0
Reserved
26.3.2 Interrupt pending register
Table 435.0x004 - IPEND - Interrupt pending register
31
16 15
EIP[31:16]
1
0
IP[15:1]
R
0
0
0
rw
rw
r
31:16
Extended Interrupt Pending n (EIP[n]) - Interrupt pending for interrupt n
15:1
Interrupt Pending n (IP[n]) - Interrupt pending for interrupt n
0
Reserved
26.3.3 Interrupt force register
Table 436.0x008 - IFORCE0 - Interrupt force register for processor 0
31
16 15
RESERVED
1
0
IF[15:1]
R
0
0
0
r
rw
r
31:16
Reserved
15:1
Interrupt Force n (IF[n]) - Force interrupt nr n.
0
Reserved
GR740-UM-DS, Jun 2016, Version 1.4
352
www.cobham.com/gaisler
GR740
26.3.4 Interrupt clear register
Table 437.0x00C - ICLEAR - Interrupt clear register
31
16 15
1
EIC[31:16]
IC[15:1]
w
w
0
R
0
31:16
r
Extended Interrupt Clear n (EIC[n]) - Writing ‘1’ to EIC[n] will clear interrupt n
15:1
Interrupt Clear n (IC[n]) - Writing ‘1’ to IC[n] will clear interrupt n
0
Reserved
GR740-UM-DS, Jun 2016, Version 1.4
353
www.cobham.com/gaisler
GR740
26.3.5 Multiprocessor status register
Table 438.0x010 - MPSTAT - Multiprocessor status register
31
28 27 26
20 19
16 15
4
3
0
NCPU
BA
RESERVED
EIRQ
RESERVED
STATUS
3
1
0
0xA
0
*
r
r
r
r
r
rw
31: 28
Number of CPUs (NCPU) - Number of CPUs in the system - 1
27
Broadcast Available (BA) - Set to ‘1’ if NCPU > 0.
26: 20
RESERVED
19: 16
Extended IRQ (EIRQ) - Interrupt number (1 - 15) used for extended interrupts. Fixed to 0 if
extended interrupts are disabled.
15: 4
RESERVED
3: 0
Power-down status of CPU[n] (STATUS[n]) - ‘1’ = power-down, ‘0’ = running.
Write STATUS[n] with ‘1’ to start processor n.
The reset value for this field is 0xE if the external signal BREAK is LOW. Otherwise the reset value
is 0xF.
26.3.6 Broadcast register
Table 439.0x014 - BRDCST - Broadcast register
31
16 15
1
RESERVED
0
BM15:1]
R
0
0
0
r
rw
r
31: 16
RESERVED
15: 1
Broadcast Mask n (BM[n]) - If BM[n] = ‘1’ then interrupt n is broadcasted (written to the Force Register of all CPUs), otherwise standard semantic applies (Pending register)
0
RESERVED
26.3.7 Watchdog control register
Table 440.0x01C - WDOGCTRL - Watchdog control register
31
27 26
20 19
16 15
4
3
0
NWDOG
Reserved
WDOGIRQ
RESERVED
4
0
NR
0
0
r
r
rw
r
rw
31: 27
WDOGMSK
Number of watchdog inputs (NWDOG) - Number of watchdog inputs that the core supports.
26: 20
RESERVED
19: 16
Watchdog interrupt (WDOGIRQ) - Selects the bit in the pending register to set when any line watchdog line selected by the WDOGMSK field is asserted.
15: 4
RESERVED
3: 0
Watchdog Mask n (WDOGMSK[n]) - If WDOGMSK[n] = ‘1’ then the assertion of watchdog input
n will lead to the bit selected by the WDOGIRQ field being set in the controller’s Interrupt Pending
Register.
Bit n in the watchdog input is connected to GPTIMER (n+1)’s, timer 4 tick output.
GR740-UM-DS, Jun 2016, Version 1.4
354
www.cobham.com/gaisler
GR740
26.3.8 Asymmetric multiprocessing control register
Table 441.0x020 - ASMPCTRL - Asymmetric multiprocessing control register
31
28 27
2
1
0
NCTRL
RESERVED
ICF L
0x3
0
0
r
r
rw rw
0
31: 28
Number of internal controllers (NCTRL) - NCTRL + 1 is the number of internal interrupt controllers
available.
27: 2
RESERVED
1
Inter-controller Force (ICF) - If this bit is set to ‘1’ all Interrupt Force Registers can be set from any
internal controller. If this bit is ‘0’, a processor’s Interrupt Force Register can only be set from the
controller to which the processor is connected. Bits in an Interrupt Force Register can only be
cleared by the controller or by writing the Interrupt Force Clear field on the controller to which the
processor is connected.
0
Lock (L) - If this bit is written to ‘1’, the contents of the Interrupt Controller Select registers is frozen..
26.3.9 Interrupt controller select register
Table 442.0x024 - ICSELR - Interrupt controller select register
31
28 27
ICSEL0
24 23
ICSEL1
20 19
ICSEL2
16 15
0
ICSEL3
RESERVED
0
0
0
0
0
rw
rw
rw
rw
r
31: 16
Interrupt controller select for processor n (ICSEL[n]) - The nibble ICSEL[n] selects the (internal)
interrupt controller to connect to processor n.
15: 0
RESERVED
26.3.10 Processor interrupt mask registers
Table 443.0x040, 0x044, 0x048, 0x04C - PIMASK0-3 - Processor 0, 1, 2, 3 interrupt mask register
31
16 15
EIM[31:16]
1
0
IM15:1]
R
0
0
R
rw
rw
r
31: 16
Extended Interrupt Mask n (EIC[n]) - Interrupt mask for extended interrupts
15: 1
Interrupt Mask n (IM[n]) - If IM[n] = ‘0’ then interrupt n is masked, otherwise it is enabled.
0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
355
www.cobham.com/gaisler
GR740
26.3.11 Processor interrupt force registers
Table 444.0x080, 0x084, 0x088, 0x08C - PIFORCE0-3 - Processor 0, 1, 2, 3 interrupt force register
31
17 16 15
FC[15:1]
R
0
wc
1
0
IF15:1]
R
0
0
0
r
rw*
r
31: 17
Interrupt Force Clear n (IFC[n]) - Interrupt force clear for interrupt n. Bits can be cleared by writing
’1’ to their position, writes of ’0’ have no effect.
16
RESERVED
15: 1
Interrupt Force n (IF[n]) - Force interrupt nr n. Bits can be asserted by writing ’1’ to their position.
Writes of ’0’ have no effect.
0
RESERVED
26.3.12 Extended interrupt acknowledge registers
Table 445.0x0C0, 0x0C4, 0x0C8, 0x0CC - PEXTACK0-3 - Processor 0, 1, 2, 3 extended interrupt acknowledge register
31
5
4
0
RESERVED
EID[4:0]
0
0
r
r
31: 5
RESERVED
4: 0
Extended interrupt ID (EID) - ID (16-31) of the most recent acknowledged extended interrupt.
26.3.13 Interrupt timestamp counter register
Table 446.0x100, 0x110 - ITCNT - Interrupt timestamp counter register
31
0
TCNT
0
r
31: 0
Timestamp Counter (TCNT) - Current value of timestamp counter. The time value is taken from the
DSU timer and is the same timetag value available in the LEON4 up-counters and in the trace buffers. The counter will only increment if the DSU and trace buffers are enabled or if at least one of the
processor up-counters is enabled.
GR740-UM-DS, Jun 2016, Version 1.4
356
www.cobham.com/gaisler
GR740
26.3.14 Timestamp control registers
Table 447.0x1n4 - ITSTMPCn - Interrupt timestamp n control register
31
27 26 25 24
TSTAMP
0x2
r
S1 S2
0
6
RESERVED
0
wc wc
5
KS
4
0
TSISEL
0
0
0
r
rw
rw
31: 27
Number of timestamp register sets (TSTAMP) - The number of available timestamp register sets.
26
Assertion Stamped (S1) - Set to ‘1’ when the assertion of the selected line has received a timestamp.
This bit is cleared by writing ‘1’ to its position. Writes of ‘0’ have no effect.
25
Acknowledge Stamped (S2) - Set to ‘1’ when the processor acknowledge of the selected interrupt
has received a timestamp. This bit can be cleared by writing ‘1’ to this position, writes of ‘0’ have no
effect. This bit can also be cleared automatically by the core, see description of the KS field below.
24: 6
RESERVED
5
Keep Stamp (KS) - If this bit is set to ‘1’ the core will keep the first stamp value for the first interrupt
until the S1 and S2 fields are cleared by software. If this bit is set to ‘0’ the core will time stamp the
most recent interrupt. This also has the effect that the core will automatically clear the S2 field whenever the selected interrupt line is asserted and thereby also stamp the next acknowledge of the interrupt.
4: 0
Timestamp Interrupt Select (TSISEL) - This field selects the interrupt line (0 - 31) to timestamp.
26.3.15 Interrupt assertion timestamp n register
Table 448.0x1n8 - ITSTMPASn - Interrupt Assertion Timestamp n register
31
0
TASSERTION
0
r
31: 0
Timestamp of Assertion (TASSERTION) - The current Timestamp Counter value is saved in this
register when timestamping is enabled and the interrupt line selected by TSISEL is asserted.
The time value used for stamping is the DSU timer, which is also available as the processor internal.
up-counter.
GR740-UM-DS, Jun 2016, Version 1.4
357
www.cobham.com/gaisler
GR740
26.3.16 Interrupt acknowledge timestamp register
Table 449.0x1nC - ITSTMPACn - Interrupt Acknowledge Timestamp n register
31
0
TACKNOWLEDGE
0
r
31: 0
Timestamp of Acknowledge (TACKNOWLEDGE) - The current Timestamp Counter value is saved
in this register when timestamping is enabled, the Acknowledge Stamped (S2) field is ‘0’, and the
interrupt selected by TSISEL is acknowledged by a processor connected to the interrupt controller.
The time value used for stamping is the DSU timer, which is also available as the processor internal.
up-counter.
26.3.17 Processor reset start address registers
Table 450.0x200 + n*4 - PRSTADDRn - Processor n reset start address register
31
12 11
0
RSTADDR
RESERVED
0xC0000
0
rw
r
31: 12
Processor reset start address (RSTADDR) - The Processor start address register at offset 0x200 +
4*n specifies the reset start address for processor n.
Note that a processor must be reset before the new reset start address is valid. It is not possible to
update the value in this register and then to correctly boot from the new address by only waking a
processor via the Multiprocessor status register. Instead use the Processor boot register to boot or
reset the processor.
11: 0
RESERVED
26.3.18 Processor boot register
Table 451.0x240 - PBOOT - Processor boot register
31
20 19
RESERVED
16 15
4
3
0
RESET[n]
RESERVED
BOOT[n]
0
0
0
0
r
rw*
r
rw*
31: 20
RESERVED
19: 16
Processor reset (RESET): Writing bit n of this field to ‘1’ will reset, but not start, processor n. When
the processor has been reset the bit will be reset to ‘0’. A processor can only be reset if it is currently
idle (in power-down, error or debug mode), if a processor is running then the write to its bit in this
field will be ignored. Multiple bits in this register may be set with one write but the register can only
be written when all bits are zero.
15: 4
RESERVED
3: 0
Processor boot (BOOT): Writing bit n of this field to ‘1’ will reset and start processor n. When the
processor has been booted the bit will be reset to ‘0’. A processor can only be started if it is currently
idle (in power-down, error or debug mode), if a processor is running then the write to its bit in this
field will be ignored. Multiple bits in this register may be set with one write but the register can only
be written when all bits are zero.
GR740-UM-DS, Jun 2016, Version 1.4
358
www.cobham.com/gaisler
GR740
26.3.19 Interrupt map registers
Table 452.0x300 + 4*n - IRQMAPn - Interrupt map register n
31
24 23
16 15
8
7
0
IRQMAP[n*4]
IRQMAP[n*4+1]
IRQMAP[n*4+2]
IRQMAP[n*4+3]
n*4
n*4+1
n*4+2
n*4+3
rw
rw
rw
rw
31: 0
Interrupt map (IRQMAP) - The Interrupt map register at offset 0x300 + 4*n specifies the mapping
for interrupt lines 4*n to 4*n+3.
The bus interrupt line 4*n+x will be mapped to the interrupt controller interrupt line specified by the
value of IRQMAP[n*4+x].
The interrupt map registers are only accessible from the first interrupt controller (starting at offset
0x300).
GR740-UM-DS, Jun 2016, Version 1.4
359
www.cobham.com/gaisler
GR740
27
General Purpose I/O Ports
27.1
Overview
Each bit in the general purpose input output port can be individually set to input or output, and can
optionally generate an interrupt. For interrupt generation, the input can be filtered for polarity and
level/edge detection.
Note that some GPIO pins are used as bootstrap pins, see section 3.1 for further information.
The design has two general purpose I/O port peripherals. The first one, GRGPIO0, with base address
0xFF902000 is connected to the 16 external GPIO signals. The second port, GRGPIO1, with base
address 0xFFA08000 is connected to a set 22 of shared pins. Pin sharing is further described in chapter 3. The only configuration difference between the two peripherals are the signals connected and the
number of connected signals.
The figure 47 shows a diagram for one I/O line.
Alternate enable
(Not used)
Direction
D
Q
Alternate
Output
(Not used)
Value
Output
Value
Input
Value
Input
Value
D
PAD
Q
Input D
Q
Value
Q
D
(Not used)
Figure 47. General Purpose I/O Port diagram
27.2
Operation
The I/O ports are implemented as bi-directional buffers with programmable output enable. The input
from each buffer is synchronized by two flip-flops in series to remove potential meta-stability. The
synchronized values can be read-out from the I/O port data register. The output enable is controlled by
the I/O port direction register. A ‘1’ in a bit position will enable the output buffer for the corresponding I/O line. The output value driven is taken from the I/O port output register.
The core supports dynamic mapping of interrupts, each I/O line can be mapped using the Interrupt
map register(s) to an interrupt line starting at interrupt 16.
Interrupt generation is controlled by three registers: interrupt mask, polarity and edge registers. To
enable an interrupt, the corresponding bit in the interrupt mask register must be set. If the edge register is ‘0’, the interrupt is treated as level sensitive. If the polarity register is ‘0’, the interrupt is active
low. If the polarity register is ‘1’, the interrupt is active high. If the edge register is ‘1’, the interrupt is
edge-triggered. The polarity register then selects between rising edge (‘1’) or falling edge (‘0’).
A GPIO pin can also be toggled when a pulse is detected on an internal signal. This is enabled via the
Pulse register in the core. This functionality is only supported for the first GPIO port, GRGPIO0.
GR740-UM-DS, Jun 2016, Version 1.4
360
www.cobham.com/gaisler
GR740
27.3
Registers
The core is programmed through registers mapped into APB address space.
Table 453. General Purpose I/O Port registers
APB address offset
Register
0x00
I/O port data register
0x04
I/O port output register
0x08
I/O port direction register
0x0C
Interrupt mask register
0x10
Interrupt polarity register
0x14
Interrupt edge register
0x18
Reserved
0x1C
Capability register
0x20
Interrupt map register 0
0x24
Interrupt map register 1
0x28
Interrupt map register 2
0x2C
Interrupt map register 3
0x30 - 0x3C
Reserved
0x40
Interrupt available register
0x44
Interrupt flag register
0x48
Reserved
0x4C
Pulse register
0x50
Reserved
0x54
I/O port output register, logical-OR
0x58
I/O port direction register, logical-OR
0x5C
Interrupt mask register, logical-OR
0x60
Reserved
0x64
I/O port output register, logical-AND
0x68
I/O port direction register, logical-AND
0x6C
Interrupt mask register, logical-AND
0x70
Reserved
0x74
I/O port output register, logical-XOR
0x78
I/O port direction register, logical-XOR
0x7C
Interrupt mask register, logical-XOR
GR740-UM-DS, Jun 2016, Version 1.4
361
www.cobham.com/gaisler
GR740
27.3.1 I/O port data register
Table 454.0x00 - DATA - I/O port data register
31
nlin
31: nlin
nlin-1: 0
nlin-1
0
RESERVED
DATA
0
*
r
r
RESERVED
I/O port input value (DATA) - Data value read from GPIO lines
Note: This field has range 15:0 for the first GPIO port, GRGPIO0, and range 21:0 for the second
GPIO port, GRGPIO1.
Reset value depends on state of external signals.
27.3.2 I/O port output register
Table 455.0x04 - OUTPUT - I/O port output register
31
nlin
nlin-1
RESERVED
31: nlin
nlin-1: 0
0
DATA
0
0
r
rw
RESERVED
I/O port output value (DATA) - Output value for GPIO lines
Note: This field has range 15:0 for the first GPIO port, GRGPIO0, and range 21:0 for the second
GPIO port, GRGPIO1.
27.3.3 I/O port direction register
Table 456.0x08 - DIRECTION - I/O port direction register
31
nlin
RESERVED
31: nlin
nlin-1: 0
nlin-1
0
DIR
0
0
r
rw
RESERVED
I/O port direction value (DIR) - 0=output disabled, 1=output enabled
Note: This field has range 15:0 for the first GPIO port, GRGPIO0, and range 21:0 for the second
GPIO port, GRGPIO1.
GR740-UM-DS, Jun 2016, Version 1.4
362
www.cobham.com/gaisler
GR740
27.3.4 Interrupt mask register
Table 457.0x0C - IMASK - Interrupt mask register
31
16 15
RESERVED
0
MASK
0
0
r
rw
31: 16
RESERVED
15: 0
Interrupt mask (MASK) - 0=interrupt masked, 1=intrrupt enabled
27.3.5 Interrupt polarity register
Table 458.0x10 - IPOL - Interrupt polarity register
31
16 15
0
RESERVED
POL
0
NR
r
rw
31: 16
RESERVED
15: 0
Interrupt polarity (POL) - 0=low/falling, 1=high/rising
27.3.6 Interrupt edge register
Table 459.0x14- IEDGE - Interrupt edge register
31
16 15
0
RESERVED
EDGE
0
NR
r
rw
31: 16
RESERVED
15: 0
Interrupt edge (EDGE) - 0=level, 1=edge
GR740-UM-DS, Jun 2016, Version 1.4
363
www.cobham.com/gaisler
GR740
27.3.7 Capability register
Table 460.0x1C- CAP - Capability register
31
19 18 17 16 15
RESERVED
13 12
PU IER IFL RESERVED
8
7
5
4
0
IRQGEN
RESERVED
NLINES
0
*
0
1
0
0x4
0
*
r
r
r
r
r
r
r
r
31: 19
RESERVED
18
Pulse register implemented (PU) - If this field is ‘1’ then the core implements the Pulse register. Set
to ‘1’ for the first GPIO controller in this implementation.
17
Input Enable register implemented (IER) - If this field is ‘1’ then the core implements the Input
enable register. Set to ‘0’ in this implementation.
16
Interrupt flag register implemented (IFL) - If this field is ‘1’ then the core implements the Interrrupt
available and Interrupt flag registers (registers at offsets 0x40 and 0x44). Set to ‘1’ in this implementation.
12: 8
Interrupt generation setting (IRQGEN) - Set to 4 to signify that the core has Interrupt map registers
allowing software to dynamically map which lines that should drive interrupt lines 16 to 19..
7: 5
RESERVED
4: 0
Number of pins in GPIO port - 1 (NLINES)
The field has value 15 for the first GPIO port and 21 for the second GPIO port.
27.3.8 Interrupt map registers
Table 461.0x20+4*n- IRQMAPRn - Interrupt map register n, where n = 0 .. 3
31
29 28
24 23
21 20
16 15
13 12
8
7
6
4
0
RESERVED
IRQMAP[i]
RESERVED
IRQMAP[i+1]
RESERVED
IRQMAP[i+2]
RESERVED
IRQMAP[i+3]
0
4*n+i
0
4*n+i+1
0
4*n+i+2
0
4*n+i+3
r
rw
r
rw
r
rw
r
rw
31: 0
IRQMAP[i] : The field IRQMAP[i] determines to which interrupt I/O line i is connected. If IRQMAP[i] is set to x, IO[i] will drive interrupt 16+x. Several I/O can be mapped to the same interrupt.
An I/O line’s interrupt generation must be enabled in the Interrupt mask register in order for the I/O
line to drive the interrupt specified by the IRQMAP field.
27.3.9 Interrupt available register
Table 462.0x40 - IAVAIL - Interrupt available register
31
0
IMASK
0xFFFF
r
31: 0
Interrupt available bit field (IMASK) - If IMASK[n] is 1 then GPIO line n can generate interrupts.
This field is read-only has has value 0xFFFF for both GPIO ports. This means that lines 15:0 can be
used for interrupt generation.
GR740-UM-DS, Jun 2016, Version 1.4
364
www.cobham.com/gaisler
GR740
27.3.10 Interrupt flag register
Table 463.0x44 - IFLAG - Interrupt flag register
31
16 15
31: 16
15: 0
0
RESERVED
IFLAG
0
0
r
r
RESERVED
IFLAG : If IFLAG[n] is set to ‘1’ then GPIO line n has generated an interrupt. Self-clearing.
The IFLAG register bit(s) will only be set while the interrupt(s) is asserted. In this implementation
this register is only practically usable when using level interrupts. For edge interrupts, an IFLAG
register bit will only be asserted during one cycle per interrupt assertion. For software compatibility
please note that the behaviour of the IFLAG register bit may be different for other LEON systems
where the IFLAG register bits may be sticky bits that are set to one when an interrupt has been generated and remain set until cleared by writing to this register.
27.3.11 Pulse register
Table 464.0x4C - PULSE - Pulse register
31
16 15
RESERVED
0
PULSE
0
0
r
rw
31: 16
RESERVED
15: 0
PULSE : If PULSE[n] is set to ‘1’ then OUTPUT register bit n will be inverted on the following
events:
Position 0: GPTIMER 0 tick 0
Position 1: GPTIMER 0 tick 1
Position 2: GPTIMER 0 tick 2
Position 3: GPTIMER 0 tick 3
Position 4: GPTIMER 0 tick 4
Position 5: GRSPWTDP CTICK -A pulse is generated when SpaceWire Time-Code is transmitted
when TDP controller is acting as initiator. A pulse is also generated when a diagnostic SpaceWire
Time-Code is generated when TDP controller is acting as target.
Position 6: GRSPWTDP JTICK - The incoming SpaceWire Time-Code provides an output pulse
when the TDP controller is acting as target, this output is used to visualize the jitter in incoming
SpaceWire Time-Codes
Position 7 to 10: GRSPWTDP external datation pulse 0 to 3
Position 11 to 15: GPTIMER latch disabled for timer units 0 to 4.
This register is only available for the first GPIO port, GRGPIO 0.
See also section 5.9.
27.3.12 Logical-OR/AND-XOR registers
Table 465.0x54-0x7C - LOR/LAND/LXOR - Logical-OR/AND/XOR registers
31
nlin
RESERVED
nlin-1
0
DATA
0
0
r
w*
GR740-UM-DS, Jun 2016, Version 1.4
365
www.cobham.com/gaisler
GR740
Table 465.0x54-0x7C - LOR/LAND/LXOR - Logical-OR/AND/XOR registers
31: nlin
RESERVED
nlin-1: 0
The logical-OR/AND/XOR registers will update the corresponding register (see table 453) according to:
New value = <Old value> logical-op <Write data>
Note: This field has range 15:0 for the first GPIO port, GRGPIO0, and range 21:0 for the second
GPIO port, GRGPIO1.
GR740-UM-DS, Jun 2016, Version 1.4
366
www.cobham.com/gaisler
GR740
28
UART Serial Interfaces
28.1
Overview
Two UART interfaces are provided for serial communications. Each UART supports data frames with
8 data bits, one optional parity bit and one stop bit. To generate the bit-rate, each UART has a programmable 20-bit clock divider. Two FIFOs are used for data transfer between the APB bus and
UART. Hardware flow-control is supported through RTSN/CTSN hand-shake signals.
CTSN
Baud-rate
generator
RXD
8*bitclk
Receiver shift register
Serial port
Controller
RTSN
TXD
Transmitter shift register
Receiver FIFO
Transmitter FIFO
APB
Figure 48. Block diagram
28.2
Operation
28.2.1 Transmitter operation
The transmitter is enabled through the TE bit in the UART control register. Data that is to be transferred is stored in the 16-byte FIFO by writing to the data register. When ready to transmit, data is
transferred from the transmitter FIFO to the transmitter shift register and converted to a serial stream
on the transmitter serial output pin. The core automatically sends a start bit followed by eight data
bits, an optional parity bit, and one stop bit (figure 49). The least significant bit of the data is sent first.
Data frame, no parity:
Start D0
D1
D2
D3
D4
D5
D6
D7 Stop
Data frame with parity:
Start D0
D1
D2
D3
D4
D5
D6
D7 Parity Stop
Figure 49. UART data frames
Following the transmission of the stop bit, if a new character is not available in the transmitter FIFO,
the transmitter serial data output remains high and the transmitter shift register empty bit (TS) will be
set in the UART status register. Transmission resumes and the TS is cleared when a new character is
GR740-UM-DS, Jun 2016, Version 1.4
367
www.cobham.com/gaisler
GR740
loaded into the transmitter FIFO. When the FIFO is empty the TE bit is set in the status register. If the
transmitter is disabled, it will immediately stop any active transmissions including the character currently being shifted out from the transmitter shift register. The transmitter holding register may not be
loaded when the transmitter is disabled or when the FIFO is full. If this is done, data might be overwritten and one or more frames are lost.
The TF status bit (not to be confused with the TF control bit) is set if the transmitter FIFO is currently
full and the TH bit is set as long as the FIFO is less than half-full (less than half of entries in the FIFO
contain data). The TF control bit enables FIFO interrupts when set. The status register also contains a
counter (TCNT) showing the current number of data entries in the FIFO.
When flow control is enabled, the CTSN input must be low in order for the character to be transmitted. If it is deasserted in the middle of a transmission, the character in the shift register is transmitted
and the transmitter serial output then remains inactive until CTSN is asserted again. If the CTSN is
connected to a receivers RTSN, overrun can effectively be prevented.
28.2.2 Receiver operation
The receiver is enabled for data reception through the receiver enable (RE) bit in the UART control
register. The receiver looks for a high to low transition of a start bit on the receiver serial data input
pin. If a transition is detected, the state of the serial input is sampled a half bit clocks later. If the serial
input is sampled high the start bit is invalid and the search for a valid start bit continues. If the serial
input is still low, a valid start bit is assumed and the receiver continues to sample the serial input at
one bit time intervals (at the theoretical centre of the bit) until the proper number of data bits and the
parity bit have been assembled and one stop bit has been detected. The serial input is shifted through
an 8-bit shift register where all bits have to have the same value before the new value is taken into
account, effectively forming a low-pass filter with a cut-off frequency of 1/8 system clock.
The receiver also has a FIFO which is identical to the one in the transmitter.
During reception, the least significant bit is received first. The data is then transferred to the receiver
FIFO and the data ready (DR) bit is set in the UART status register as soon as the FIFO contains at
least one data frame. The parity, framing and overrun error bits are set at the received byte boundary,
at the same time as the receiver ready bit is set. The data frame is not stored in the FIFO if an error is
detected. Also, the new error status bits are or:ed with the old values before they are stored into the
status register. Thus, they are not cleared until written to with zeros from the AMBA APB bus. If both
the receiver FIFO and shift registers are full when a new start bit is detected, then the character held in
the receiver shift register will be lost and the overrun bit will be set in the UART status register. A
break received (BR) is indicated when a BREAK has been received, which is a framing error with all
data received being zero.
If flow control is enabled, then the RTSN will be negated (high) when a valid start bit is detected and
the receiver FIFO is full. When the holding register is read, the RTSN will automatically be reasserted
again.
The RF status bit (not to be confused with the RF control bit) is set when the receiver FIFO is full.
The RH status bit is set when the receiver FIFO is half-full (at least half of the entries in the FIFO contain data frames). The RF control bit enables receiver FIFO interrupts when set. A RCNT field is also
available showing the current number of data frames in the FIFO.
28.3
Baud-rate generation
Each UART contains a 20-bit down-counting scaler to generate the desired baud-rate. The scaler is
clocked by the system clock and generates a UART tick each time it underflows. It is reloaded with
the value of the UART scaler reload register after each underflow. The resulting UART tick frequency
should be 8 times the desired baud-rate. If the EC bit is set, the ticks will be generated with the same
frequency as the external clock input instead of at the scaler underflow rate. In this case, the frequency
of external clock must be less than half the frequency of the system clock.
GR740-UM-DS, Jun 2016, Version 1.4
368
www.cobham.com/gaisler
GR740
28.4
Loop back mode
If the LB bit in the UART control register is set, the UART will be in loop back mode. In this mode,
the transmitter output is internally connected to the receiver input and the RTSN is connected to the
CTSN. It is then possible to perform loop back tests to verify operation of receiver, transmitter and
associated software routines. In this mode, the outputs remain in the inactive state, in order to avoid
sending out data.
28.5
FIFO debug mode
FIFO debug mode is entered by setting the debug mode bit in the control register. In this mode it is
possible to read the transmitter FIFO and write the receiver FIFO through the FIFO debug register.
The transmitter output is held inactive when in debug mode. A write to the receiver FIFO generates an
interrupt if receiver interrupts are enabled.
28.6
Interrupt generation
Two different kinds of interrupts are available: normal interrupts and FIFO interrupts. For the transmitter, normal interrupts are generated when transmitter interrupts are enabled (TI), the transmitter is
enabled and the transmitter FIFO goes from containing data to being empty. FIFO interrupts are generated when the FIFO interrupts are enabled (TF), transmissions are enabled (TE) and the UART is
less than half-full (that is, whenever the TH status bit is set). This is a level interrupt and the interrupt
signal is continuously driven high as long as the condition prevails. The receiver interrupts work in
the same way. Normal interrupts are generated in the same manner as for the holding register. FIFO
interrupts are generated when receiver FIFO interrupts are enabled, the receiver is enabled and the
FIFO is half-full. The interrupt signal is continuously driven high as long as the receiver FIFO is halffull (at least half of the entries contain data frames).
To reduce interrupt occurrence a delayed receiver interrupt is available. It is enabled using the delayed
interrupt enable (DI) bit. When enabled a timer is started each time a character is received and an
interrupt is only generated if another character has not been received within 4 character + 4 bit times.
If receiver FIFO interrupts are enabled a pending character interrupt will be cleared when the FIFO
interrupt is active since the character causing the pending irq state is already in the FIFO and is
noticed by the driver through the FIFO interrupt.
There is also a separate interrupt for break characters. When enabled an interrupt will always be generated immediately when a break character is received even when delayed receiver interrupts are
enabled. When break interrupts are disabled no interrupt will be generated for break characters when
delayed interrupts are enabled.
When delayed interrupts are disabled the behavior is the same for the break interrupt bit except that an
interrupt will be generated for break characters if receiver interrupt enable is set even if break interrupt is disabled.
An interrupt can also be enabled for the transmitter shift register. When enabled the core will generate
an interrupt each time the shift register goes from a non-empty to an empty state.
GR740-UM-DS, Jun 2016, Version 1.4
369
www.cobham.com/gaisler
GR740
28.7
Registers
The core is controlled through registers mapped into APB address space.
Table 466.UART registers
APB address offset
Register
0x0
UART Data register
0x4
UART Status register
0x8
UART Control register
0xC
UART Scaler register
0x10
UART FIFO debug register
28.7.1 UART Data Register
Table 467. UART data register
31
8
7
0
RESERVED
DATA
N/R
rw*
7: 0
Receiver holding register or FIFO (read access)
7: 0
Transmitter holding register or FIFO (write access)
28.7.2 UART Status Register
Table 468. UART status register
31
26 25
20 19
11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
RCNT
TCNT
RF TF RH TH FE PE OV BR TE TS DR
0
0
0
0
0
0
0
0
1
1
0
r
r
r
r
r
r
rw rw rw rw
r
r
r
0
0
31: 26
Receiver FIFO count (RCNT) - shows the number of data frames in the receiver FIFO.
25: 20
Transmitter FIFO count (TCNT) - shows the number of data frames in the transmitter FIFO.
10
Receiver FIFO full (RF) - indicates that the Receiver FIFO is full.
9
Transmitter FIFO full (TF) - indicates that the Transmitter FIFO is full.
8
Receiver FIFO half-full (RH) -indicates that at least half of the FIFO is holding data.
7
Transmitter FIFO half-full (TH) - indicates that the FIFO is less than half-full.
6
Framing error (FE) - indicates that a framing error was detected.
5
Parity error (PE) - indicates that a parity error was detected.
4
Overrun (OV) - indicates that one or more character have been lost due to overrun.
3
Break received (BR) - indicates that a BREAK has been received.
2
Transmitter FIFO empty (TE) - indicates that the transmitter FIFO is empty.
1
Transmitter shift register empty (TS) - indicates that the transmitter shift register is empty.
0
Data ready (DR) - indicates that new data is available in the receiver holding register.
GR740-UM-DS, Jun 2016, Version 1.4
370
www.cobham.com/gaisler
GR740
28.7.3 UART Control Register
Table 469. UART control register
31 30
15 14 13 12 11 10
FA
RESERVED
SI
DI
9
8
7
6
5
4
3
BI DB RF TF EC LB FL PE PS TI
0
NR
0
2
1
0
RI TE RE
1
NR NR NR NR NR NR
r
rw rw rw rw rw rw rw rw rw rw rw rw rw rw rw
NR NR NR NR
0
0
31
FIFOs available (FA) - Set to 1 when receiver and transmitter FIFOs are available. When 0, only
holding register are available.
30: 15
RESERVED
14
Transmitter shift register empty interrupt enable (SI) - When set, an interrupt will be generated when
the transmitter shift register becomes empty. See section 28.6 for more details.
13
Delayed interrupt enable (DI) - When set, delayed receiver interrupts will be enabled and an interrupt will only be generated for received characters after a delay of 4 character times + 4 bits if no
new character has been received during that interval. This is only applicable if receiver interrupt
enable is set. See section 28.6 for more details.
12
Break interrupt enable (BI) - When set, an interrupt will be generated each time a break character is
received. See section 16.6 for more details.
11
FIFO debug mode enable (DB) - when set, it is possible to read and write the FIFO debug register.
10
Receiver FIFO interrupt enable (RF) - when set, Receiver FIFO level interrupts are enabled.
9
Transmitter FIFO interrupt enable (TF) - when set, Transmitter FIFO level interrupts are enabled.
8
External Clock (EC) - if set, the UART scaler will be clocked by UARTI.EXTCLK.
7
Loop back (LB) - if set, loop back mode will be enabled.
6
Flow control (FL) - if set, enables flow control using CTS/RTS (when implemented).
5
Parity enable (PE) - if set, enables parity generation and checking (when implemented).
4
Parity select (PS) - selects parity polarity (0 = even parity, 1 = odd parity) (when implemented).
3
Transmitter interrupt enable (TI) - if set, interrupts are generated when characters are transmitted
(see section 28.6 for details).
2
Receiver interrupt enable (RI) - if set, interrupts are generated when characters are received (see section 28.6 for details).
1
Transmitter enable (TE) - if set, enables the transmitter.
0
Receiver enable (RE) - if set, enables the receiver.
28.7.4 UART Scaler Register
Table 470. UART scaler reload register
31
20 19
0
RESERVED
SCALER RELOAD VALUE
NR
rw
19:0
Scaler reload value
28.7.5 UART FIFO Debug Register
Table 471. UART FIFO debug register
31
8
RESERVED
7
0
DATA
N/R
rw
7: 0
Transmitter holding register or FIFO (read access)
7: 0
Receiver holding register or FIFO (write access)
GR740-UM-DS, Jun 2016, Version 1.4
371
www.cobham.com/gaisler
GR740
29
SPI Controller supporting master and slave operation
29.1
Overview
The core provides a link between the AMBA APB bus and the Serial Peripheral Interface (SPI) bus
and can be dynamically configured to function either as a SPI master or a slave. The SPI bus parameters are highly configurable via registers. Core features also include configurable word length, bit
ordering, clock gap insertion and automatic slave select. All SPI modes are supported and also a 3wire mode where one bidirectional data line is used. In slave mode the core synchronizes the incoming clock and can operate in systems where other SPI devices are driven by asynchronous clocks.
SPICTRL
A
M
B
A
A
P
B
Mode register
Event register
Mask register
Control
Transmit
FIFO
Master ctrl
Com. register
Transmit register
Receive register
Clock gen.
Receive
FIFO
Slave select reg.
Slave ctrl
S
y
n
c
r
e
g
i
s
t
e
r
s
SPI_SCK
SPI_MISO
SPI_MOSI
SPI_SEL
SPI_SLVSEL[1:0]
Figure 50. Block diagram
29.2
Operation
29.2.1 SPI transmission protocol
The SPI bus is a full-duplex synchronous serial bus. Transmission starts when a master selects a slave
through the slave’s Slave Select (SPI_SLVSEL) signal and the clock line SCK transitions from its idle
state. Data is transferred from the master through the Master-Output-Slave-Input (SPI_MOSI) signal
and from the slave through the Master-Input-Slave-Output (SPI_MISO) signal. In a system with only
one master and one slave, the Slave Select input of the slave may be always active and the master does
not need to have a slave select output. If the core is configured as a master it will monitor the SPISEL
signal to detect collisions with other masters, if SPI_SEL is activated the master will be disabled.
During a transmission on the SPI bus data is either changed or read at a transition of SPI_SCK. If data
has been read at edge n, data is changed at edge n+1. If data is read at the first transition of SPI_SCK
the bus is said to have clock phase 0, and if data is changed at the first transition of SCK the bus has
clock phase 1. The idle state of SPI_SCK may be either high or low. If the idle state of SPI_SCK is
low, the bus has clock polarity 0 and if the idle state is high the clock polarity is 1. The combined values of clock polarity (CPOL) and clock phase (CPHA) determine the mode of the SPI bus. Figure 51
shows one byte (0x55) being transferred MSb first over the SPI bus under the four different modes.
Note that the idle state of the MOSI line is ‘1’ and that CPHA = 0 means that the devices must have
data ready before the first transition of SPI_SCK. The figure does not include the SPI_MISO signal,
the behavior of this line is the same as for the SPI_MOSI signal. However, due to synchronization
issues the SPI_MISO signal will be delayed when the core is operating in slave mode, please see section 29.2.5 for details.
GR740-UM-DS, Jun 2016, Version 1.4
372
www.cobham.com/gaisler
GR740
CPOL = 0
CPHA = 0
SCK
Mode 0
MOSI
CPHA = 1
SCK
Mode 1
MOSI
CPOL = 1
CPHA = 0
SCK
Mode 2
MOSI
CPHA = 1
SCK
Mode 3
MOSI
Figure 51. SPI transfer of byte 0x55 in all modes
29.2.2 3-wire transmission protocol
The core can be configured to operate in 3-wire mode, where the controller uses a bidirectional
dataline instead of separate data lines for input and output data. In 3-wire mode the bus is thus a halfduplex synchronous serial bus. Transmission starts when a master selects a slave through the slave’s
Slave Select (SPI_SLVSEL) signal and the clock line SPI_SCK transitions from its idle state. Only
the Master-Output-Slave-Input (SPI_MOSI) signal is used for data transfer in 3-wire mode. The
SPI_MISO signal is not used.
The direction of the first data transfer is determined by the value of the 3-wire Transfer Order (TTO)
field in the core’s Mode register. If TTO is ‘0’, data is first transferred from the master (through the
MOSI signal). After a word has been transferred, the slave uses the same data line to transfer a word
back to the master. If TTO is ‘1’ data is first transferred from the slave to the master. After a word has
been transferred, the master uses the MOSI line to transfer a word back to the slave.
The data line transitions depending on the clock polarity and clock phase in the same manner as in SPI
mode. The aforementioned slave delay of the SPI_MISO signal in SPI mode will affect the SPI_MOSI signal in 3-wire mode, when the core operates as a slave.
29.2.3 Receive and transmit queues
The core’s transmit queue consists of the transmit register and the transmit FIFO. The receive queue
consists of the receive register and the receive FIFO. The total number of words that can exist in each
queue is thus the FIFO depth plus one. When the core has one or more free slots in the transmit queue
it will assert the Not full (NF) bit in the event register. Software may only write to the transmit register
when this bit is asserted. When the core has received a word, as defined by word length (LEN) in the
Mode register, it will place the data in the receive queue. When the receive queue has one or more elements stored the Event register bit Not empty (NE) will be asserted. The receive register will only
contain valid data if the Not empty bit is asserted and software should not access the receive register
GR740-UM-DS, Jun 2016, Version 1.4
373
www.cobham.com/gaisler
GR740
unless this bit is set. If the receive queue is full and the core receives a new word, an overrun condition will occur. The received data will be discarded and the Overrun (OV) bit in the Event register will
be set.
The core will also detect underrun conditions. An underrun condition occurs when the core is
selected, via SPISEL, and the SCK clock transitions while the transmit queue is empty. In this scenario the core will respond with all bits set to ‘1’ and set the Underrun (UN) bit in the Event register.
An underrun condition will never occur in master mode. When the master has an empty transmit
queue the bus will go into an idle state.
29.2.4 Clock generation
The core only generates the clock in master mode, the generated frequency depends on the system
clock frequency and the Mode register fields DIV16, FACT, and PM. Without DIV16 the SCK frequency is:
AMBAclockfrequency
SCKFrequency = -------------------------------------------------------------------- 4 –  2  FACT     PM + 1 
With DIV16 enabled the frequency of SCK is derived through:
AMBAclockfrequency
SCKFrequency = -------------------------------------------------------------------------------16   4 –  2  FACT     PM + 1 
Note that the fields of the Mode register, which includes DIV16, FACT and PM, should not be
changed when the core is enabled.
29.2.5 Slave operation
When the core is configured for slave operation it does not drive any SPI signal until the core is
selected, via the SPI_SEL input, by a master. If the core operates in SPI mode when SPI_SEL goes
low the core configures SPI_MISO as an output and drives the value of the first bit scheduled for
transfer. If the core is configured into 3-wire mode the core will first listen to the SPI_MOSI line and
when a word has been transferred drive the response on the SPI_MOSI line. If the core is selected
when the transmit queue is empty it will transfer a word with all bits set to ‘1’ and the core will report
an underflow.
Since the core synchronizes the incoming clock it will not react to transitions on SPI_SCK until two
system clock cycles have passed. This leads to a delay of three system clock cycles when the data output line should change as the result of a SPI_SCK transition. This constrains the maximum input
SPI_SCK frequency of the slave to (system clock) / 8 or less. The controlling master must also allow
the decreased setup time on the slave data out line.
The core can also filter the SCK input. The value of the PM field in the Mode register defines for how
many system clock cycles the SCK input must be stable before the core accepts the new value. If the
PM field is set to zero, then the maximum SCK frequency of the slave is, as stated above, (system
clock) / 8 or less. For each increment of the PM field the clock period of SCK must be prolonged by
two times the system clock period as the core will require longer time discover and respond to SCK
transitions.
29.2.6 Master operation
When the core is configured for master operation it will transmit a word when there is data available
in the transmit queue. When the transmit queue is empty the core will drive SPI_SCK to its idle state.
If the SPI_SEL input goes low during master operation the core will abort any active transmission and
GR740-UM-DS, Jun 2016, Version 1.4
374
www.cobham.com/gaisler
GR740
the Multiple-master error (MME) bit will be asserted in the Event register. If a Multiple-master error
occurs the core will be disabled. Note that the core will react to changes on SPI_SEL even if the core
is operating in loop mode and that the core can be configured to ignore SPI_SEL by setting the IGSEL
field in the Mode register.
29.3
Registers
The core is programmed through registers mapped into APB address space.
Table 472.SPI controller registers
APB address offset
Register
0x00
Capability register
0x04-0x1C
Reserved
0x20
Mode register
0x24
Event register
0x28
Mask register
0x2C
Command register
0x30
Transmit register
0x34
Receive register
0x38
Slave Select register
0x3C
Automatic slave select register
0x3F-0xFF
Reserved
GR740-UM-DS, Jun 2016, Version 1.4
375
www.cobham.com/gaisler
GR740
29.3.1 Capability register
Table 473.0x00 - CAP - Capability register
31
24 23
20 19 18 17 16 15
8
7
6
5
4
0
SSSZ
MAXWLEN
T
W
E
N
A
M
O
D
E
A
S
E
L
A
S
S
E
N
FDEPTH
SR
FT
REV
2
0
1
0
1
1
4
1
0
0x5
r
r
r
r
r
r
r
r
r
r
31: 24
Slave Select register size (SSSZ) -This field contains the number of available signals: 2.
23: 20
Maximum word Length (MAXWLEN) - The maximum word length supported by the core: 
0b0000 is 4-16, and 32-bit word length.
19
Three-wire mode Enable (TWEN) - ‘1’, the core supports three-wire mode.
18
Auto mode (AMODE) - ‘0’
17
Automatic slave select available (ASELA) - ‘1’, core has support for setting slave select signals
automatically.
16
Slave Select Enable (SSEN) - ‘1’, the core has a slave select register.
15: 8
FIFO depth (FDEPTH) - This field contains the depth (4) of the core’s internal FIFOs. The number
of words the core can store in each queue is FDEPTH+1, since the transmit and receive registers can
contain one word each.
7
SYNCRAM (SR) - ‘1’, signals type of internal buffers. No impact for software.
6: 5
Fault-tolerance (FT) - “00”, internal buffers is implemented with radiation hardended flip-flops.
4: 0
Core revision (REV) - Core has revision 5.
29.3.2 Mode register
Table 474.0x20 - MODE - Mode register
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
R
L
O
O
P
C
P
O
L
C
P
H
A
D
I
V
1
6
R
E
V
M
S
E
N
0
0
0
0
0
0
0
0
0
r
rw rw rw rw rw rw rw
rw
LEN
PM
9
T
W
E
N
A
S
E
L
F
A
C
T
O
D
0
0
0
0
0
0
rw
rw rw rw rw
rw
CG
8
7
6
5
4
3
2
1
0
T
A
C
T
T
O
I
G
S
E
L
C
I
T
E
R
0
0
0
0
0
0
rw
rw rw rw rw
r
A
S
E
L
D
E
L
31
RESERVED
30
Loop mode (LOOP) - When this bit is set, and the core is enabled, the core’s transmitter and receiver
are interconnected and the core will operate in loopback mode. The core will still detect, and will be
disabled, on Multiple-master errors.
29
Clock polarity (CPOL) - Determines the polarity (idle state) of the SCK clock.
28
Clock phase (CPHA) - When CPHA is ‘0’ data will be read on the first transition of SCK. When
CPHA is ‘1’ data will be read on the second transition of SCK.
27
Divide by 16 (DIV16) - Divide system clock by 16, see description of PM field below and see section 29.2.4 on clock generation. This bit has no significance in slave mode.
26
Reverse data (REV) - When this bit is ‘0’ data is transmitted LSB first, when this bit is ‘1’ data is
transmitted MSB first. This bit affects the layout of the transmit and receive registers.
25
Master/Slave (MS) - When this bit is set to ‘1’ the core will act as a master, when this bit is set to ‘0’
the core will operate in slave mode.
24
Enable core (EN) - When this bit is set to ‘1’ the core is enabled. No fields in the mode register
should be changed while the core is enabled. This can bit can be set to ‘0’ by software, or by the core
if a multiple-master error occurs.
GR740-UM-DS, Jun 2016, Version 1.4
376
www.cobham.com/gaisler
GR740
Table 474.0x20 - MODE - Mode register
23: 20
Word length (LEN) - The value of this field determines the length in bits of a transfer on the SPI bus.
Values are interpreted as:
0b0000 - 32-bit word length
0b0001-0b0010 - Illegal values
0b0011-0b1111 - Word length is LEN+1, allows words of length 4-16 bits.
19: 16
Prescale modulus (PM) - This value is used in master mode to divide the system clock and generate
the SPI SCK clock. The value in this field depends on the value of the FACT bit.
If bit 13 (FACT) is ‘0’:The system clock is divided by 4*(PM+1) if the DIV16 field is ‘0’ and
16*4*(PM+1) if the DIV16 field is set to ‘1’. The highest SCK frequency is attained when PM is set
to 0b0000 and DIV16 to ‘0’, this configuration will give a SCK frequency that is (system clock)/4.
With this setting the core is compatible with the SPI register interface found in MPC83xx SoCs.
If bit 13 (FACT) is ‘1’: The system clock is divided by 2*(PM+1) if the DIV16 field is ‘0’ and
16*2*(PM+1) if the DIV16 field is set to ‘1’. The highest SCK frequency is attained when PM is set
to 0b0000 and DIV16 to ‘0’, this configuration will give a SCK frequency that is (system clock)/2.
In slave mode the value of this field defines the number of system clock cycles that the SCK input
must be stable for the core to accept the state of the signal. See section 29.2.5.
15
Three-wire mode (TW) - If this bit is set to ‘1’ the core will operate in 3-wire mode.
14
Automatic slave select (ASEL) - If this bit is set to ‘1’ the core will swap the contents in the Slave
select register with the contents of the Automatic slave select register when a transfer is started and
the core is in master mode. When the transmit queue is empty, the slave select register will be
swapped back. Note that if the core is disabled (by writing to the core enable bit or due to a multiplemaster-error (MME)) when a transfer is in progress, the registers may still be swapped when the core
goes idle. Also see the ASELDEL field which can be set to insert a delay between the slave select
register swap and the start of a transfer.
13
PM factor (FACT) - If this bit is 1 the core’s register interface is no longer compatible with the
MPC83xx register interface. The value of this bit affects how the PM field is utilized to scale the SPI
clock. See the description of the PM field.
12
Open drain mode (OD) - If this bit is set to ‘0’, all pins are configured for operation in normal mode.
If this bit is set to ‘1’ all pins are set to open drain mode. The pins driven from the slave select register are not affected by the value of this bit.
11: 7
Clock gap (CG) - The value of this field is only significant in master mode. The core will insert CG
SCK clock cycles between each consecutive word. This only applies when the transmit queue is kept
non-empty. After the last word of the transmit queue has been sent the core will go into an idle state
and will continue to transmit data as soon as a new word is written to the transmit register, regardless
of the value in CG. A value of 0b00000 in this field enables back-to-back transfers.
6
Automatic Slave Select Delay (ASELDEL) - If the core is configured to use automatic slave select
(ASEL field set to ‘1’) the core will insert a delay corresponding to ASELDEL*(SPI SCK cycle
time)/2 between the swap of the slave select registers and the first toggle of the SCK clock. As an
example, if this field is set to “10” the core will insert a delay corresponding to one SCK cycle
between assigning the Automatic slave select register to the Slave select register and toggling SCK
for the first time in the transfer.
5
Toggle Automatic slave select during Clock Gap (TAC) - If this bit is set, and the ASEL field is set,
the core will perform the swap of the slave select registers at the start and end of each clock gap. The
clock gap is defined by the CG field and must be set to a value >= 2 if this field is set.
4
3-wire Transfer Order (TTO) - This bit controls if the master or slave transmits a word first in 3-wire
mode.If this bit is ‘0’, data is first transferred from the master to the slave. If this bit is ‘1’, data is
first transferred from the slave to the master.
3
Ignore SPISEL input (IGSEL) - If this bit is set to ‘1’ then the core will ignore the value of the SPISEL input.
2
RESERVED
1
Require Clock Idle for Transfer End (CITE) - If this bit is ‘0’ the core will regard the transfer of a
word as completed when the last bit has been sampled. If this bit is set to ‘1’ the core will wait until
it has set the SCK clock to its idle level (see CI field) before regarding a transfer as completed. This
setting only affects the behavior of the TIP status bit, and automatic slave select toggling at the end
of a transfer, when the clock phase (CP field) is ‘0’.
0
RESERVED (R) - Read as zero and should be written as zero to ensure forward compatibility.
GR740-UM-DS, Jun 2016, Version 1.4
377
www.cobham.com/gaisler
GR740
29.3.3 Event register
Table 475.0x24 - EVENT - Event register
31 30
15 14 13 12 11 10
TIP
RESERVED
0
r
9
8
LT
R OV UN M NE NF
M
E
0
0
0
r
wc
r
0
0
0
wc wc wc
7
0
RESERVED
0
0
0
r
r
r
31
Transfer in progress (TIP) - This bit is ‘1’ when the core has a transfer in progress. Writes have no
effect. This bit is set when the core starts a transfer and is reset to ‘0’ once the core considers the
transfer to be finished. Behavior affected by setting of CITE field in Mode register.
30: 15
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
14
Last character (LT) - This bit is set when a transfer completes if the transmit queue is empty and the
LST bit in the Command register has been written. This bit is cleared by writing ‘1’, writes of ‘0’
have no effect.
13
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
12
Overrun (OV) - This bit gets set when the receive queue is full and the core receives new data. The
core continues communicating over the SPI bus but discards the new data. This bit is cleared by writing ‘1’, writes of ‘0’ have no effect.
11
Underrun (UN) - This bit is only set when the core is operating in slave mode. The bit is set if the
core’s transmit queue is empty when a master initiates a transfer. When this happens the core will
respond with a word where all bits are set to ‘1’. This bit is cleared by writing ‘1’, writes of ‘0’ have
no effect.
10
Multiple-master error (MME) - This bit is set when the core is operating in master mode and the SPISEL input goes active. In addition to setting this bit the core will be disabled. This bit is cleared by
writing ‘1’, writes of ‘0’ have no effect.
9
Not empty (NE) - This bit is set when the receive queue contains one or more elements. It is cleared
automatically by the core, writes have no effect.
8
Not full (NF) - This bit is set when the transmit queue has room for one or more words. It is cleared
automatically by the core when the queue is full, writes have no effect. This field is only updated
when the core is enabled (EN field of Mode register is set to ’1’).
7: 0
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
GR740-UM-DS, Jun 2016, Version 1.4
378
www.cobham.com/gaisler
GR740
29.3.4 Mask register
Table 476.0x28 - MASK - Mask register
31 30
15 14 13 12 11 10
T
I
P
E
RESERVED
0
rw
31
9
8
O
V
E
U
N
E
M
M
E
E
N
E
E
E
N
F
E
RESERVED
0
0
0
0
0
1
0
r
rw rw rw rw rw
r
L
T
E
R
0
0
r
rw
7
0
Transfer in progress enable (TIPE) - When this bit is set the core will generate an interrupt when the
TIP bit in the Event register transitions from ‘0’ to ‘1’.
30: 15
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
14
Last character enable (LTE) - When this bit is set the core will generate an interrupt when the LT bit
in the Event register transitions from ‘0’ to ‘1’.
13
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
12
Overrun enable (OVE) - When this bit is set the core will generate an interrupt when the OV bit in
the Event register transitions from ‘0’ to ‘1’.
11
Underrun enable (UNE) - When this bit is set the core will generate an interrupt when the UN bit in
the Event register transitions from ‘0’ to ‘1’.
10
Multiple-master error enable (MMEE) - When this bit is set the core will generate an interrupt when
the MME bit in the Event register transitions from ‘0’ to ‘1’.
9
Not empty enable (NEE) - When this bit is set the core will generate an interrupt when the NE bit in
the Event register transitions from ‘0’ to ‘1’.
8
Not full enable (NFE) - When this bit is set the core will generate an interrupt when the NF bit in the
Event register transitions from ‘0’ to ‘1’.
7: 0
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
29.3.5 Command register
Table 477.0x2C - CMD - Command register
31 30
23 22 21
RESERVED
L
S
T
RESERVED
0
0
0
r
w
r
31: 23
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
22
Last (LST) - After this bit has been written to ‘1’ the core will set the Event register bit LT when a
character has been transmitted and the transmit queue is empty. If the core is operating in 3-wire
mode the Event register bit is set when the whole transfer has completed. This bit is automatically
cleared when the Event register bit has been set and is always read as zero.
21: 0
RESERVED (R) - Read as zero and should be written to zero to ensure forward compatibility.
GR740-UM-DS, Jun 2016, Version 1.4
379
www.cobham.com/gaisler
GR740
29.3.6 Transmit register
Table 478.0x30 - TX - Transmit register
31
0
TDATA
0
w
31: 0
Transmit data (TDATA) - Writing a word into this register places the word in the transmit queue.
This register will only react to writes if the Not full (NF) bit in the Event register is set. The layout of
this register depends on the value of the REV field in the Mode register:
Rev = ‘0’: The word to transmit should be written with its least significant bit at bit 0.
Rev = ‘1’: The word to transmit should be written with its most significant bit at bit 31.
29.3.7 Receive register
Table 479.0x34 - RX - Receive register
31
0
RDATA
0
r
31: 0
Receive data (RDATA) - This register contains valid receive data when the Not empty (NE) bit of the
Event register is set. The placement of the received word depends on the Mode register fields LEN
and REV:
For LEN = 0b0000 - The data is placed with its MSb in bit 31 and its LSb in bit 0.
For other lengths and REV = ‘0’ - The data is placed with its MSB in bit 15.
For other lengths and REV = ‘1’ - The data is placed with its LSB in bit 16.
To illustrate this, a transfer of a word with eight bits (LEN = 7) that are all set to one will have the
following placement:
REV = ‘0’ - 0x0000FF00
REV = ‘1’ - 0x00FF0000
29.3.8 Slave select register
Table 480.0x38 - SLVSEL - Slave select register
31
2
1
0
RESERVED
SLVSEL
0
0b11
r
rw
31: 2
RESERVED
1: 0
Slave select (SLVSEL) - The core’s slave select signals are mapped to this register on bits 1:0. Software is responsible for activating the correct slave select signals.
GR740-UM-DS, Jun 2016, Version 1.4
380
www.cobham.com/gaisler
GR740
29.3.9 Automatic slave select register
Table 481.0x38 - ASLVSEL - Automatic slave select register
31
2
RESERVED
1
0
A
S
L
V
S
E
L
0
0
r
rw
31: 2
RESERVED
1: 0
Automatic Slave select (ASLVSEL) - The core’s slave select signals are assigned from this register
when the core is about to perform a transfer and the ASEL field in the Mode register is set to ‘1’.
After a transfer has been completed the core’s slave select signals are assigned the original value in
the slave select register.
GR740-UM-DS, Jun 2016, Version 1.4
381
www.cobham.com/gaisler
GR740
30
Clock gating unit
30.1
Overview
The clock gating unit provides a mean to save power by disabling the clock to unused functional
blocks.
30.2
Operation
The operation of the clock gating unit is controlled through three registers: the unlock, clock enable
and core reset registers. The clock enable register defines if a clock is enabled or disabled. A ‘1’ in a
bit location will enable the corresponding clock, while a ‘0’ will disable the clock. The core reset register allows to generate a reset signal for each generated clock. A reset will be generated as long as the
corresponding bit is set to ‘1’. The bits in clock enable and core reset registers can only be written
when the corresponding bit in the unlock register is 1. If a bit in the unlock register is 0, the corresponding bits in the clock enable and core reset registers cannot be written.
To gate the clock for a core, the following procedure should be applied:
1. Disable the core through software to make sure it does not initialize any AHB accesses
2. Write a 1 to the corresponding bit in the unlock register
3. Write a 0 to the corresponding bit in the clock enable register
4. Write a 0 to the corresponding bit in the unlock register
To enable the clock for a core, the following procedure should be applied
1. Write a 1 to the corresponding bit in the unlock register
2. Write a 1 to the corresponding bit in the core reset register
3. Write a 1 to the corresponding bit in the clock enable register
4. Write a 0 to the corresponding bit in the clock enable register
5. Write a 0 to the corresponding bit in the core reset register
6. Write a 1 to the corresponding bit in the clock enable register
7. Write a 0 to the corresponding bit in the unlock register
The cores connected to the clock gating unit are defined in the table below:
Table 482.Clocks controlled by CLKGATE unit
Bit
Functional module
0
GRETH 10/100/1000 Mbit Ethernet MAC 0
1
GRETH 10/100/1000 Mbit Ethernet MAC 1
2
SpaceWire router
3
PCI master/target controller
4
MIL-STD-1553B controller
5
CAN controller
6
LEON4 Statistics unit
7
UART 0
8
UART 1
9
SPI Controller
10
PROM/IO memory controller
GR740-UM-DS, Jun 2016, Version 1.4
382
www.cobham.com/gaisler
GR740
The clock gating unit also provides gating for the processor cores and floating-point units. A processor core will be automatically gated off when it enters power down mode. A FPU will be gated off
when processor core connected to the FPU has floating-point disabled or when the processor core is in
power down mode.
Processor/FPU clock gating can be disabled by writing 0x000F000F to the CPU/FPU override register.
30.3
Registers
Table 483 shows the clock gating unit registers.
Table 483.Clock gating unit registers
APB address offset
Register
0x00
Unlock register
0x04
Clock enable register
0x08
Core reset register
0x0C
CPU/FPU override register
0x10-0xFF
Reserved
GR740-UM-DS, Jun 2016, Version 1.4
383
www.cobham.com/gaisler
GR740
30.3.1 Unlock register
Table 484.0x00 - UNLOCK - Unlock register
31
11 10
0
RESERVED
UNLOCK
0
0
r
rw
31: 11
RESERVED
10: 0
Unlock clock enable and reset registers (UNLOCK) - The bits in clock enable and core reset registers can only be written when the corresponding bit in this field is 1.
30.3.2 Clock enable register
Table 485.0x04 - CLKEN - Clock enable register
31
11 10
0
RESERVED
ENABLE
0
*
r
rw
31: 11
RESERVED
10: 0
Cock enable (ENABLE) - A ‘1’ in a bit location will enable the corresponding clock, while a ‘0’ will
disable the clock.
The reset value of this register is set by bootstrap signals. See section 4.9.
30.3.3 Core reset register
Table 486. 0x08 - RESET - Reset register
31
11 10
0
RESERVED
RESET
0
*
r
rw
31: 11
RESERVED
10: 0
Reset (RESET) - A reset will be generated as long as the corresponding bit is set to ‘1’.
The reset value of this register is set by bootstrap signals. See section 4.9.
GR740-UM-DS, Jun 2016, Version 1.4
384
www.cobham.com/gaisler
GR740
30.3.4 CPU/FPU override register
Table 487. 0x0c - OVERRIDE - CPU/FPU override register
31
20 19
RESERVED
16 15
4
3
0
FOVERRIDE
RESERVED
OVERRIDE
0
0
0
0
r
rw
r
rw
31: 20
RESERVED
19: 16
Override FPU clock gating (FOVERRIDE) - If bit n of this field is set to ’1’ then the clock for FPU
n will be active regardless of the value of %PSR.EF.
15: 4
RESERVED
3: 0
Override CPU clock gating (OVERRIDE) - If bit n of this field is set to ’1’ then the clock for processor n and FPU n will always be active.
GR740-UM-DS, Jun 2016, Version 1.4
385
www.cobham.com/gaisler
GR740
31
LEON4 Statistics Unit (Performance Counters)
31.1
Overview
The LEON4 Statistics Unit (L4STAT) is used count events in the LEON4 processor and the AHB bus,
in order to create performance statistics for various software applications.
L4STAT consists of sixteen 32-bit counters that increment on a selected event. The counters roll over
to zero when reaching their maximum value, but can also be automatically cleared on reading to facilitate statistics building over longer periods. Each counter has a control register where the event type is
selected. The control registers also indicates which particular processor core is monitored. The table
488 below shows the event types that can be monitored.
Table 488.Event types and IDs
ID
Event description
Processor events:
0x00
Instruction cache miss
0x01
Instruction MMU TLB miss
0x02
Instruction cache hold
0x03
Instruction MMU hold
0x08
Data cache (read) miss
0x09
Data MMU TLB miss
0x0A
Data cache hold
0x0B
Data MMU hold
0x10
Data write buffer hold
0x11
Total instruction count
0x12
Integer instructions
0x13
Floating-point unit instruction count
0x14
Branch prediction miss
0x15
Execution time, excluding debug mode
0x17
AHB utilization (per AHB master)
0x18
AHB utilization (total, master/CPU selection is ignored)
0x22
Integer branches
0x28
CALL instructions
0x30
Regular type 2 instructions
0x38
LOAD and STORE instructions
0x39
LOAD instructions
0x3A
STORE instructions
AHB events (counted via LEON4 Debug Support Unit):
0x40
AHB IDLE cycles. Filtered on CPU/AHBM if SU(1) = ‘1
0x41
AHB BUSY cycles. Filtered on CPU/AHBM if SU(1) = ‘1
0x42
AHB NON-SEQUENTIAL transfers. Filtered on CPU/AHBM if SU(1) = ‘1
0x43
AHB SEQUENTIAL transfers. Filtered on CPU/AHBM if SU(1) = ‘1
0x44
AHB read accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x45
AHB write accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x46
AHB byte accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x47
AHB half-word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x48
AHB word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x49
AHB double word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
GR740-UM-DS, Jun 2016, Version 1.4
386
www.cobham.com/gaisler
GR740
Table 488.Event types and IDs
ID
Event description
0x4A
AHB quad word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x4B
AHB eight word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x4C
AHB waitstates. Filtered on CPU/AHBM if SU(1) = ‘1
0x4D
AHB RETRY responses. Filtered on CPU/AHBM if SU(1) = ‘1
0x4E
AHB SPLIT responses. Filtered on CPU/AHBM if SU(1) = ‘1
0x4F
AHB SPLIT delay. Filtered on CPU/AHBM if SU(1) = ‘1
0x50
AHB bus locked. Filtered on CPU/AHBM if SU(1) = ‘1
0x51-0x5F
Reserved
Device specific events (may be marked as user defined events in generic software drivers):
0x60
L2 cache hit (external event 0, CPU/AHBM field can select AHB master)
0x61
L2 cache miss (external event 1, CPU/AHBM field can select AHB master)
0x62
L2 cache bus access (external event 2, CPU/AHBM field can select AHB master)
0x63
L2 cache tag correctable error (external event 3, CPU/AHBM field must be set to 0)
0x64
L2 cache tag uncorrectable error (external event 4, CPU/AHBM field must be set to 0)
0x65
L2 cache data correctable error (external event 5, CPU/AHBM field must be set to 0)
0x66
L2 cache data uncorrectable error (external event 6, CPU/AHBM field must be set to 0)
0x67
IOMMU cache lookup (external event 7, CPU/AHBM field must be set to 0)
0x68
IOMMU table walk (external event 8, CPU/AHBM field must be set to 0)
0x69
IOMMU access error/denied (external event 9, CPU/AHBM field must be set to 0)
0x6A
IOMMU access OK (external event 10, CPU/AHBM field must be set to 0)
0x6B
IOMMU access passthrough (external event 11, CPU/AHBM field must be set to 0)
0x6C
IOMMU cache/TLB miss (external event 12, CPU/AHBM field must be set to 0)
0x6D
IOMMU cache/TLB hit (external event 13, CPU/AHBM field must be set to 0)
0x6E
IOMMU cache/TLB parity error (external event 14, CPU/AHBM field must be set to 0)
AHB events (only available if core is connected to a standalone AHB trace buffer):
0x70
AHB IDLE cycles
0x71
AHB BUSY cycles. Filtered on CPU/AHBM if SU(1) = ‘1’
0x72
AHB NON-SEQUENTIAL transfers. Filtered on CPU/AHBM if SU(1) = ‘1
0x73
AHB SEQUENTIAL transfers. Filtered on CPU/AHBM if SU(1) = ‘1
0x74
AHB read accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x75
AHB write accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x76
AHB byte accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x77
AHB half-word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x78
AHB word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x79
AHB double word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x7A
AHB quad word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x7B
AHB eight word accesses. Filtered on CPU/AHBM if SU(1) = ‘1
0x7C
AHB waitstates. Filtered on CPU/AHBM if SU(1) = ‘1
0x7D
AHB RETRY responses. Filtered on CPU/AHBM if SU(1) = ‘1
0x7E
AHB SPLIT responses. Filtered on CPU/AHBM if SU(1) = ‘1
0x7F
AHB SPLIT delay. Filtered on CPU/AHBM if SU(1) = ‘1
Events generated from REQ/GNT signals
GR740-UM-DS, Jun 2016, Version 1.4
387
www.cobham.com/gaisler
GR740
Table 488.Event types and IDs
ID
Event description
0x80 - 0x8F
Active when master selected by CPU/AHBM field has request asserted while grant is
asserted for the master correspoding to the least significant nibble of the event ID.
The following map is a map between least siginificant nibble of this event and CPU/AHBM
field to masters in the system
8:6 - Masters 2, 1, 0 on memory AHB bus
5:0 - Masters on Processor AHB bus
0x90 - 0x9F
Active when master selected by CPU/AHBM field has request asserted while grant is deasserted for the master correspoding to the least significant nibble of the event ID. See events
0x80 - 0x8F for a description of masters corresponding to the different events.
Note that IDs 0x39 (LOAD instructions) and 0x3A (STORE instructions) will both count all LDST
and SWAP instructions. The sum of events counted for 0x39 and 0x3A may therefore be larger than
the number of events counted with ID 0x38 (LOAD and STORE instructions).
The documentation for the Debug Support Unit contains more information on events 0x40 - 0x5F, see
section 13.3.2. Please note that the statistical outputs from the DSU may be subject to AHB trace buffer filters. The same applies to events 0x70 to 0x7F that come from the AHBTRACE standalone AHB
trace buffer.
Collecting statistics on interrupt latency is possible using the interrupt timestamping mechnism
described in section 26.2.6. Interrupt latency can not be measured with the LEON4 statistics unit.
Master indexes are listed in section 2.5.
31.2
Multiple APB interfaces
The core has two AMBA APB interfaces, the first is connected via the Processor AHB bus and the
second is connected via the Debug AHB bus. The first APB interface always has precedence when
both interfaces handle write operations to the same address.
GR740-UM-DS, Jun 2016, Version 1.4
388
www.cobham.com/gaisler
GR740
31.3
Registers
The L4STAT unit is programmed through registers mapped into APB address space.
Table 489. L4STAT counter control register
APB address offset
Register
0x00
Counter 0 value register
0x04
Counter 1 value register
0x08
Counter 2 value register
0x0C
Counter 3 value register
0x10
Counter 4 value register
0x14
Counter 5 value register
0x18
Counter 6 value register
0x1C
Counter 7 value register
0x20
Counter 8 value register
0x24
Counter 9 value register
0x28
Counter 10 value register
0x2C
Counter 11 value register
0x30
Counter 12 value register
0x34
Counter 13 value register
0x38
Counter 14 value register
0x3C
Counter 15 value register
0x40 - 0x7C
Reserved
0x80
Counter 0 control register
0x84
Counter 1 control register
0x88
Counter 2 control register
0x8C
Counter 3 control register
0x90
Counter 4 control register
0x94
Counter 5 control register
0x98
Counter 6 control register
0x9C
Counter 7 control register
0xA0
Counter 8 control register
0xA4
Counter 9 control register
0xA8
Counter 10 control register
0xAC
Counter 11 control register
0xB0
Counter 12 control register
0xB4
Counter 13 control register
0xB8
Counter 14 control register
0xBC
Counter 15 control register
0xC0 - 0xFC
Reserved
0x100
Counter 0 max/latch register
0x104
Counter 1 max/latch register
0x108
Counter 2 max/latch register
0x10C
Counter 3 max/latch register
0x110
Counter 4 max/latch register
0x114
Counter 5 max/latch register
0x118
Counter 6 max/latch register
GR740-UM-DS, Jun 2016, Version 1.4
389
www.cobham.com/gaisler
GR740
Table 489. L4STAT counter control register
APB address offset
Register
0x11C
Counter 7 max/latch register
0x120
Counter 8 max/latch register
0x124
Counter 9 max/latch register
0x128
Counter 10 max/latch register
0x12C
Counter 11 max/latch register
0x130
Counter 12 max/latch register
0x134
Counter 13 rmax/latch egister
0x138
Counter 14 rmax/latch egister
0x13C
Counter 15 max/latch register
0x140 - 0x17C
Reserved
0x180
Timestamp register
0x184 - 0x1FC
Reserved
31.3.1 Counter value registers
Table 490.0x00-0x3C - CVAL0-15 - Counter 0-15 value register
31
0
CVAL
NR
rw
31: 0
Counter value (CVAL) - This register holds the current value of the counter. If the core has been
implemented with support for keeping the maximum count (MC field of Counter control register is
‘1’) and the Counter control register field CD is ‘1’, then the value displayed by this register will be
the maximum counter value reached with the settings in the counter’s control register. Writing to this
register will write both to the counter and, if implemented, the hold register for the maximum
counter value.
GR740-UM-DS, Jun 2016, Version 1.4
390
www.cobham.com/gaisler
GR740
31.3.2 Counter control registers
Table 491.0x80-0xCC - CCTRL0-15 - Counter 0-15 control register
31
28 27
23 22 21 20 19 18 17 16 15 14 13 12 11
NCPU
NCNT
0x3
0xF
1
1
1
1
r
r
r
r
r
r
31: 28
MC IA DS EE AE EL CD
SU
CL EN
1 NR NR
NR
NR
r
rw
rw rw
4
3
0
EVENT ID
CPU/AHBM
0
NR
NR
rw rw
rw
rw
Number of CPU (NCPU) - Number of supported processors - 1
27: 23
Number of counters (NCNT) - Number of implemented counters - 1
22
Maximum count (MC) - If this field is ‘1’ then this counter has support for keeping the maximum
count value.
21
Internal AHB count (IA) - If this field is ‘1’ the core supports events 0x17 and 0x18
20
DSU support (DS) - If this field is ‘1’ the core supports events 0x40-0x5F
19
External events (EE) - If this field is ‘1’ the core supports external events (events 0x60 - 0x6F)
18
AHBTRACE Events (AE) - If this field is ‘1’ the core supports events 0x70 - 0x7F.
17
Event Level (EL) - The value of this field determines the level where the counter keeps running
when the CD field below has been set to ‘1’. If this field is ‘0’ the counter will count the time
between event assertions. If this field is ‘1’ the counter will count the cycles where the event is
asserted. This field can only be set if the MC field of this register is ‘1’.
16
Count maximum duration (CD) - If this bit is set to ‘1’ the core will save the maximum time the
selected event has been at the level specified by the EL field. This also means that the counter will be
reset when the event is activated or deactivated depending on the value of the EL field.
When this bit is set to ‘1’, the value shown in the counter value register will be the maximum current
value which may be different from the current value of the counter.
This field can only be set if the MC field of this register is ‘1’.
15: 14
Supervisor/User mode filter (SU) - “01” - Only count supervisor mode events, “10” - Only count
user mode events, others values - Count events regardless of user or supervisor mode. This setting
only applies to events 0x0 - 0x3A.
When SU = “1x” only events generated by the CPU/AHB master specified in the CPU/AHBM field
will be counted. This applies to events 0x40 - 0x7F
13
Clear counter on read (CL) - If this bit is set the counter will be cleared when the counter’s value is
read. The register holding the maximum value will also be cleared, if implemented.
If an event occurs in the same cycle as the counter is cleared by a read then the event will not be
counted. The counter latch register can be used to guarantee that no events are lost
12
Enable counter (EN) - Enable counter
11: 4
Event ID to be counted
3: 0
CPU or AHB master to monitor.(CPU/AHBM) - The value of this field does not matter when selecting one of the events coming from the Debug Support Unit or one of the external events.
GR740-UM-DS, Jun 2016, Version 1.4
391
www.cobham.com/gaisler
GR740
31.3.3 Counter max/latch registers
Table 492.0x100-0x13C - CSVAL0-15 - Counter 0-15 max/latch register
31
0
CSVAL
NR
rw*
31: 0
Counter max/latch value (CSVAL) - This register holds the current value of the counter max/latch
register.
If the counter control register field CD is ‘1’, then the value displayed by this register will be the
maximum counter value reached with the settings in the counter’s control register.
If the counter control register field CDis ‘0’, then the value displayed by this register is the latched
(saved) counter value. The counter value is saved whenever a write access is made to the core in
address range 0x100 - 0x1FC (all counters are saved simultaneously). If the counter control register
CL field is set, then the current counter value will be cleared when the counter value is saved into
this register.
31.3.4 Timestamp register
Table 493.0x180 - TSTAMP - Timestamp register
31
0
TSTAMP
NR
rw*
31: 0
Timestamp (TSTAMP) - Timestamp saved at latch of counters. The timestamp value uses the DSU
timer as time source. The same timer is available as the processor’s internal up-counter, for interrupt
timestamping and in the time-tag of the trace buffers. The time value is saved whenever a write
access is made to the core in address range 0x100 - 0x1FC.
GR740-UM-DS, Jun 2016, Version 1.4
392
www.cobham.com/gaisler
GR740
32
AHB Status Registers
32.1
Overview
AHB status registers store information about AMBA AHB accesses triggering an error response.
There is a status register and a failing address register capturing the control and address signal values
of a failing AMBA bus transaction, or the occurrence of a correctable error being signaled from a fault
tolerant core.
The system has two AHB status register cores. One monitoring the Processor AHB bus and one monitoring the Slave I/O AHB bus. Both cores are accessed via the AMBA APB bus. The memory scrubber core, described in section 11, on the Memory AHB bus also provides the same functionality as an
AHB status register.
32.2
Operation
32.2.1 Errors
The registers monitor AMBA AHB bus transactions and store the current HADDR, HWRITE,
HMASTER and HSIZE internally. The monitoring are always active after startup and reset until an
error response (HRESP = “01”) is detected. When the error is detected, the status and address register
contents are frozen and the New Error (NE) bit is set to one. At the same time an interrupt is generated, as described hereunder.
The fault tolerant memory controllers and L2 cache containing EDAC signal an un-correctable error
as an AMBA error response, so that it can be detected by the processor as described above.
32.2.2 Correctable errors
Not only error responses on the AHB bus can be detected. The PROM/IO controller has a correctable
error signal that is asserted each time a correctable error is detected. When such an error is detected,
the effect will be the same as for an AHB error response. The only difference is that the Correctable
Error (CE) bit in the status register is set to one when a correctable error is detected.
When the CE bit is set the interrupt routine can acquire the address containing the correctable error
from the failing address register and correct it. When it is finished it resets the NE bit and the monitoring becomes active again. Interrupt handling is described in detail hereunder.
Note that only the AHB status register monitoring the Slave I/O AHB bus reacts to correctable errors.
Correctable errors on the Processor AHB bus are reported via the L2 cache and correctable errors
from the memory controllers on the Memory AHB bus are reported via the memory scrubber core.
32.2.3 Interrupts
The interrupt is connected to the interrupt controller to inform the processor of the error condition.
The normal procedure is that an interrupt routine handles the error with the aid of the information in
the status registers. When it is finished it resets the NE bit and the monitoring becomes active again.
Interrupts are generated for both AMBA error responses and correctable errors as described above.
32.3
Registers
The core is programmed through registers mapped into APB address space.
Table 494.AHB Status registers
APB address offset
Registers
0x00
AHB Status register
0x04
AHB Failing address register
GR740-UM-DS, Jun 2016, Version 1.4
393
www.cobham.com/gaisler
GR740
Table 495.0x00 - AHBS - AHB Status register
31
10
RESERVED
9
8
CE NE
0
0
r
rw rw
0
7
6
3
2
0
HWRITE
HMASTER
HSIZE
N/R
N/R
N/R
r
r
r
31: 10
RESERVED
9
Correctable Erro (CE) - . Set if the detected error was caused by a correctable error and zero otherwise. Never set for register monitoring Processor AHB bus. For the register monitoring the Slave I/O
bus, this bit indicates that a correctable error was detected by the PROM/IO controller.
8
New Error (NE) - Deasserted at start-up and after reset. Asserted when an error is detected. Reset by
writing a zero to it.
7
HWRITE (HWRITE) - The HWRITE signal of the AHB transaction that caused the error.
6: 3
HMASTER (HMASTER) - The HMASTER signal of the AHB transaction that caused the error.
Table 496.0x04 - AHBFAR - AHB Failing address register
31
0
HADDR
N/R
r
31: 0
Failing address (HADDR) - The HADDR value of the AHB transaction that caused the error.
GR740-UM-DS, Jun 2016, Version 1.4
394
www.cobham.com/gaisler
GR740
33
Register for bootstrap signals
33.1
Overview
This interface provides a programmable register that controls the internal values of bootstrap signals
in the design.
33.2
Operation
The core contains one register of 20 bits that is mapped into APB address space. The value in bits 15:0
of this register is propagated to other peripherals in the design. The reset value of the register is taken
from the GPIO signals and from the bootstrap signals PLL_IGNLOCK, PCIMODE_ENABLE,
MEM_CLKSEL and MEM_IFWIDTH. The value of the GPIO and other bootstrap signal is latched
when the device’s internal reset signal is deasserted.
33.3
Registers
The peripheral provides one register mapped into APB address space.
Table 497.General purpose register registers
APB address offset
Register
0x00
Bootstrap register
0x04 - 0xFF
RESERVED (first register is aliased in all words in this area)
Table 498.0x00 - BOOTSTRAP - Bootstrap register
31
20 19 18 17 16 15
RESERVED
B4 B3 B2 B1
0
GPIO
*
rw*
31: 20
RESERVED
19
Reset value of PLL_IGNLOCK bootstrap signal (B4) - The value in this register can be written but
the changed value does not affect system operation.
18
Reset value of PCIMODE_ENABLE bootstrap signal (B3) - The value in this register can be written
but the changed value does not affect system operation.
17
Reset value of MEM_CLKSEL bootstrap signal (B2) - The value in this register can be written but
the changed value does not affect system operation.
16
Reset value of MEM_IFWIDTH (B1) - The value in this register can be written but the changed
value does not affect system operation.
15: 0
Bootstrap value (BOOTSTRAP) - Bootstrap value of external general purpose I/O signal
GPIO(15:0). Writes to this register will override the current value of the bootstrap signals. Writes do
NOT affect the value of the external GPIO signal of the function of the GPIO port.
GR740-UM-DS, Jun 2016, Version 1.4
395
www.cobham.com/gaisler
GR740
34
Temperature sensor controller
34.1
Overview
This peripheral provides an interface to the on-chip temperature sensor. It allows software to control
the temperature sensor, saves maximum and minimum temperature values and allows generation of
alarms.
Temperature sensor functionality has been disabled on the GR740-XX, GR740-CP and GR740-IP
devices described by this data sheet.
GR740-UM-DS, Jun 2016, Version 1.4
396
www.cobham.com/gaisler
GR740
35
Register Bank For I/O and PLL configuration registers
35.1
Overview
The core provides an array of programmable registers that are used to control pin sharing and allows
controlling the PLL reconfiguration unit.
35.2
Operation
35.2.1 Pin multiplexing control
The core controls the pin multiplexing for the PROM/IO interface shared pins.The pins are controlled
by a combination of two register bits according to table 499 below. The pin corresponding to each register bit position is described in section 3.2.1.
Table 499.Mapping between register bit and pin function
FTMEN ALTEN Pin function
1
*
PROM/IO interface
0
1
Alternate I/O interface
0
0
GPIO2
35.2.2 LVDS pad enable control
The LVDS enable control registers allow to turn off unused LVDS output drivers for the spacewire
links, as well as the differential memory clock output, in order to reduce power consumption. The single-ended memory clock output can also be disabled using this register, in case the differential one is
only used. All pads are reset to an enabled state.
35.2.3 Pad drive strength control
Register bits allow adjusting the drive strength of the non-differential pads in the design. The pads
have been divided by function into 20 groups, as shown in table below.
GR740-UM-DS, Jun 2016, Version 1.4
397
www.cobham.com/gaisler
GR740
Table 500.Groups for pad drive strength control
Group
Description
Pins
0
SDRAM outgoing clock
mem_clk_out
1
SDRAM address/command
mem_wen, mem_rasn, mem_cke[1:0], mem_casn, mem_ba,
mem_addr[12:0]
2
SDRAM chip-select
mem_sn[1:0], mem_addr[14:13]
3
SDRAM data-mask
mem_dqm[11:0]
4
SDRAM data non-muxed
mem_dq[79:64, 41, 32:0]
5
SDRAM data / PCI address /
ETH1
mem_dq[95:80, 63:48, 38]
6
SDRAM data / PCI command
mem_dq[47:43, 40:39, 37:34]
7
SDRAM data / PCI request
mem_dq[42]
8
SDRAM data / PCI inta
mem_dq[33]
9
ETH0
eth0_txer, eth0_txd[7:0], eth0_txen
10
PROMIO control non-muxed
promio_oen, promio_wen, promio_read, prom_cen[0]
11
PROMIO control / CAN
prom_cen[1], io_sn
12
PROMIO address / UART
promio_addr[27:26]
13
PROMIO addr / 1553
promio_addr[25:20]
14
PROMIO addr / UART/SPWD
promio_addr[19:16]
15
PROMIO address non-muxed
promio_addr[15:0]
16
PROMIO data MSB
promio_data[15:8]
17
PROMIO data LSB
promio_data[7:0]
18
GPIO
gpio[15:0]
19
Miscellaneous
dsu_active, proc_errorn, eth0_mdio, eth0_mdc, spi_miso, spi_mosi, spi_sck, spi_slvsel[1:0], pll_locked[5:0], wdogn, jtag_tdo,
gr1553_busatxin, gr1553_busbtxin
35.2.4 PLL reconfiguration interface
Three registers allow the PLLs in the design to be reconfigured. One of them is a read-only register
indicating the current configuration, the other two registers are for specifying the new configuration
and commanding the reprogramming.
When the PLL is reprogrammed, the entire system will be reset and the new configuration will be
reflected in the current configuration register. An extra flag field is included in the configuration
which does not fill any function, this can be used by software to preserve a few bits of information
across the restart.
The relation between configuration words and resulting PLL programming is described in section 4.5.
35.2.5 Lockdown register
These different configuration can be locked to prevent further modification using the lockdown register. Locking can be revokable or permanent (until next full system reset). Note that the lockdown register is not reset when the PLL is reconfigured but only when the external reset signal is asserted.
GR740-UM-DS, Jun 2016, Version 1.4
398
www.cobham.com/gaisler
GR740
35.3
Registers
The core is programmed through registers mapped into APB address space.
Table 501.General purpose register registers
APB address offset
Register
0x00
FTMCTRL function enable
0x04
Alternative function enable
0x08
LVDS and memclk pad enable
0x0C
PLL new configuration
0x10
PLL reconfigure command
0x14
PLL current configuration (read-only)
0x18
Drive strength configuration register 1
0x1C
Drive strength configuration register 2
0x20
Config lockdown register
0x24 - 0xFF
Reserved
35.3.1 FTMCTRL enable register
Table 502.0x00 - FTMFUNC - FTMCTRL function enable register
31
22 21
0
RESERVED
FTMEN
0
*
r
rw
31: 22
RESERVED
21: 0
Pinmux FTMCTRL function enable (FTMEN) - Bit mask corresponding to table 24, used in conjunction with Alternative function enable to determine pin function. If set to 1, pin is used as FTMCTRL function, if 0 alternate or GPIO function
Reset value determined by bootstrap signal GPIO[15], see section 3.2.
35.3.2 Alternative function enable register
Table 503.0x04 - ALTFUNC - Alternative function enable register
31
22 21
0
RESERVED
ALTEN
0
0x3fffff
r
rw
31: 22
RESERVED
21: 0
Pinmux alternative function enable (ALTFN) - Bit mask corresponding to table 24, used in conjunction with FTMCTLR function enable register to determine pin function. If set to 1, pin is used as
FTMCTRL or alternative function, if 0 FTMCTRL or GPIO function
GR740-UM-DS, Jun 2016, Version 1.4
399
www.cobham.com/gaisler
GR740
35.3.3 LVDS and memory clock pad enable register
Table 504.0x08 - LVDSMCLK - LVDS and memory clock pad enable register
31
18 17 16 15
RESERVED
8
7
0
S
M
E
M
D
M
E
M
RESERVED
SPWOE
0
1
1
0
0xFF
r
rw rw
r
rw
31: 18
RESERVED
17
Enable single-ended SDRAM clock output (SMEM) - 1=enabled, 0=disabled
16
Enable differential SDRAM clock output (DMEM) - 1=enabled, 0=disabled
15: 8
RESERVED
7: 0
Enable LVDS output drivers for SPW links 7...0 (SPWOE) - 1=enabled, 0=disabled
35.3.4 PLL new configuration register
Table 505.0x0C - PLLNEWCFG - PLL new configuration register
31
19 28 27 26
RESERVED SWTAG
r
18 17
SPWPLLCFG
9
8
0
MEMPLLCFG
SYSPLLCFG
0
0
0
0
rw
rw
rw
rw
31: 29
RESERVED
28: 27
Software tag (SWTAG) - Not fed to PLL. Can be used freely as tag data.
26: 18
New SPWPLL configuration (SPWPLLCFG) - To be used when reprogramming, see table 32
17: 9
New MEMPLL configuration (MEMPLLCFG) - To be used when reprogramming, see table 31
8: 0
New SYSPLL configuration (SYSPLLCFG) - To be used when reprogramming, see table 30
35.3.5 PLL reconfigure command register
Table 506.0x10 - PLLRECFG - PLL reconfigure command register
31
3
RESERVED
0
RECONF
0
0
r
rw
31: 3
RESERVED
2: 0
Reconfigure PLL (RECONF) - Write "000" then "111" to initiate reconfiguration
GR740-UM-DS, Jun 2016, Version 1.4
2
400
www.cobham.com/gaisler
GR740
35.3.6 PLL current configuration register
Table 507.0x14 - PLLCURCFG - PLL current configuration register
31
19 28 27 26
18 17
RESERVED SWTAG
r
9
8
0
SPWPLLCFG
MEMPLLCFG
SYSPLLCFG
0
0b000010000
0b000001010
0b000010101
r
r
r
r
31: 29
RESERVED
28: 27
Software tag (SWTAG) - Can be used freely as tag data.
26: 18
Current SPWPLL configuration (SPWPLLCFG) - See table 32
17: 9
Current MEMPLL configuration (MEMPLLCFG) - See table 31
8: 0
Current SYSPLL configuration (SYSPLLCFG) - See table 30
35.3.7 Drive strength configuration registers
Table 508.0x18 - DRVSTR1 - Drive strength configuration register 1
31
20 19 18 17 16 15 14 13 12 11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
S9
S8
S7
S6
S5
S4
S3
S2
S1
S0
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
0
r
31: 20
RESERVED
19: 18
Drive strength setting for output group 9 (S9)
17: 16
Drive strength setting for output group 8 (S8)
15: 14
Drive strength setting for output group 7 (S7)
13: 12
Drive strength setting for output group 6 (S6)
11: 10
Drive strength setting for output group 5 (S5)
9: 8
Drive strength setting for output group 4 (S4)
7: 6
Drive strength setting for output group 3 (S3)
5: 4
Drive strength setting for output group 2 (S2)
3: 2
Drive strength setting for output group 1 (S1)
1: 0
Drive strength setting for output group 0 (S0)
Table 509.0x1C - DRVSTR2 - Drive strength configuration register 2
31
20 19 18 17 16 15 14 13 12 11 10
RESERVED
9
8
7
6
5
4
3
2
1
0
S19
S18
S17
S16
S15
S14
S13
S12
S11
S10
rw
rw
rw
rw
rw
rw
rw
rw
rw
rw
0
r
31: 20
RESERVED
19: 18
Drive strength setting for output group 19 (S19)
17: 16
Drive strength setting for output group 18 (S18)
15: 14
Drive strength setting for output group 17 (S17)
13: 12
Drive strength setting for output group 16 (S16)
11: 10
Drive strength setting for output group 15 (S15)
9: 8
Drive strength setting for output group 14 (S14)
7: 6
Drive strength setting for output group 13 (S13)
5: 4
Drive strength setting for output group 12 (S12)
GR740-UM-DS, Jun 2016, Version 1.4
401
www.cobham.com/gaisler
GR740
Table 509.0x1C - DRVSTR2 - Drive strength configuration register 2
3: 2
Drive strength setting for output group 11 (S11)
1: 0
Drive strength setting for output group 10 (S10)
35.3.8 Configuration lockdown register
Table 510.0x20 - LOCKDOWN - Configuration lockdown register
31
24 23
RESERVED
16 15
8
7
0
PERMANENT
RESERVED
REVOCABLE
0
0
0
0
r
rw*
r
rw
31: 24
RESERVED
23: 16
Permanent lock bit mask (PERMANENT) - Bit N for register at offset 4*N, 1=locked, 0=unlocked
15: 8
RESERVED
7: 0
Revocable lock bit mask (REVOCABLE) - Bit N for register at offset 4*N, 1=locked, 0=unlocked
Bits that are set to 1 in this field can only be cleared with a full system reset.
GR740-UM-DS, Jun 2016, Version 1.4
402
www.cobham.com/gaisler
GR740
36
SpaceWire - Time Distribution Protocol Controller
36.1
Overview
This core provides basic time keeping functions such as Elapsed Time counter according to the
CCSDS Unsegmented Code specification. It provides support for setting and sampling the Elapsed
Time counter. It also includes a frequency synthesizer with which a binary frequency is generated to
drive the Elapsed Time counter. This interface also implements the SpaceWire - Time Distribution
Protocol (TDP). The protocol provides capability to transfer time values and synchronise them
between onboard users of SpaceWire network. The time values are transferred as CCSDS Time Codes
and synchronisation is performed through SpaceWire Time-Codes. The core also provides datation
services. The AMBA APB bus is used for configuration, control and status handling.
36.2
Protocol
The initiator and target maintain their own time locally. The Time Distribution Protocol provides the
means for transferring time of initiator to targets and for providing a synchronization point in time.
The time is transferred by means of an RMAP write command carrying a CCSDS Time Code (time
message). The synchronization event is signalled by means of transferring a SpaceWire Time-Code.
The transfer of the SpaceWire Time-Code is synchronized with time maintained by the initiator. To
distinguish which SpaceWire Time-Code is to be used for synchronization, the value of SpaceWire
Time-Code is transferred from initiator to target by means of an RMAP write command prior to actual
transmission of SpaceWire Time-Code itself. When there is more than one target the CCSDS Time
Code need to be transferred to each individual target separately [SPWCUC].
36.3
Functionality
The block diagram below shows how the controller is connected to the system.
Processor
AHB
AMBA
AHB
SpaceWire
Router
APB
AMBA
APB
Slave
SPWTDP
Time-Code/Interrupt
Figure 52. Block diagram
The foreseen usage of this core is to distribute and synchronise time between an initiator SPWTDP
core and one or more target SPWTDP (slave) cores using the SpaceWire interface for communication
between them.
The system can act as initiator (time master) and target being able to send and receive SpaceWire
Time-Codes. The initiator requires SpaceWire link interface implements an RMAP initiator. The Target requires SpaceWire link interface implements an RMAP target. The SPWTDP component is a part
GR740-UM-DS, Jun 2016, Version 1.4
403
www.cobham.com/gaisler
GR740
of this system providing SpaceWire Time-Codes, CCSDS Time Codes, datation, time-stamping of
distributed interrupts, support for transmission of CCSDS Time Codes through RMAP and support
for latency measurement and correction. In this implementation the CCSDS Time Codes carried
between the SpaceWire network is based on CCSDS Unsegmented Code format (CUC) which is
explained below [CCSDS]. The table below shows an example Preamble Field (P-Field) which corresponds to 40 bits of coarse time and 24 bits of fine time.
36.3.1 CCSDS Unsegmented Code: Preamble Field (P-Field)
Table 511.CCSDS Unsegmented Code P-Field definition
Bit
Value
Interpretation
0
“1”
Extension flag, P-Field extended with 2nd octet
1-3
“010”
Agency-defined epoch (Level 2)
Time code identification
4-5
“11”
(number of octets of coarse time) + 1
Detail bits for information on the code
6-7
“11”
(number of octets of fine time)
8
“0”
9-10
“01”
Number of additional octets of the
coarse time.
added to octet 1
11-13
“000”
Number of additional octets of the fine
time.
added to octet 1
Extension flag, P-Field not extended with 3rd octet
14-15
RESERVED
36.3.2 CCSDS Unsegmented Code: Time Field (T-Field)
For the unsegmented binary time codes described herein, the T-Field consists of a selected number of
contiguous time elements, each element being one octet in length. An element represents the state of 8
consecutive bits of a binary counter, cascaded with adjacent counters, which rolls over at a modulo of
256.
Table 512.Example CCSDS Unsegmented Code T-Field with 32 bit coarse and 24 bit fine time
CCSDS Unsegmented Code
Preamble
Time Field
Field
Coarse time
-
231
0:15
0
224
Fine time
223
216
215
28
27
20
2-1
31
32
2-8
2-9
2-15
2-16
2-24
55
The basic time unit is the second. The T-Field coarse time (seconds) can be maximum 56 bits and
minimum 8 bits. The T-Field fine time (sub seconds) can be maximum 80 bits and minimum of 0 bits.
The number of bits representing coarse and fine time implemented in this core can be obtained by
reading the DPF bits of Datation Preamble Field register.
The coarse time code elements are a count of the number of seconds elapsed from the initial time
value. This code is not UTC-based and leap second corrections do not apply according to CCSDS.
36.3.3 Time generation
The core consist of time generator which is the source for time in this system. The core may act as initiator or a target but both have their respective time generator. The Elapsed Time (ET) counter is
implemented complying with the CUC T-Field. The number of bits representing coarse and fine time
of a ET counter implemented in a design can be obtained by reading the DPF bits of Datation Preamble Field register.
GR740-UM-DS, Jun 2016, Version 1.4
404
www.cobham.com/gaisler
GR740
The counter is incremented on the system clock only when enabled by the frequency synthesizer. The
binary frequency required to determine the counter increment is derived from the system clock using
a frequency synthesizer (FS). The frequency synthesizer is incremented with a pre-calculated increment value, which matches the available system clock frequency. The frequency synthesizer generates
a tick every time it wraps around, which makes the ET time counter to step forward with the precalculated increment value. The output of frequency synthesizer is used for enabling the increment of ET
counter. The increment rate of the ET counter and frequency synthesizer counter should be set according to the system clock frequency. The ET counter increment rate is set by providing values to ETINC
bits in Configuration 2 register and frequency synthesizer counter is set by providing values to FSINC
bits in Configuration 1 register. The following table specifies some example ETINC and FSINC values for some frequencies. The below values are also obtained for this core’s current implementation
consist of Coarse time width 32, Fine time width 24 and Frequency synthesizer width of 30. To calculate for other frequencies and configuration refer the spreadsheet provided along with this document.
Table 513.Example values of ETINC and FSINC for corresponding frequencies
Frequency
ETINC
FSINC
50 MHz
0
360287970
250 MHz
0
72057594
33333333
2
135107990
The following section describes the cores capabilities if it configured as initiator or target.
36.3.4 Initiator
An initiator is a SpaceWire node distributing CCSDS Time Codes and SpaceWire Time-Codes. It is
also an RMAP initiator, capable of transmitting RMAP commands and receiving RMAP replies.
There is only one active initiator in a SpaceWire network during a mission phase.
The initiator performs the following tasks
•
Transmission of SpaceWire Time-Codes
The SpaceWire Time-Codes are provided by this component and transmission of those codes to targets should be performed by a SpaceWire interface.
•
Transmission of CCSDS Time Codes through RMAP
•
Datation, time-stamping and latency measurement
36.3.5 Target
A target is a SpaceWire node receiving CCSDS Time Codes and SpaceWire Time-Codes. A target is
also an RMAP target, capable of receiving RMAP commands and transmitting RMAP replies. There
can be one or more targets in a SpaceWire network.
The target performs the following tasks
•
Reception of SpaceWire Time-Codes
The SpaceWire Time-Codes sent from initiator are received by SpaceWire interface and provided to
this component in target.
•
Reception of CCSDS Time Codes through RMAP
•
Qualification of received time messages (CCSDS Time Codes) using SpaceWire Time-Codes
•
Initialization and Synchronisation of received CCSDS Time Codes with Elapsed Time counter
available in this component
•
Datation, time-stamping and latency correction
•
Jitter and drift mitigation (the MA bit in Status 0 register specify the availability of this service)
GR740-UM-DS, Jun 2016, Version 1.4
405
www.cobham.com/gaisler
GR740
36.3.6 Configuring initiator and target
The core is interfaced via an AMBA Advanced Peripheral Bus (APB) slave interface, providing a register view that is compatible with the Time Distribution Protocol (TDP). The core must be configured
according to the requirement either as initiator or target.
•
Initializing initiator
The initiator transmits the SpaceWire Time-Codes out of the core only when the Transmit Enable TE
bit in Configuration 0 register is enabled. The ET counter in initiator can be initialized (to provide any
initial value). Initialization is done by writing a time value into the Command Elapsed Time registers
available in the command field, the NC bit in the Control register of command field should be enabled
to initialize the time value stored in the Command Elapsed Time registers to be the local time (Transmit Enable TE bit in Configuration 0 register must be enabled). The NC bit in the Control register will
disable itself when the time is initialized. The INSYNC bit in Status 0 register will enable when initialization is performed. The MAPPING bits in Configuration 0 register determines the interval
between SpaceWire Time-Code transmissions which is explained in detail in the section below.
The target time must be configured with time values from the initiator. The targets register space must
be configured and controlled through RMAP by an initiator to achieve time synchronisation. The target time synchronisation is explained in detail under the section initialization and synchronisation of
target through RMAP.
36.3.7 SpaceWire Time-Code
SpaceWire Time-Codes are continuously transmitted from an initiator node (time master) to all slave
nodes. The transmission of the SpaceWire Time-Code is synchronized with the ET counter in the initiator node. The six bits of the Time-Code time information correspond to six bits of the local ET
counter (MAPPING bits in Configuration 0 register determines its exact mapping and interval
between SpaceWire Time-Code transmissions). Value of 0b00000 for MAPPING bits in Configuration 0 register will send SpaceWire Time-Code at every Second. When the value is 0b00001 SpaceWire Time-Codes are sent at every 0.5 Seconds interval and so on (maximum value of MAPPING can
be 0b11111 but this value cannot be more than the number of bits implemented as fine time). The ET
bits with lower weights than the size bits mapped to Time Codes time information bits are all zero at
time of SpaceWire Time-Codes transmission. The Table below shows an example Local ET
counter and Mapping. If the Coarse time is 32 bits and Fine time is 24 bits and mapping value is 6
then 0 to 31 is coarse(32 bits), 32 to 55 is fine time and mapped SpaceWire Time-Code is 32 to 37.
Table 514. Example Local ET counter with Mapping values
0
25 26 27 28 29 30 31 32 33 34 35 36 37 38
Mapping Values
0
If the Mapping value is 6 then the mapped SpaceWire
Time-Codes is 32 to 37
If the Mapping value is 0 then the mapped SpaceWire
Time-Codes is 26 to 31
If the Mapping value is 5 then the mapped SpaceWire
Time-Codes is 31 to 36
1
2
3
4
5
6
7
55
24
32 33 34 35 36 37
26 27 28 29 30 31
31 32 33 34 35 36
If the Mapping value is 7 then the mapped SpaceWire
Time-Codes is 33 to 38
33 34 35 36 37 38
36.3.8 Initialization and synchronisation of target through RMAP
An initiator must provide the time values and set the target in order to get the time synchronized. The
below text explains how an initiator can synchronise the target.
The SPWTC in Control register of initiator core component should be configured initially with a
SpaceWire Time-Code value at which the time message needed to be transferred. When the SpaceWire Time-Code generated internally using the ET counter matches the SPWTC in Control register a
Time Message TM interrupt will be generated (TME bit Time Message Enable should be enabled in
GR740-UM-DS, Jun 2016, Version 1.4
406
www.cobham.com/gaisler
GR740
the Interrupt Enable register). Based on this interrupt the local time (ET counter) in initiator should be
accessed from the Datation registers and used to calculate the time message needed to be transmitted.
•
Time message generation
The Time message transmitted using RMAP should be an exact mapping of the Command field
(explained under Registers section). The Time message transmitted should write the Command field
available in target. Control register available in Command field specify weather the target should be
initialized or synchronized, at which SpaceWire Time-Codes it should happen (synchronization
event) and details of coarse and fine time available in the time message. The New code NC bit available in Control register should be enabled and if the target should be initialized then Init Sync IS bit in
Control register must be enabled otherwise target will be synchronized.
The Command Elapsed Time in time message are calculated from the local time (ET counter) available in the initiator. The local time can be obtained by reading the Datation Field of initiator component. While reading the Datation registers always the total implemented coarse time and fine time
must be read in order (from 0 till the implemented Datation Elapsed Time registers). The DPF of
Datation Preamble Field register gives the coarse and fine time implemented which gives the total
local ET counter (coarse + fine width).
The current implementation has 32 bit coarse and 24 bit fine time then it is enough to access the first
two Datation Elapsed Time registers (0 and 1). The 32 bits of Datation Elapsed Time 0 and only the
most significant 24 bits (31 to 8) of Datation Elapsed Time 1 registers (32 + 24 =56 bits) represents
the local time. These 56 bits only be used for Command Elapsed time (time message) calculation.
The SpaceWire Time-Codes at which the Time Message interrupt generated is embedded in the local
ET counter. The Command Elapsed time which is transmitted as time message should be an incremented time value of this SpaceWire Time-Code and Command Elapsed time bits with lower weights
than the size bits mapped to SpaceWire Time-Code time information bits are all must be zero.
The incremented time value is to make the initialization or synchronisation of time message in target
will happen after the reception of qualifying SpaceWire Time-Codes. The qualifying SpaceWire
Time-Code is embedded in the Command Elapsed time (part of time message) sent from initiator.
This qualifying SpaceWire Time-Code value should also be written in the SPWTC in Control section
of the time message.
•
Time qualification in target
In target, the Command field will contain the time message when it is written by the initiator through
RMAP. When the SPWTC of Control register in Command field matches with a received SpaceWire
Time-Code then initialization or synchronization will occur (according to NC bit and IS bit in the
Control register) to the local ET counter of the target SPWTDP component. When the local ET
counter is initialized or synchronized the NC bit in the control register will disable itself. The
INSYNC bit in Status 0 register will enable when initialization is performed specifying the target is
initialized. Initialization completely writes time message values into the implemented local Elapsed
time counter and synchronisation verifies whether the time message Command Elapsed Time and
local Elapsed Time counter matches till the mapped SpaceWire Time-Code level (with a tolerance of
previous value) and only modifies the local Elapsed Time if their is a mismatch. If the target is not
implemented with jitter and mitigation unit then the synchronisation forces the target time (ET
counter) with the time message received.
For example, the initiator can create time message exactly at 0x00000001 coarse time and 0x040000
fine time (32 bit coarse time and 24 bit fine time, mapping value of 6 i.e. 64 SpaceWire Time-Codes
per second, time message is generated at 0b000001 SpaceWire Time-Code), the value in the time
message to be sent to the target can be coarse time 0x00000002 and 0x040000 fine time, (32 bit
coarse time and 24 bit fine time, mapping value of 6, time message is qualified at the next reception of
0b000001 SpaceWire Time-Code, i.e. after a second). Both SPWTC in Control registers available in
the initiator and target can be 0b000001 for this example. The time is synchronized after a second in
this example. Depending on the frequency of SpaceWire Time-Codes and data link rate several different combination of ways to achieve time synchronisation is possible.
GR740-UM-DS, Jun 2016, Version 1.4
407
www.cobham.com/gaisler
GR740
36.3.9 Latency measurement using Time-Stamps
The incoming and outgoing SpaceWire Distributed Interrupts are time stamped in initiator and target.
The initiator calculates latency based on these time stamp values. The time stamped values in target
are accessed from initiator through RMAP. The Latency Enable LE bit in Configuration 0 register
must be enabled between the two nodes in the SpaceWire network for which the latency is to be calculated. The core supports 32 distributed interrupts and acknowledgement (Interrupt and acknowledgement numbers 0 to 31). The distributed interrupt transmission from initiator (which is the origin for
latency calculation) is controlled by a mask register STM available in Configuration 3 register and
SpaceWire time code register TSTC available in Time-Stamp SpaceWire Time-Code and Preamble
Field Tx register, these registers specifies how often and at which time code distributed interrupt is
transmitted and time stamping is performed.
The time stamping can be performed in two methods (only Interrupts or Interrupts and Acknowledgement), the DI bit in Configuration 3 register of SPWTDP component in target should be configured to
specify which type of method is used. If only distributed interrupts (no acknowledgement) are used
then DI bit should be 0. The transmitted and received distributed interrupts INTX and INRX in the
Configuration 0 registers of both initiator and target must be configured with the interrupt number
which will be used for the latency measurement. For example if the INTX in initiator Configuration 0
is configured with 0b00100 then the target INRX should be configured with the same value. Similarly
if the INTX in target Configuration 0 is configured to be 0b00101 then the initiator INRX should be
configured with the same value. Initially initiator sends a distributed interrupt when the conditions are
matched (STM and TSTC registers match) and when the target received this distributed interrupt it
will send another interrupt which will be received by the initiator. At each end transmission and reception is time stamped (current local time is stored in Time Stamp registers) and interrupt transmitted is
INTX and received interrupt is checked whether it received INRX.
If both distributed interrupts and acknowledgement method is to be used then DI bit should be 1. The
transmitted and received distributed interrupts INTX and INRX in the Configuration 0 registers of
both initiator and target can have the same interrupt number (the acknowledgement number for a particular interrupt will be same as interrupt number). Similar to the previous method at each end transmission and reception is time stamped which will be used for latency calculations.
The Latency calculation can be started in initiator based on DIR (distributed interrupt received) interrupt available in Interrupt Status register (the interrupt should be enabled in the Interrupt Enable register). The latency is calculated form the time stamp registers based on the equation explained below
Latency = ((initiator time stamp Rx - initiator time stamp Tx) - (target time stamp Tx - target time
stamp Rx)) /2
By calculating the Latency value repeatedly (at least for about 128 times, more number of times provides increased accuracy) and taking an average of it will provide the final latency value. The initiator
should transfer the latency correction information to the Latency Field registers in the target by means
of RMAP transfer. When the latency values are written it will be adjusted to local time in the target.
36.3.10 External Datation
The core provides external datation services, there are four external datation services implemented
which can time stamp the Elapsed Time counter when the conditions for a respective event (time
stamping) occurs. The event on which time stamp must occur is configurable individually (using the
respective mask registers EDMx and also a dedicated mask bit is available for each of the input
events) for all the external datation services.
Each of the four external datation services implemented has its own mask EDMx, status EDS and
time EDxETx registers. (here the x suffix represent 0, 1, 2 and 3 respect to individual registers available)
All the external datation services share the same event inputs (32 inputs).
GR740-UM-DS, Jun 2016, Version 1.4
408
www.cobham.com/gaisler
GR740
The table below describes the inputs connected.
Table 515. Input Events on which time stamp occurs.
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10
9
8
7
6
5
4
3
2
1
0
irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq irq LS irq irq irq irq irq irq irq irq irq irq
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11
9 8 7 6 5 4 3 2 1 0
31: 11
The Interrupt lines (11 to 31) are connected as input events. The respective mask bit must be enabled
to time stamp on this event and when the condition matches (Interrupt occurs) the time stamp values
are available in respective EDxETx registers.
10
Latch-Save (LS) : Input 10 is special case.
MIL-STD-1553B controller RTSYNC event.
The Elapsed Time is continuously latched when a valid command is detected by the controller (the
corresponding mask register bit 10 must be enabled).
When RTSYNC event is reported by the MIL-STD-1553B controller the latched value will remain
same (saved at when the previous latch condition met) and all the mask bit previously enabled will
be cleared. The RTSYNC event is necessary to finalize the condition match.
9: 0
The Interrupt lines (0 to 9) are connected as input events. The respective mask bit must be enabled to
time stamp on this event and when the condition matches (Interrupt occurs) the time stamp values
are available in respective EDxETx registers.
Any condition match for a particular external datation service will clear its respective mask register
EDMx (clears all the mask bits and must be set again inorder to achieve an another time stamp). An
output pulse is also produced when conditions are met (respective TDP controller datation pulse output is driven high for a clock period). The EDS bit in Status Register 0 will go high when the condition matches and cleared when the latched elapsed time is read. The purpose of this status register is to
ensure that all the implemented coarse and fine time are read. Reading the lowest implemented fine
time makes the status register to go low.
36.4
Registers
The core is programmed through registers mapped into AMBA APB address space.
Table 516.Registers
APB address offset
Register
0x000-0x00F
Configuration Field
0x000
Configuration 0
0x004
Configuration 1
0x008
Configuration 2
0x00C
Configuration 3
0x010 - 0x01F
Status Field
0x010
Status 0
0x014
Status 1
0x018
RESERVED
0x01C
RESERVED
APB address offset
Register
0x020 - 0x03F
Command Field
0x020
Control
0x024
Command Elapsed Time 0
0x028
Command Elapsed Time 1
0x02C
Command Elapsed Time 2
GR740-UM-DS, Jun 2016, Version 1.4
409
www.cobham.com/gaisler
GR740
APB address offset
Register
0x030
Command Elapsed Time 3
0x034
Command Elapsed Time 4
0x038
RESERVED
0x03C
RESERVED
0x040 - 0x05F
Datation Field
0x040 - 0x05F
Datation Field
0x040
Datation Preamble Field
0x044
Datation Elapsed Time 0
0x048
Datation Elapsed Time 1
0x04C
Datation Elapsed Time 2
0x050
Datation Elapsed Time 3
0x054
Datation Elapsed Time 4
0x058
RESERVED
0x05C
RESERVED
0x060 - 0x09F
Time-Stamp Field
0x060
Time-Stamp Preamble Field Rx
0x064
Time-Stamp Elapsed Time 0 Rx
0x068
Time-Stamp Elapsed Time 1 Rx
0x06C
Time-Stamp Elapsed Time 2 Rx
0x070
Time-Stamp Elapsed Time 3 Rx
0x074
Time-Stamp Elapsed Time 4 Rx
0x078
RESERVED
0x07C
RESERVED
0x080
Time-Stamp SpaceWire Time-Code and Preamble Field Tx
0x084
Time-Stamp Elapsed Time 0 Tx
0x088
Time-Stamp Elapsed Time 1 Tx
0x08C
Time-Stamp Elapsed Time 2 Tx
0x090
Time-Stamp Elapsed Time 3 Tx
0X094
Time-Stamp Elapsed Time 4 Tx
0x098
RESERVED
0x09C
RESERVED
0x0A0-0x0BF
Latency Field
0x0A0
Latency Preamble Field
0x0A4
Latency Elapsed Time 0
0x0A8
Latency Elapsed Time 1
0x0AC
Latency Elapsed Time 2
0x0B0
Latency Elapsed Time 3
0x0B4
Latency Elapsed Time 4
0x0B8
RESERVED
0x0BC
RESERVED
0x0C0
Interrupt Enable
0x0C4
Interrupt Status
0x0C8
Delay Count
0x0CC-0x0FF
RESERVED
0x100-0x18F
External Datation Field
GR740-UM-DS, Jun 2016, Version 1.4
410
www.cobham.com/gaisler
GR740
APB address offset
Register
0x100
External Datation 0 Mask
0x104
External Datation 1 Mask
0x108
External Datation 2 Mask
0x10C
External Datation 3 Mask
0x110-0x12F
External Datation 0 Time
0x110
External Datation 0 Preamble Field
0x114
External Datation 0 Elapsed Time 0
0x118
External Datation 0 Elapsed Time 1
0x11C
External Datation 0 Elapsed Time 2
0x120
External Datation 0 Elapsed Time 3
0x124
External Datation 0 Elapsed Time 4
0x128
RESERVED
0x12C
RESERVED
0x130-0x14F
External Datation 1 Time
0x150-0x16F
External Datation 2 Time
0x170-0x18F
External Datation 3 Time
0x190-0x1FF
RESERVED
Table 517.0x000 - CONF0 - Configuration 0
31
25 24 23
17 16 15 14 13 12
RESERVED
JE
RESERVED
0
0
0
0
r
rw
r
31: 25
LE AE
8
7
6
TD
-
5
4
-
MAPPING
SEL
0
0
0b00110
0
0b00
rw rw
r
rw
rw
rw
3
2
1
0
ME RE TE RS
0
0
0
0
rw rw rw rw
RESERVED
24
Jitter Correction Enable (JE) - (not usable in this implementation).
23: 17
RESERVED
16:
Latency Enable(LE) - To calculate latency between an initiator and target this bit must be enabled in
both of them.
15:
AMBA Interrupt Enable (AE) - The interrupts (explained in interrupt registers) in this core will generate an AMBA interrupt only when this bit is enabled.
14 13
RESERVED
12: 8
Mapping (MAP) - Defines mapping of SpaceWire Time-Codes versus CCSDS Time-code.
Value 0b00000 will send SpaceWire Time-Codes every Second,
Value 0b00001 will send SpaceWire Time-Codes every 0.5 Second,
Value 0b00010 will send SpaceWire Time-Codes every 0.25 Second,
Value 0b00011 will send SpaceWire Time-Codes every 0.125 Second
The maximum value it can take is 0b11000.
7:
Enable TDP (TD) - This is to indicate that the TDP provides SpaceWire Time-codes and Distributed
interrupts.
6:
RESERVED
5:
4
Select (SEL) - Select for SpaceWire Time-Codes and Distributed Interrupt transmission and reception, one of 0 through 3, (must always be 0b00 in this implementation).
3:
Mitigation Enable (ME) - (not usable in this implementation).
2:
Receiver Enable (RE) - Enabling this will make the core to act as target.
1
Transmit Enable (TE) - Enabling this will make the core to act as initiator.
The core can act only as an initiator or target, both TE and RE cannot be enabled at the same time.
0
Reset (RS) - Reset core. Makes complete reset when enabled, self clears itself (to disable).
GR740-UM-DS, Jun 2016, Version 1.4
411
www.cobham.com/gaisler
GR740
Table 518. 0x004 - CONF 1 - Configuration 1
31 30 29
0
-
FSINC
0
0
r
rw
31: 30
RESERVED
29: 0
Frequency synthesizer (FSINC) - Increment value of the Frequency Synthesizer which is added to
the counter every system clock cycle. It defines the frequency of the synthesized reference time.
Refer the spreadsheet provided along with this document to obtain this value.
Table 519. 0x008 - CONF 2 - Configuration 2
31
8
7
0
CV
ETINC
0
0
rw
rw
31: 8
Compensation Value (CV) - (not usable in this implementation)
7: 0
Elapsed Time Increment (ETINC) - Value of the Elapsed Time counter is to be incremented each
time when the Frequency Synthesizer wraps around.
Refer the spreadsheet provided along with this document to obtain this value.
Table 520. 0x00C - CONF3 - Configuration 3
31
22 21
16 15
-
11 10
9
DI
5
4
INRX
0
STM
-
INTX
0
0
0
0
0
0
r
rw
r
rw
rw
rw
31: 22
RESERVED
21: 16
SpaceWire Time-Code Mask (STM) - Mask For TSTC register available at Time-Stamp SpaceWire
Time-Code and Preamble Field Tx register.
Value all bits zero will send Distributed interrupts at all SpaceWire Time-Codes irrespective of any
values in TSTC register.
Value all ones will send Distributed interrupts at complete match of SpaceWire Time-Code with
TSTC register.
(only for initiator)
15: 11
RESERVED
10:
Distributed Interrupt (DI) - Distributed Interrupt method, when set interrupt and acknowledge mode
else only interrupt mode. (only for target)
9: 5
Interrupt Received (INRX) - The distributed interrupt number received by initiator or target.
4: 0
Interrupt Transmitted (INTX) - The distributed interrupt number transmitted by initiator or target.
Table 521. 0x010 - STAT 0 - Status Register 0
31 30 28 27
24 23 22
16 15 14 13
MA
-
EDS
-
FW
-
0
0
0
0
0
0
2
1
0
CW
8
7
-
3
LC
TCQ
INSYNC
0
0
0
0
0
r
31:
Mitigation available (MA) - Mitigation unit available
If value ’1’ Drift and Jitter mitigation unit available.
If value ’0’ Drift and Jitter mitigation unit not available.
In this implementation Mitigation unit is not available.
(only for target)
30: 28
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
412
www.cobham.com/gaisler
GR740
Table 521. 0x010 - STAT 0 - Status Register 0
27: 24
External Datation Status (EDS) - When conditions matched for external datation this bit will go
high. This bit will go low when all the implemented time values are read.
24: External Datation 0 Status bit
25: External Datation 1 Status bit
26: External Datation 2 Status bit
27: External Datation 3 Status bit
23
RESERVED
22: 16
Fine Width (FW) - Fine width of command CCSDS Time Code received. Calculated from Preamble
field of Command Register.
15: 14
RESERVED
13: 8
Coarse Width (CW) Coarse width of command CCSDS Time Code received, calculated from Preamble field of Command Register.
7: 3
RESERVED
2
Latency Corrected (LC) - Goes high when the latency value is written into latency registers in target
(only for target).
1
Time Message Qualified (TCQ)- Time message is qualified by SpaceWire Time-Codes.
0
In Sync (INSYNC) - In Synchronization at Time code level, enabled when time values are Initialized
or Synchronized.
Table 522. 0x014 - STAT 1 - Status Register 1
31 30 29
0
-
IV
0
0
r
r
31: 30
RESERVED
29: 0
Increment Variation (IV) - (not usable in this implementation)
Table 523. 0x020 - CTRL - Control
31 30 29
NC IS
0
24
-
23
16 15
0
SPWTC
CPF
0
0
0
0
rw rw
r
rw
rw
31:
New Command (NC) - New command is set to provide a new time value.
30:
Init or Sync (IS) - ’1’ Initialization of received time message
’0’ Synchronisation of received time message
(only for target).
29: 24
RESERVED
23: 16
Spacewire Time-code (SPWTC) - Spacewire Time-code value used for initialization and synchronisation.
In initiator the SpaceWire Time-Codes generated internally using the local ET counter matches this
register a Time Message TM interrupt will be generated which is used to send Time message over the
SpaceWire network.
In target this register should match the received SpaceWire Time-code for time qualification.
15: 0
Command Preamble Field (CPF) - The number of coarse and fine time available in Command
Elapsed Time registers should be mentioned in this field. Based on this preamble field the target will
initialize or synchronise the local ET counter (only for target).
Table 524. 0x024 - CET0 - Command Elapsed Time 0
31
0
CET0
0
GR740-UM-DS, Jun 2016, Version 1.4
413
www.cobham.com/gaisler
GR740
Table 524. 0x024 - CET0 - Command Elapsed Time 0
rw
31: 0
Command Elapsed Time 0 (CET0) - Initialize or Synchronise local ET counter value (0 to 31).
Table 525. 0x028 - CET1 - Command Elapsed Time 1
31
0
CET1
0
rw
31: 0
Command Elapsed Time 1 (CET1) - Initialize or Synchronise local ET counter value (32 to 63).
Table 526.0x02C - CET2 - Command Elapsed Time 2
31
0
CET2
0
rw
31: 0
Command Elapsed Time 2 (CET2) - Initialize or Synchronise local ET counter value (64 to 95).
Table 527.0x030 - CET0 - Command Elapsed Time 3
31
0
CET3
0
rw
31: 0
Command Elapsed Time 3 (CET3) - Initialize or Synchronise local ET counter value (96 to 127).
Table 528. 0x034 - CET4 - Command Elapsed Time 4
31
24 23
0
CET4
RESERVED
0
0
rw
r
31: 24
Command Elapsed Time 4 (CET4) - Initialize or Synchronise local ET counter value (128 to 135).
23: 0
RESERVED
Table 529.0x040 - DPF - Datation Preamble Field
31
16 15
0
RESERVED
DPF
0
0x2f00
r
r
31: 16
RESERVED
15: 0
Datation Preamble Field (DPF) - The number of coarse and fine time implemented can be obtained
from this Preamble Field.
Table 530.0x044 - DET0 - Datation Elapsed Time 0
31
0
DET0
0
r
GR740-UM-DS, Jun 2016, Version 1.4
414
www.cobham.com/gaisler
GR740
Table 530.0x044 - DET0 - Datation Elapsed Time 0
Datation Elapsed Time 0 (DET0) - CCSDS Time Code value (0 to 31) of local ET counter
value.
31: 0
Table 531.0x048 - DET1 - Datation Elapsed Time 1
31
0
DET1
0
r
Datation Elapsed Time 1 (DET1) - CCSDS Time Code value (32 to 63) of local ET counter
value.
31: 0
Table 532.0x04C - DET2 - Datation Elapsed Time 2
31
0
DET2
0
r
Datation Elapsed Time 2 (DET2) - CCSDS Time Code value (64 to 95) of local ET counter
value.
31: 0
All registers are only readable.
Table 533.0x050 - DET3 - Datation Elapsed Time 3
31
0
DET3
0
r
Datation Elapsed Time 3 (DET3) - CCSDS Time Code value (96 to 127) of local ET counter
value.
31: 0
Table 534.0x054 - DET4 - Datation Elapsed Time 4
31
24 23
0
DET4
RESERVED
0
0
r
r
31: 24
Datation Elapsed Time 4 (DET4) - CCSDS Time Code value (128 to 135) of local ET
counter value.
23: 0
RESERVED
Table 535. 0x060 - TRPFRX - Time-Stamp Preamble Field Rx
31
16 15
RESERVED
0
TRPF
0
0x2f00
r
r
31: 16
RESERVED
15: 0
Time stamp Preamble Field (TRPF) - The number of coarse and fine time implemented can be
obtained from this Preamble Field.
Table 536. 0x064 - TR0 - Time Stamp Elapsed Time 0 Rx
31
0
TR0
GR740-UM-DS, Jun 2016, Version 1.4
415
www.cobham.com/gaisler
GR740
Table 536. 0x064 - TR0 - Time Stamp Elapsed Time 0 Rx
0
r
31: 0
Time Stamp Elapsed Time 0 Rx (TR0) - Time stamped local ET value (0 To 31) when distributed interrupt
received.
Table 537. 0x068 - TR1 - Time Stamp Elapsed Time 1 Rx
31
0
TR1
0
r
31: 0
Time Stamp Elapsed Time 1 Rx (TR1) - Time stamped local ET value (32 to 63) when distributed interrupt
received.
Table 538. 0x06C - TR2 - Time Stamp Elapsed Time 2 Rx
31
0
TR2
0
r
31: 0
Time Stamp Elapsed Time 2 Rx (TR2) - Time stamped local ET value (64 to 95) when distributed interrupt
received.
Table 539. 0x070 - TR3- Time Stamp Elapsed Time 3 Rx
31
0
TR3
0
r
31: 0
Time Stamp Elapsed Time 3 Rx (TR3) - Time stamped local ET value (96 to 127) when distributed interrupt
received.
Table 540. 0x074 - TR4 - Time Stamp Elapsed Time 4 Rx
31
24 23
0
TR4
RESERVED
0
0
r
r
31: 24 Time Stamp Elapsed Time 4 Rx (TR4) - Time stamped local ET value (128 to 135) when distributed interrupt received.
23: 0
RESERVED
Table 541. 0x080 - TTPFTX - Time-Stamp SpaceWire Time-Code and Preamble Field Tx
31
24 23
16 15
0
TSTC
RESERVED
TTPF
0
0
0x2f00
rw
r
r
31: 24
Time stamp time code (TSTC) - Time stamp on this time-code value, used for time stamping when
this register matched with SpaceWire Time-Codes. The mask for this matching is available in configuration register 3.(only for initiator)
23: 16
RESERVED
15: 0
Time stamp Preamble Field (TTPF) - The number of coarse and fine time implemented can be
obtained from this Preamble Field.
GR740-UM-DS, Jun 2016, Version 1.4
416
www.cobham.com/gaisler
GR740
Table 542. 0x084 - TT0 - Time Stamp Elapsed Time 0 Tx
31
0
TT0
0
r
Time Stamp Elapsed Time 0 Tx (TT0) - Time stamped local ET value (0 to 31) when distributed
interrupt transmitted.
31: 0
Table 543. 0x088 - TT1 - Time Stamp Elapsed Time 1 Tx
31
0
TT1
0
r
Time Stamp Elapsed Time 1 Tx (TT1) - Time stamped local ET value (32 to 63) when distributed
interrupt transmitted.
31: 0
Table 544. 0x08C - TT2 - Time Stamp Elapsed Time 2 Tx
31
0
TT2
0
r
31: 0
Time Stamp Elapsed Time 2 Tx (TT2) - Time stamped local ET value (64 to 95) when distributed interrupt
transmitted.
Table 545. 0x090 - TT3 - Time Stamp Elapsed Time 3 Tx
31
0
TT3
0
r
Time Stamp Elapsed Time 3 Tx (TT3) - Time stamped local ET value (96 to 127) when distributed
interrupt transmitted.
31: 0
Table 546. 0x094 - TT4 - Time Stamp Elapsed Time 4 Tx
31
24 23
0
TT4
RESERVED
0
0
r
r
31: 24 Time Stamp Elapsed Time 4 Tx (TT4) - Time stamped local ET value (128 to 135) when distributed
interrupt transmitted.
23: 0
RESERVED
Table 547. 0x0A0 - LPF- Latency Preamble Field
31
16 15
0
RESERVED
LPF
0
0x2f00
r
r
31: 16
RESERVED
15: 0
Latency Preamble Field (LPF) - The number of coarse and fine time implemented can be obtained
from this Preamble Field.(only for target)
GR740-UM-DS, Jun 2016, Version 1.4
417
www.cobham.com/gaisler
GR740
Table 548. 0xA4 - LE0 -Latency Elapsed Time 0
31
0
LE0
0
rw
31: 0
Latency Elapsed Time Value 0 (LE0) - Latency Value (0 to 31) written by initiator.(only for target)
Table 549. 0xA8 - LE1 -Latency Elapsed Time 1
31
0
LE1
0
rw
31: 0
Latency Elapsed Time Value 1 (LE1) - Latency Value (32 to 63) written by initiator.(only for target)
Table 550. 0xAC - LE2 -Latency Elapsed Time 2
31
0
LE2
0
rw
31: 0
Latency Elapsed Time Value 2 (LE2) - Latency Value (64 to 95) written by initiator.(only for target)
Table 551. 0xB0 - LE3 -Latency Elapsed Time 3
31
0
LE3
0
rw
31: 0
Latency Elapsed Time Value 3 (LE3) - Latency Value (96 to 127) written by initiator.(only for target)
Table 552. 0xB4 - LE4 -Latency Elapsed Time 4
31
24 23
0
LE4
RESERVED
0
0
rw
r
31: 24 Latency Elapsed Time Value 4 (LE4) - Latency Value (128 to 135) written by initiator. (only for target)
23: 0
RESERVED
GR740-UM-DS, Jun 2016, Version 1.4
418
www.cobham.com/gaisler
GR740
Table 553. 0x0C0 - IE - Interrupt Enable
31
10
-
9
8
7
EDIE3
EDIE2
EDIE1
6
5
4
EDIE0
DITE
DIRE
RESERVED
0
r
rw
3
TTE
2
1
TME
TRE
31: 10
RESERVED
9
External Datation Interrupt Enable 3 (EDIE3)
8
External Datation Interrupt Enable 2 (EDIE2)
7
External Datation Interrupt Enable 1 (EDIE1)
6
External Datation Interrupt Enable 0 (EDIE0)
5
Distributed Interrupt TransmittedInterrupt Enable (DITE)
4
Distributed interrupt Received Received Interrupt Enable(DIRE)
3
Time-Code Transmitted Interrupt Enable (TTE) - SpaceWire Time-Code Transmitted Interrupt
Enable (only for initiator)
0
SE
2
Time Message transmit Interrupt Enable (TME) - (only for initiator)
1
Time-Code Received Interrupt Enable (TRE) - SpaceWire Time-Code Received Interrupt Enable
(only for target)
0
Sync Interrupt Enable (SE) (only for target)
Table 554. 0xC4 - IS -Interrupt Status
31
10
RESERVED
9
8
7
EDI3
EDI2
EDI1
6
EDI0
5
4
DIT
DIR
0
0
r
wc
GR740-UM-DS, Jun 2016, Version 1.4
419
3
TT
2
TM
1
TR
0
S
www.cobham.com/gaisler
GR740
Table 554. 0xC4 - IS -Interrupt Status
31: 10
RESERVED
9
External Datation Interrupt 3 (EDI3) - Generated when conditions for External Datation 3 is
matched.
8
External Datation Interrupt 2 (EDI2) - Generated when conditions for External Datation 2 is
matched.
7
External Datation Interrupt 1 (EDI1) - Generated when conditions for External Datation 1is
matched.
6
External Datation Interrupt 0 (EDI0) - Generated when conditions for External Datation 0 is
matched.
5
Distributed Interrupt Transmitted (DIT) - Generated when distributed interrupt is transmitted
(Latency calculation should be enabled)
4
Distributed interrupt Received (DIR) - Generated when distributed interrupt is Received (Latency
calculation should be enabled)
3
Time-Codes Transmitted (TT) - Generated when SpaceWire Time-Codes is transmitted (only for
initiator)
2
Transmit Time Message (TM) - Generated when the conditions for transmitting time message
occurred, based on this time message should be transmitted from initiator (only for initiator)
1
Time-Code Received (TR) - Generated when SpaceWire Time-Code is received (only for target)
0
Target initialized or synchronized (S) - Generated when the target is initialized or synchronized with
initiator (only for target)
Table 555. 0xC8 - DC - Delay Count
31
15 14
0
-
DC
0
0x7FFF
r
r
31: 15
RESERVED
14: 0
Delay Count (DC) - Delay induced between SpaceWire Time-Codes and Distributed Interrupt transmission in system clock units. The delay introduced is the value in this register multiplied by the
system clock.
(only for initiator)0x7FFF
Table 556. 0x100 - EDM0 - External Datation 0 Mask
31
0
EDM0
0x00000000
rw
31: 0
External Datation Mask (EDM0) - External datation can be enabled by writing ‘1’ into the bit for
that corresponding external input. When conditions are matched the Elapsed Time will be latched.
The latched values are available at External Datation 0 Time Register.
All the mask bits will go low after any one of the conditions with respect to the enabled mask bits.are
matched.
Table 557. 0x110 - EDPF0 - External Datation 0 Preamble Field
31
16 15
0
-
EDPF0
0
0x2f00
r
r
31: 16
RESERVED
15: 0
External Datation Preamble Field (EDPF0) - The number of coarse and fine time implemented can
be obtained from this Preamble Field.
GR740-UM-DS, Jun 2016, Version 1.4
420
www.cobham.com/gaisler
GR740
Table 558.0x114 - ED0ET0 - External Datation 0 Elapsed Time 0
31
0
ED0ET0
0
r
31: 0
External Datation Elapsed Time 0 (ED0ET0) - Latched CCSDS Time Code value (0 to 31) of local
ET counter.
Table 559.0x118 - ED0ET1 - External Datation 0 Elapsed Time 1
31
0
ED0ET1
0
r
31: 0
External Datation Elapsed Time 1 (ED0ET1) - Latched CCSDS Time Code value (32 to 63) of local
ET counter.
Table 560.0x11C - ED0ET2 - External Datation 0 Elapsed Time 2
31
0
ED0ET2
0
r
31: 0
External Datation 0 Elapsed Time 2 (ED0ET2) - Latched CCSDS Time Code value (64 to 95) of
local ET counter.
Table 561.0x120 - ED0ET3 - External Datation 0 Elapsed Time 3
31
0
ED0ET3
0
r
31: 0
External Datation 0 Elapsed Time 3 (ED0ET3) - Latched CCSDS Time Code value (96 to 127) of
local ET counter.
Table 562. 0x124 - ED0ET4 - External Datation 0 Elapsed Time 4
31
24 23
0
ED0ET4
RESERVED
0
0
r
r
31: 24
External Datation 0 Elapsed Time 4 (ED0ET4) - Latched CCSDS Time Code value (128 to 135) of
local ET counter.
23: 0
RESERVED
Note:
Reserved register fields should be written as zeroes and masked out on read.
The registers which are not mentioned either as only for initiator or target are used in both initiator
and target.
The Definition of External Datation 1 Mask, External Datation 2 Mask and External Datation 3 Mask
registers are exactly same as External Datation 0 Mask Register.
The Definition of External Datation 1 Time, External Datation 2 Time and External Datation 3 Time
registers are exactly same as External Datation 0 Time Registers (i.e. External Datation 0 Preamble
Field and External Datation 0 Elapsed Time 0,1,2,3,4).
GR740-UM-DS, Jun 2016, Version 1.4
421
www.cobham.com/gaisler
GR740
GR740-UM-DS, Jun 2016, Version 1.4
422
www.cobham.com/gaisler
GR740
37
AMBA AHB controller with plug&play support
37.1
Overview
The AMBA AHB controller is a combined AHB arbiter, bus multiplexer and slave decoder according
to the AMBA 2.0 standard. Each AHB bus in the system has one AHB controller.
MASTER
MASTER
AHBCTRL
ARBITER/
DECODER
SLAVE
SLAVE
Figure 53. AHB controller block diagram
37.2
Operation
37.2.1 Arbitration
The AHB controller supports round-robin arbitration. In round-robin mode, priority is rotated one
step after each AHB transfer. If no master requests the bus, the last owner will be granted (bus parking).
37.2.2 Decoding
Decoding (generation of HSEL) of AHB slaves is done using the plug&play method explained in the
GRLIB User’s Manual. A slave can occupy any binary aligned address space with a size of 1 - 4096
MiB. A specific I/O area is also decoded, where slaves can occupy 256 byte - 1 MiB. Access to
unused addresses will cause an AHB error response.
37.2.3 Plug&play information
The plug&play information is mapped on a read-only address area on each AHB bus except the Master I/O AHB bus. See the memory map in section 2.3 for the Plug&play area base addresses of the
buses in the system.
The master information is placed on the first 2 KiB of the block (0xFFFFF000 - 0xFFFFF800 for the
Processor AHB bus), while the slave information is placed on the second 2 KiB block. Each unit
occupies 32 bytes, which means that the area has place for 64 masters and 64 slaves. The address of
the plug&play information for a certain unit is defined by its bus index. The address for masters is
thus 0xFFFFF000 + n*32, and 0xFFFFF800 + n*32 for slaves.
GR740-UM-DS, Jun 2016, Version 1.4
423
www.cobham.com/gaisler
GR740
31
Identification Register
00
24 23
VENDOR ID
12 11 10 9
DEVICE ID
04
USER-DEFINED
08
USER-DEFINED
0C
USER-DEFINED
00
5 4
VERSION
0
IRQ
BAR0 10
HADDR
ADDR
00
P C
MASK
MASK
TYPE
BAR1 14
ADDR
00
P C
MASK
TYPE
BAR2 18
ADDR
00
P C
MASK
TYPE
BAR3 1C
ADDR
00
P C
MASK
TYPE
Bank Address Registers
31
20 19 18 17 16 15
P = Prefetchable
C = Cacheable
4 3
0
TYPE
0001 = APB I/O space
0010 = AHB Memory space
0011 = AHB I/O space
Figure 54. AHB plug&play information record
GR740-UM-DS, Jun 2016, Version 1.4
424
www.cobham.com/gaisler
GR740
38
AMBA AHB/APB bridge with plug&play support
38.1
Overview
The AMBA AHB/APB bridge is a APB bus master according the AMBA 2.0 standard. The system
contains three AHB/APB bridges. Two on the Slave I/O AHB bus and one on the Debug AHB bus.
AHB/APB Bridge
AHB BUS
APBO[0]
APB SLAVE
AHBSI
APBO[n]
AHB Slave
Interface
AHBSO[n]
APB SLAVE
•••
APBI
Figure 55. AHB/APB bridge block diagram
38.2
Operation
38.2.1 Decoding
Decoding (generation of PSEL) of APB slaves is done using the plug&play method explained in the
GRLIB IP Library User’s Manual. A slave can occupy any binary aligned address space with a size of
256 bytes - 1 MiB. Writes to unassigned areas will be ignored, while reads from unassigned areas will
return an arbitrary value. AHB error responses will never be generated.
38.2.2 Plug&play information
The plug&play information is mapped on a read-only address area at the top 4 KiB of each bridge’s
address space. Each plug&play block occupies 8 bytes. The address of the plug&play information for
a certain unit is defined by its bus index. If the bridge is mapped on AHB address 0xF0000000, the
address for the plug&play records is thus 0xF00FF000 + n*8.
31
APB Plug&play record
24 23
VENDOR ID
0x00
12 11 10 9
DEVICE ID
ADDR
0x04
31
C/P
20 19
00
5
Configuration word
IRQ
MASK
16 15
0
4
VERSION
BAR
TYPE
4 3
0
Figure 56. APB plug&play information
GR740-UM-DS, Jun 2016, Version 1.4
425
www.cobham.com/gaisler
GR740
39
Electrical description
39.1
Absolute maximum ratings
Estimated lifetime is greater than 20 years (TBC). Exposure to absolute maximim rating conditions
for extended periods may affect reliability and greater stress may cause permanent damage to the
device.
Table 563.Absolute maximum DC ratings
39.2
Parameter
Max
Unit
VDD12 relative to GND
TBD
V
VDIG25 relative to VSS25
TBD
V
VDIG33 relative to VSS33
TBD
V
Operating conditions
Table 564 below shows recommended DC operating conditions.
Table 564.Recommended DC operating conditions
Parameter
Symbol
Digital core supply
voltage
I/O bank supply
Analog PLL supply
39.3
Conditions
Minimum
Typical
Maximum
Unit
VDD12
TBD
1.200
TBD
V
VDIG25
TBD
2.500
TBD
V
VDIG33
TBD
3.300
TBD
V
VDDPLLA
TBD
1.200
TBD
V
1.200
TBD
Digital PLL supply
VDDPLLD
TBD
HIGH (logic 1) level
input switching voltage
(LVCMOS)
VIH
TBD
TBD
V
LOW (logic 0) level
input switching voltage
(LVCMOS)
VIL
TBD
TBD
V
Input voltages, leakage currents and capacitances
Data for leakage currents and capacitances is not included in this datasheet. Cobham Gaisler plans to
provide a (IBIS/SPICE) model of the device for users that want to do custom PCB designs.
GR740-UM-DS, Jun 2016, Version 1.4
426
www.cobham.com/gaisler
GR740
39.4
Power supplies
The device has the following power domains:
Table 565.Power domains
Name
Description
Voltage (V)
VDD12
Digital core supply
1.2
GND
Digital core ground
0
VDDPLLA
Analog PLL supply
1.2
VSSPLLA
Analog PLL ground
0
VDDPLLD
Digital PLL supply
1.2
VSSPLLD
Digital PLL ground
0
VDIG33
I/O bank supply
3.3
VSS33
I/O bank ground
0
VDIG25
I/O bank supply
2.5
VSS25
I/O bank ground
0
Notes
Used for LVDS IOs
Table 566 provides detailed specifications for the power supplies listed in table 565.
Table 566.Detailed power supply specifications
Symbol
Conditions
Minimum
Typical
Maximum
TBD
Unit
V
V
V
V
V
V
39.4.1 Power sequence
TBD
39.5
AC characteristics
39.5.1 Test conditions
For the SDRAM, Ethernet and PCI interfaces, timing is tested using a impedance matched transmission line setup. The propagation delays in the test fixture is calibrated out as part of the test procedure,
making the equivalent test load for the timing data into a pure 50 ohm resistive load. The interfaces
are tested with the pad drive strength set to maximum.
It is up to the end user to translate the timing data to data relevant for the system. An IBIS model of
the drivers can be provided to aid in this process.
Note: This section will be expanded in a future release of the data sheet.
39.5.2 Clocks
Table 567 below summarizes required/recommended conditions for some of the design input clocks
that connect to on-chip PLLs. For the remaining clocks please see the interface-specific timing in the
subsections below.
GR740-UM-DS, Jun 2016, Version 1.4
427
www.cobham.com/gaisler
GR740
The AC characteristics presented are preliminary and will be updated based on static timing
analysis results and revised once devices have been manufactured and tested.
Table 567.Recommended AC operating conditions
Parameter
Symbol
Minimum
Nominal
Maximum
Unit
Note(s)
fsys_clk
TBD
50
TBD
MHz
1
40
50
60
%
TBD
50
TBD
MHz
40
50
60
%
TBD
100
TBD
MHz
MEM_EXTCLK duty cycle
40
50
60
%
MEM_CLK_IN frequency
TBD
100
TBD
Externally fed clocks
SYS_CLK frequency
SYS_CLK duty cycle
SPW_CLK frequency
fspw_clk
SPW_CLK duty cycle
MEM_EXTCLK frequency
fmem_extclk
MEM_CLK_IN duty cycle
1
1, 2
TBD
50
TBD
%
PCI_CLK frequency for 33 MHz
operation
fpci_clk=33MHz
33
33
33
MHz
3
PCI_CLK frequency for 66 MHz
operation
fpci_clk=66MHZ
66
66
66
MHz
3
TBD
50
TBD
%
125
125
125
MHz
TBD
50
TBD
%
40
250
TBD
MHz
10
400
TBD
MHz
PCI_CLK duty cycle
ETH*_GTXCLK frequency
feth_gtxclk
ETH*_GTXCLK duty cycle
Internally generated clocks
AMBA system clock frequency
fsoc_systemk
Internal SPW clock frequency
4
1 Min/max
values are maximum for on-chip PLL input, limits for the generated frequencies must also be satisfied.
Nominal frequency required for correct operation with PLL power-up configuration and will change if reconfigured.
2 Assuming MEM_CLKSEL has been set high so that MEM_EXTCLK is used
3 The PCI interface will not be fully compliant to the PCI specification. Interface will be characterised at a later date.
4 The minimum clock frequency for the on-chip system is determined by functional interface limitations and has not
been fully characterised. A clock frequency over 40 MHz is required for the Ethernet interfaces to function at gigabit
speeds.
39.5.3 Processor error mode signal timing
The timing waveforms and timing parameters are shown in figure 57 and are defined in table 568.
clk
tLEON4_0
tLEON4_0
errorn
Figure 57. Timing waveforms
Table 568.Timing parameters
Name
Parameter
Reference edge
tLEON4_0
clock to output delay
rising clk edge
1) This
Min
0
1)
Max
TBD
Unit
2)
ns
parameter is determined by static timing analysis and is not tested
GR740-UM-DS, Jun 2016, Version 1.4
428
www.cobham.com/gaisler
GR740
39.5.4 64-bit PC133 SDRAM Controller with Reeed-Solomon EDAC timing
The timing waveforms and timing parameters are shown in figure 58 and are defined in table 569.
mem_clk_in
tSDRAM11
mem_casn, mem_rasn
mem_wen, mem_sn[]
mem_dqm[]
write
nop
read
nop
nop
term
nop
nop
nop
tSDRAM11
mem_addr[]
tSDRAM12
tSDRAM14
tSDRAM13
mem_dq[]
tSDRAM15
Figure 58. Timing waveforms - SDRAM accesses
Table 569.Timing parameters - SDRAM accesses
Name
Parameter
Reference edge
Min
Max
Unit
tSDRAM11
clock to output delay
rising sdr_sdclk edge TBD
TBD
ns
tSDRAM12
clock to data output delay
rising sdr_sdclk edge TBD
TBD
ns
tSDRAM13
data clock to data tri-state delay
rising sdr_sdclk edge 0 1)
TBD 2)
ns
tSDRAM14
data input to clock setup
rising sdr_sdclk edge TBD
TBD
ns
tSDRAM15
data input from clock hold
rising sdr_sdclk edge TBD
TBD
ns
Notes:
1) This
parameter is guaranteed by design and is not tested
2) This
parameter is determined by static timing analysis and is not tested
GR740-UM-DS, Jun 2016, Version 1.4
429
www.cobham.com/gaisler
GR740
39.5.5 DSU signals timing
The timing waveforms and timing parameters are shown in figure 59 and are defined in table 570.
clk
dsu_active
tDSU0
dsu_en, break
tDSU1
tDSU0
tDSU2
Figure 59. Timing waveforms
Table 570.Timing parameters
Name
Parameter
Reference edge
tDSU0
clock to output delay
rising clk edge
tDSU1
input to clock hold
tDSU2
input to clock setup
Min
Max
Unit
1)
ns
0
1)
TBD
rising clk edge
-
3)
- 3)
ns
rising clk edge
- 3)
- 3)
ns
Notes:
1) This
parameter is guaranteed by design and is not tested
2) This
parameter is determined by static timing analysis and is not tested
3) The break and dsu_en signals are re-synchronized internally. These signals do not have to meet any setup or hold
requirements. As the dsu_en signal controls clock gating for the Debug AHB bus the signal’s value should be kept constant from power-up.
39.5.6 JTAG interface timing
The timing waveforms and timing parameters are shown in figure 60 and are defined in table 571.
tAHBJTAG0
tAHBJTAG1
jtag_tck
tAHBJTAG2
jtag_tdi, jtag_tms
tAHBJTAG4
tAHBJTAG3
jtag_tdo
Figure 60. Timing waveforms
Table 571.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tAHBJTAG0
clock period
-
100
-
ns
tAHBJTAG1
clock low/high period
-
40
-
ns
tAHBJTAG2
data input to clock setup
rising jtag_tck edge
12
-
ns
tAHBJTAG3
data input from clock hold
rising jtag_tck edge
TBD
-
ns
tAHBJTAG4
clock to data output delay
falling jtag_tck edge
0
TBD
ns
GR740-UM-DS, Jun 2016, Version 1.4
430
www.cobham.com/gaisler
GR740
39.5.7 Gigabit Ethernet Media Access Controller (MAC) w. EDCL timing
The timing waveforms and timing parameters are shown in figure 61 and are defined in table 572.
clock
outputs
tGRETH0
tGRETH0
inputs
tGRETH1
tGRETH2
Figure 61. Timing waveforms
Table 572.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tGRETHTXCLK0
Ethernet MII transmit clock
(eth*_txclk) period
-
40 1)
-
ns
tGRETHRXCLK0
Ethernet MII receive clock (eth*_rxclk) period
-
40 1), 3)
-
ns
tGRETHGTXCLK0
Ethernet GMII transmit clock
(eth*_gtxclk) period
-
8 1)
-
ns
tGRETHRXCLK1
Ethernet GMII receive clock period
(eth*_rxclk)
-
8 1), 3)
-
ns
tGRETH0MII
transmitter clock to output delay
rising (MII) clock edge
TBD
TBD
ns
tGRETH0GMII
transmitter clock to output delay
rising (GMII) clock edge
TBD
TBD
ns
tGRETH1MII/GMII
input to receiver clock hold
rising clock edge
TBD
-
ns
tGRETH2MII/GMII
input to receiver clock setup
rising clock edge
TBD
-
ns
1 The
_crs, _col, _mdio, _mdint inputs are re-synchronized internally. The signals do not have to meet any setup or hold
requirements.
2 The _mdio and _mdc outputs are low speed signals without any timing relationship with the _rxclk, _txclk or _gtxclk
signals.
3 eth*_rxclk is used in both MII and GMII mode, with different frequencies.
GR740-UM-DS, Jun 2016, Version 1.4
431
www.cobham.com/gaisler
GR740
39.5.8 SpaceWire router interface timing
The specifies the timing for the Spacewire router links that are interfaced via LVDS signals. The timing waveforms are shown in figure 62. Timing parameters are defined in table 573.
tSPW0
spw_clk
spw_txd, spw_txdn
spw_txs, spw_txsn
tSPW1
spw_rxd, spw_rxdn
spw_rxs, spw_rxsn
tSPW2
tSPW4
tSPW1
tSPW3
tSPW4
spw_txd, spw_txdn
spw_txs, spw_txsn
tSPW4
tSPW5
tSPW5
tSPW6
spw_rxd, spw_rxdn
tSPW5
spw_rxs, spw_rxsn
tSPW6
tSPW7
spw_txd, spw_txdn
spw_txs, spw_txsn
Figure 62. Timing waveforms
Table 573.Timing parameters
Name
Parameter
Reference edge
tSPW0
transmit clock period
-
Min
Max
Unit
20
20
ns
tSPW1
clock to output delay
-
-
-
not applicable
tSPW2
input to clock hold
-
-
-
not applicable
tSPW3
input to clock setup
-
-
-
not applicable
tSPW4
output data bit period
-
2.5
-
ns
tSPW5
input data bit period
-
2.5
-
ns
tSPW6
data & strobe edge separation
-
TBD 2) 3) -
ns
tSPW7
data & strobe output skew
-
-
ns
1)
TBD 3)
1) Incoming
spw_clk is used as input for PLL that generates internal SpaceWire clock. The data given is suitable for the
default PLL configuration.
2) Assuming
spacewire PLL used in nominal configuration
3)
Edge separation and skew limits refer to each pair of data/strobe signals separately. Global skew and separation over the
entire set of eight pairs is not specified.
GR740-UM-DS, Jun 2016, Version 1.4
432
www.cobham.com/gaisler
GR740
39.5.9 SpaceWire debug interface timing
This specifies the timing for the spacewire debug interface that shares pins with the slow interfaces.
The timing waveforms are shown in figure 62. Timing parameters are defined in table 573.
tSPW0
spw_clk
spwd_txd
spwd_txs
tSPWD1
spwd_rxd
spwd_rxs
tSPWD2
tSPWD4
tSPWD1
tSPWD3
tSPWD4
spwd_txd
tSPW4
spwd_txs
tSPWD5
tSPWD5
tSPWD6
spwd_rxd
tSPWD5
spwd_rxs
tSPWD6
tSPWD7
spwd_txd
spwd_txs
Figure 63. Timing waveforms
Table 574.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tSPWD1
clock to output delay
-
-
-
not applicable
tSPWD2
input to clock hold
-
-
-
not applicable
tSPWD3
input to clock setup
-
-
-
not applicable
tSPWD4
output data bit period
-
20
-
ns
tSPWD5
input data bit period
-
20
-
ns
tSPWD6
data & strobe edge separation
-
TBD 2) 3) -
ns
tSPWD7
data & strobe output skew
-
-
ns
TBD 3)
1) Incoming spw_clk is used as input for PLL that generates internal SpaceWire clock. The data given is suitable for the
default PLL configuration.
2) Assuming
3) Verified
spacewire PLL used in nominal configuration
by static timing analysis only, not tested
GR740-UM-DS, Jun 2016, Version 1.4
433
www.cobham.com/gaisler
GR740
39.5.10 PCI interface timing
The timing waveforms and timing parameters are shown in figure 64 and are defined in table 575.
pci_clk
pci_ad[ ], pci_cbe[ ],
pci_frame, pci_irdy,
pci_trdy, pci_stop,
pci_idsel, pci_devsel,
pci_perr, pci_serr,
pci_par, pci_int
tPCIFT0
pci_ad[ ], pci_cbe[ ],
pci_frame, pci_irdy,
pci_trdy, pci_stop,
pci_idsel, pci_devsel,
pci_perr, pci_serr,
pci_par
tPCIFT0
tPCIFT1
tPCIFT2
pci_gnt, pci_req
tPCIFT3
tPCIFT3
pci_gnt, pci_req
tPCIFT4
tPCIFT5
Figure 64. Timing waveforms
Table 575.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tPCICLK0
PCI clock period
-
15, 30
-
ns
tPCIFT0
clock to output delay
rising pci_clk edge
2
11 1)
ns
tPCIFT1
input to clock hold
rising pci_clk edge
0
-
ns
tPCIFT2
input to clock setup
rising pci_clk edge
8 1)
-
ns
tPCIFT3
clock to output delay
rising pci_clk edge
2
11 1)
ns
tPCIFT4
input to clock hold
rising pci_clk edge
0
-
ns
tPCIFT5
input to clock setup
rising pci_clk edge
8 1)
-
ns
1) Interface
is not fully compliant for PCI 2.3 66 MHz operation. Both input and output delay are relaxed compared to
PCI 2.3 specification for 66 MHz operation.
GR740-UM-DS, Jun 2016, Version 1.4
434
www.cobham.com/gaisler
GR740
39.5.11 MIL-STD-1553B / AS15531 interface timing
The timing waveforms and timing parameters are shown in figure 65 and are defined in table 576.
gr1553_clk
t1553BRM0
outputs
t1553BRM2
inputs
t1553BRM1
Figure 65. Timing waveforms
Table 576.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
t1553BRM0
clock to data output delay
rising gr1553_clk edge
0 1)
40 2)
ns
t1553BRM1
data input to clock setup
rising gr1553_clk edge
- 3)
- 3)
ns
t1553BRM2
data input from clock hold
rising gr1553_clk edge
- 3)
- 3)
ns
1) Guaranteed by design, not tested
2) Guaranteed by static timing analysis, not tested
3) The inputs are asynchronous to the clock and are internally resynchronized to gr1553_clk
GR740-UM-DS, Jun 2016, Version 1.4
435
www.cobham.com/gaisler
GR740
39.5.12 Fault-tolerant 8/16-bit PROM/IO memory interface timing
The timing waveforms and timing parameters are shown in figures 66 and 67, and are defined in table
577.
clk
tFTMCTRL0
promio_addr[]
tFTMCTRL1
tFTMCTRL1
prom_cen[]
tFTMCTRL2
tFTMCTRL2
promio_wen
tFTMCTRL3, tFTMCTRL4
promio_data[]
(output)
tFTMCTRL5
clk
promio_addr[]
prom_cen[]
tFTMCTRL6
tFTMCTRL6
promio_oen
tFTMCTRL7
tFTMCTRL8
promio_data[]
(input)
tFTMCTRL10
tFTMCTRL9
promio_brdyn
Figure 66. Timing waveforms - PROM accesses
GR740-UM-DS, Jun 2016, Version 1.4
436
www.cobham.com/gaisler
GR740
clk
tFTMCTRL0
promio_addr]
tFTMCTRL1
tFTMCTRL1
io_sn[]
tFTMCTRL2
tFTMCTRL2
promio_wen
tFTMCTRL3, tFTMCTRL4
promio_data[]
(output)
tFTMCTRL5
clk
promio_addr[]
io_sn[]
tFTMCTRL6
tFTMCTRL6
promio_oen
tFTMCTRL7
tFTMCTRL8
promio_data[]
(input)
tFTMCTRL10
tFTMCTRL9
promio_brdyn
Figure 67. Timing waveforms - I/O accesses
Table 577.Timing parameters - PROM and I/O accesses
Name
Parameter
tFTMCTRL0
address clock to output delay
tFTMCTRL1
Reference edge
Min
Max
Unit
0
2)
TBD
3)
ns
0
2)
TBD 3)
ns
rising clk edge
1)
clock to output delay
rising clk edge
1)
tFTMCTRL2
clock to output delay
rising clk edge 1)
0 2)
TBD 3)
ns
tFTMCTRL3
clock to data output delay
rising clk edge 1)
TBD 3)
TBD 3)
ns
tFTMCTRL4
clock to data non-tri-state delay
rising clk edge 1)
0 2)
TBD 3)
ns
tFTMCTRL5
clock to data tri-state delay
rising clk edge 1)
TBD 3)
TBD 3)
ns
tFTMCTRL6
clock to output delay
rising clk edge 1)
0 2)
TBD 3)
ns
tFTMCTRL7
data input to clock setup
rising clk edge 1)
TBD 3)
-
ns
tFTMCTRL8
data input from clock hold
rising clk edge 1)
-
-
ns
tFTMCTRL9
input to clock setup
rising clk edge 1)
TBD 3)
-
ns
tFTMCTRL10
input from clock hold
rising clk edge 1)
-
-
ns
1)
Timing values are relative to the internal clock for the PROM/IO memory controller.
2) Guaranteed by design, not tested
3) Verified by static timing analysis, not tested
GR740-UM-DS, Jun 2016, Version 1.4
437
www.cobham.com/gaisler
GR740
39.5.13 Watchdog signal timing
The timing waveforms and timing parameters are shown in figure 68 and are defined in table 578.
clk
tGPTIMER0
tGPTIMER1
wdog
Figure 68. Timing waveforms
Table 578.Timing parameters
Name
Parameter
Reference edge
tGPTIMER0
clock to output tri-state
rising clk edge
Min
Max
Unit
1)
TBD
ns
0
tGPTIMER1
clock to output delay
rising clk edge
0 1)
TBD
ns
1) Guaranteed by design, not tested
39.5.14 General Purpose I/O interface timing
The timing waveforms and timing parameters are shown in figure 69 and are defined in table 579.
clk
gpio[ ]
(output)
tGRGPIO0
tGRGPIO0
tGRGPIO1
tGRGPIO2
gpio[ ]
(output)
gpio[ ]
(input)
tGRGPIO3
tGRGPIO4
Figure 69. Timing waveforms
Table 579.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tGRGPIO0
clock to output delay
rising clk edge
0 1)
TBD 2)
ns
tGRGPIO1
clock to non-tri-state delay
rising clk edge
0 1)
TBD 2)
ns
tGRGPIO2
clock to tri-state delay
rising clk edge
0 1)
TBD 2)
ns
tGRGPIO3
input to clock hold
rising clk edge 3)
-
-
ns
tGRGPIO4
input to clock setup
rising clk edge 3)
-
-
ns
1) Guaranteed
2)
by design, not tested.
Verified by static timing analysis, not tested
3) The
gpio inputs are re-synchronized internally. The signals do not have to meet any setup or hold requirements.
GR740-UM-DS, Jun 2016, Version 1.4
438
www.cobham.com/gaisler
GR740
39.5.15 UART interface timing
The timing waveforms and timing parameters are shown in figure 70 and are defined in table 580.
AMBA clk
_txd[], _rtsn[]
tAPBUART0
_rxd[], _ctsn[]
tAPBUART1
tAPBUART0
tAPBUART2
Figure 70. Timing waveforms
Table 580.Timing parameters
Name
Parameter
Reference edge
Min
Max
Unit
tAPBUART0
clock to output delay
rising clk edge
0 1)
TBD 2)
ns
tAPBUART1
input to clock hold
rising clk edge 3)
-
-
ns
tAPBUART2
input to clock setup
rising clk edge 3)
-
-
ns
1) Guaranteed
2) Verified
3) The
by design, not tested.
by static timing analysis, not tested
_cstn and _rxd inputs are re-synchronized internally. These signals to not have to meet any setup or hold require-
ments.
GR740-UM-DS, Jun 2016, Version 1.4
439
www.cobham.com/gaisler
GR740
39.5.16 SPI controller timing
The timing waveforms and timing parameters are shown in figure 71 and are defined in table 581.
AMBA clk
spi_sck/miso/mosi/slvsel
(output)
tSPICTRL0
tSPCTRLI0
tSPICTRL1
tSPICTRL2
spi_sck/miso/mosi
(output)
spi_sck/miso/mosi/sel
(input)
tSPICTRL3
tSPICTRL4
Figure 71. Timing waveforms
Table 581.Timing parameters
Name
Parameter
Reference edge
tSPICTRL0
clock to output delay
rising clk edge
0
tSPICTRL1
clock to non-tri-state delay
rising clk edge
tSPICTRL2
clock to tri-state delay
tSPICTRL3
tSPICTRL4
1) Guaranteed
2) Verified
Min
Max
Unit
2)
ns
0 1)
TBD 2)
ns
rising clk edge
0 1)
TBD 2)
ns
input to clock hold
rising clk edge 3)
-
-
ns
input to clock setup
rising clk edge 3)
-
-
ns
1)
TBD
by design, not tested.
by static timing analysis, not tested
3) The
spi_sck/miso/mosi/spisel inputs are re-synchronized internally. The signals do not have to meet any setup or hold
requirements. However, the input to clock setup value restricts the maximum SPI frequency.
GR740-UM-DS, Jun 2016, Version 1.4
440
www.cobham.com/gaisler
GR740
39.5.17 CAN Controller interface timing
The timing waveforms and timing parameters are shown in figure 72 and are defined in table 582.
clk
tGRCAN0
cantx[]
tGRCAN2
canrx[]
tGRCAN3
tGRCAN1
tGRCAN3
Figure 72. Timing waveforms
Table 582.Timing parameters
Name
Parameter
Reference edge
tGRCAN0
clock to data output delay
rising clk edge
0
tGRCAN1
data input to clock setup
rising clk edge
tGRCAN2
data input from clock hold
tGRCAN3
clock to output delay
1) Guaranteed
2) Verified
Min
Max
Unit
2)
ns
0 1)
TBD 2)
ns
rising clk edge 3)
-
-
ns
rising clk edge 3)
-
-
ns
1)
TBD
by design, not tested.
by static timing analysis, not tested
3) The can inputs are re-synchronized internally. The signals do not have to meet any setup or hold requirements. However,
the input to clock setup value restricts the maximum SPI frequency.
GR740-UM-DS, Jun 2016, Version 1.4
441
www.cobham.com/gaisler
GR740
40
Mechanical description
40.1
Component and package
The device has a LGA625 package.
40.2
Package pinout diagram
Figure 73. Package footprint
Note: Figure will be updated in a later release of this advanced data sheet.
GR740-UM-DS, Jun 2016, Version 1.4
442
www.cobham.com/gaisler
GR740
40.3
Pin assignment
The pin assignment in table 583 shows the implementation characteristics of each signal, indicating
how each pin has been configured in terms of electrical levels, voltage, slew rate, drive capability and
internal pull-up or pull-down in the device.
Table 583.Pin assignment
Position
Signal Name
A1
GND
I/O
Level
Volt.
[V]
Slew
Drive
[mA]
Load
[pF]
Pull
Polarity
Note
GND
A2
GND
A3
PROMIO_ADDR[6]
O
PROM
GND
A4
PROMIO_ADDR[2]
O
PROM
A5
PROMIO_WEN
O
PROM
A6
PROMIO_DATA[14]
IO
PROM
A7
PROMIO_DATA[10]
IO
PROM
A8
PROMIO_DATA[6]
IO
PROM
A9
PROMIO_DATA[2]
IO
PROM
A10
GR1553_BUSATXIN
O
MIL-1553
A11
SPI_MOSI
IO
SPI
A12
SYS_CLK
I
Sys/spw CLK
A13
MEM_EXTCLK
I
Sys/spw CLK
A14
SYS_EXTLOCK
I
Sys/spw CLK
A15
JTAG_TCK
I
JTAG
A16
JTAG_TRST
I
JTAG
A17
GPIO[14]
IO
GPIO
A18
GPIO[12]
IO
GPIO
A19
GPIO[8]
IO
GPIO
A20
GPIO[4]
IO
GPIO
A21
GPIO[0]
IO
GPIO
A22
PLL_BYPASS[0]
I
A23
GND
GND
A24
GND
GND
A25
GND
GND
Bootstrap
B1
GND
B2
PROMIO_ADDR[9]
O
PROM
GND
B3
PROMIO_ADDR[8]
O
PROM
B4
PROMIO_ADDR[4]
O
PROM
B5
PROMIO_ADDR[0]
O
PROM
B6
PROMIO_READ
O
PROM
B7
PROMIO_DATA[12]
IO
PROM
B8
PROMIO_DATA[8]
IO
PROM
B9
PROMIO_DATA[4]
IO
PROM
B10
GR1553_CLK
I
MIL-1553
B11
SPI_MISO
IO
SPI
B12
SPW_CLK
I
Sys/spw CLK
B13
SYS_RESETN
I
Sys/spw CLK
B14
JTAG_TMS
I
JTAG
GR740-UM-DS, Jun 2016, Version 1.4
443
www.cobham.com/gaisler
GR740
Table 583.Pin assignment
Level
Volt.
[V]
Slew
Drive
[mA]
Load
[pF]
Position
Signal Name
I/O
Pull
Polarity
Note
B15
JTAG_TDO
O
JTAG
B16
JTAG_TDI
I
JTAG
B17
GPIO[7]
IO
GPIO
B18
GPIO[10]
IO
GPIO
B19
GPIO[6]
IO
GPIO
B20
GPIO[2]
IO
GPIO
B21
PCIMODE_ENABLE
I
Bootstrap
B22
PLL_BYPASS[1]
I
Bootstrap
B23
PLL_IGNLOCK
I
Bootstrap
B24
PLL_LOCKED[4]
O
Bootstrap
B25
GND
C1
PROMIO_ADDR[10]
O
PROM
C2
PROMIO_ADDR[12]
O
PROM
C3
PROMIO_ADDR[7]
O
PROM
C4
PROMIO_ADDR[3]
O
PROM
C5
PROMIO_OEN
O
PROM
C6
PROMIO_DATA[15]
IO
PROM
C7
PROMIO_DATA[9]
IO
PROM
C8
PROMIO_DATA[5]
IO
PROM
C9
PROMIO_DATA[0]
IO
PROM
C10
GR1553_BUSBTXIN
O
MIL-1553
C11
SPI_SCK
IO
SPI
C12
SPI_SEL
I
SPI
C13
SPI_SLVSEL[0]
O
SPI
C14
SPI_SLVSEL[1]
O
SPI
C15
GPIO[13]
IO
GPIO
C16
GPIO[9]
IO
GPIO
C17
GPIO[3]
IO
GPIO
C18
GPIO[1]
IO
GPIO
GND
C19
DSU_ACTIVE
O
Bootstrap
C20
MEM_IFWIDTH
I
Bootstrap
C21
VSS2V5
VSS2V5
C22
VSS2V5
VSS2V5
C23
VSS2V5
VSS2V5
C24
VSS2V5
VSS2V5
C25
VSS2V5
VSS2V5
D1
PROMIO_ADDR[14]
O
PROM
D2
PROMIO_ADDR[16]
O
PROM
D3
PROMIO_ADDR[11]
O
PROM
D4
PROMIO_ADDR[5]
O
PROM
D5
PROMIO_BRDYN
I
PROM
D6
PROMIO_DATA[13]
IO
PROM
D7
PROMIO_DATA[11]
IO
PROM
D8
PROMIO_DATA[3]
IO
PROM
GR740-UM-DS, Jun 2016, Version 1.4
444
www.cobham.com/gaisler
GR740
Table 583.Pin assignment
I/O
Level
Volt.
[V]
Slew
Drive
[mA]
Load
[pF]
Position
Signal Name
Pull
Polarity
Note
D9
VSS3V3
VSS3V3
D10
VDIG3V3
VDIG3V3
D11
VSS3V3
VSS3V3
D12
VDIG3V3
VDIG3V3
D13
VSS3V3
VSS3V3
D14
GPIO[15]
IO
GPIO
D15
GPIO[11]
IO
GPIO
D16
GPIO[5]
IO
GPIO
D17
BREAK
I
Bootstrap
D18
DSU_EN
I
Bootstrap
D19
WDOGN
O
Bootstrap
D20
PLL_LOCKED[2]
O
Bootstrap
D21
VSS2V5
D22
SPW_TXS_P[7]
O
SpaceWire
D23
SPW_TXS_N[7]
O
SpaceWire
D24
SPW_TXD_P[7]
O
SpaceWire
D25
SPW_TXD_N[7]
O
SpaceWire
E1
PROMIO_ADDR[18]
PROM
E2
PROMIO_ADDR[20]
PROM
E3
VDIG3V3
VDIG3V3
E4
PROMIO_ADDR[15]
PROM
VSS2V5
E5
PROMIO_ADDR[13]
PROM
E6
PROMIO_ADDR[1]
PROM
E7
PROMIO_DATA[7]
PROM
E8
PROMIO_DATA[1]
PROM
E9
VDIG3V3
VDIG3V3
E10
VSS3V3
VSS3V3
E11
VDIG3V3
VDIG3V3
E12
AGNDPLL1V2_SYSPLL
DLL
E13
AGNDPLL1V2_MEMPLL
DLL
E14
AGNDPLL1V2_SPWPLL
DLL
E15
PROC_ERRORN
Bootstrap
E16
MEM_CLKSEL
Bootstrap
E17
PLL_BYPASS[2]
Bootstrap
E18
PLL_LOCKED[5]
Bootstrap
E19
PLL_LOCKED[3]
Bootstrap
E20
PLL_LOCKED[0]
Bootstrap
E21
VSS2V5
VSS2V5
E22
SPW_RXS_P[7]
SpaceWire
E23
SPW_RXS_N[7]
SpaceWire
E24
SPW_RXD_P[7]
SpaceWire
E25
SPW_RXD_N[7]
SpaceWire
F1
PROMIO_ADDR[22]
PROM
F2
PROMIO_ADDR[24]
PROM
GR740-UM-DS, Jun 2016, Version 1.4
445
www.cobham.com/gaisler
GR740
Table 583.Pin assignment
I/O
Level
Volt.
[V]
Slew
Drive
[mA]
Load
[pF]
Position
Signal Name
Pull
Polarity
Note
F3
VSS3V3
VSS3V3
F4
PROMIO_ADDR[19]
PROM
F5
PROMIO_ADDR[17]
PROM
F6
VDIG3V3
VDIG3V3
F7
VSS3V3
VSS3V3
F8
VDIG3V3
VDIG3V3
F9
VSS3V3
VSS3V3
F10
VDIG3V3
VDIG3V3
F11
VSS3V3
VSS3V3
F12
AVDDPLL1V2_SYSPLL
DLL
F13
AVDDPLL1V2_MEMPLL
DLL
F14
AVDDPLL1V2_SPWPLL
DLL
F15
VSS3V3
VSS3V3
F16
VDIG3V3
VDIG3V3
F17
VSS3V3
VSS3V3
F18
VDIG3V3
VDIG3V3
F19
VSS3V3
VSS3V3
F20
PLL_LOCKED[1]
Bootstrap
F21
VSS2V5
VSS2V5
F22
SPW_TXS_P[6]
SpaceWire
F23
SPW_TXS_N[6]
SpaceWire
F24
SPW_TXD_P[6]
SpaceWire
F25
SPW_TXD_N[6]
SpaceWire
G1
PROMIO_ADDR[26]
PROM
G2
IO_SN
PROM
G3
VDIG3V3
VDIG3V3
G4
PROMIO_ADDR[23]
PROM
G5
PROMIO_ADDR[21]
PROM
G6
VSS3V3
VSS3V3
G7
VDIG3V3
VDIG3V3
G8
VSS3V3
VSS3V3
G9
VDIG3V3
VDIG3V3
G10
VSS3V3
VSS3V3
G11
VDIG3V3
VDIG3V3
G12
DVDDPLL1V2_SYSPLL
DLL
G13
DVDDPLL1V2_MEMPLL
DLL
G14
DVDDPLL1V2_SPWPLL
DLL
G15
VDIG3V3
VDIG3V3
G16
VSS3V3
VSS3V3
G17
VDIG3V3
VDIG3V3
G18
VSS3V3
VSS3V3
G19
VDIG3V3
VDIG3V3
G20
VSS3V3
VSS3V3
G21
VSS2V5
VSS2V5
GR740-UM-DS, Jun 2016, Version 1.4
446
www.cobham.com/gaisler
GR740
Table 583.Pin assignment
Level
Slew
Drive
[mA]
Load
[pF]
Signal Name
G22
SPW_RXS_P[6]
SpaceWire
G23
SPW_RXS_N[6]
SpaceWire
G24
SPW_RXD_P[6]
SpaceWire
G25
SPW_RXD_N[6]
SpaceWire
H1
PROM_CEN[1]
PROM
H2
PROM_CEN[0]
PROM
H3
VSS3V3
VSS3V3
H4
PROMIO_ADDR[27]
PROM
H5
PROMIO_ADDR[25]