KeyStone C66x Multicore SoC Overview MMI Applications Team October 2011 KeyStone Overview • • • • • KeyStone Architecture CorePac & Memory Subsystem Interfaces and Peripherals Coprocessors and Accelerators Debug Performance improvement Enhanced DSP core C66x ISA 100% upward object code compatible 4x performance improvement for multiply operation 32 16-bit MACs Improved support for complex arithmetic and matrix computation C67x+ C67x 2x registers Native instructions for IEEE 754, SP&DP Advanced VLIW architecture Enhanced floatingpoint add capabilities C674x 100% upward object code compatible with C64x, C64x+, C67x and c67x+ Best of fixed-point and floating-point architecture for better system performance and faster time-to-market. FLOATING-POINT VALUE Preliminary Information under NDA - subject to change C64x+ SPLOOP and 16-bit instructions for smaller code size Flexible level one memory architecture iDMA for rapid data transfers between local memories C64x Advanced fixedpoint instructions Four 16-bit or eight 8-bit MACs Two-level cache FIXED-POINT VALUE KeyStone Device Features Memory Subsystem – Up to 1 MB Local L2 memory per core – Up to 4 MB Multicore Shared Memory (MSM) – Multicore Shared Memory Controller (MSMC) – Boot ROM, DDR3-1600 MHz (64-bit) Application-Specific Coprocessors – 2x TCP3d: Turbo Decoder – TCP3e: Turbo Encoder – 2x FFT (FFT/IFFT and DFT/IDFT) Coprocessor – 4x VCP2 for voice channel decoding Multicore Navigator – Queue Manager – Packet DMA Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 HyperLink TeraNet Multicore Navigator Network Coprocessor – Packet Accelerator – Security Accelerator Embedded Trace Buffer (ETB) & System Trace Buffer (STB) Smart Reflex Enabled 40 nm High-Performance Process E thernet Swit ch SGM II x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Interfaces – High-speed Hyperlink bus – 4x Serial RapidIO Rev 2.1 – 2x 10/100/1000 Ethernet SGMII ports w/ embedded switch – 2x PCIe Generation II – Six-lane Antenna Interface (AIF2) for Wireless Applications o WCDMA, WiMAX, LTE, GSM, TD-SCDMA, TD-LTE o Up to 6.144-Gbps – Additional Serials: I2C, SPI, GPIO, UART Queue Manager Switch C66x CorePac – 1 to 8 C66x Fixed/Floating-Point CorePac DSP Cores at up to 1.25 GHz – Backward-compatible with C64x+ and C67x+ cores – Fixed and Floating Point Operations – RSA instruction set extensions – Chip-rate processing (downlink & uplink) – Reed-Muller decoding (CorePac 1 and 2 only) Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Memory (cache/RAM) EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem • 1 to 8 C66x CorePac DSP Cores operating at up to 1.25 GHz – Fixed and Floating Point Operations – Code compatible with other C64x+ and C67x+ devices • L1 Memory can be partitioned as cache or SRAM – 32KB L1P per core – 32KB L1D per core – Error Detection for L1P – Memory Protection • Dedicated and Shared L2 Memory – 512 KB to 1 MB Local L2 per core – 2 to 4 MB Multicore Shared Memory (MSM) – Multicore Shared Memory Controller (MSMC) – Error detection and correction for all L2 memory – MSM available to all cores and can be either program or data • Boot ROM Memory Expansion Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF CorePac & Memory Subsystem Memory Expansion MSMC • Multicore Shared Memory Controller (MSMC) Debug & Trace • Arbitrates CorePac and SoC master access to shared memory • Provides a direct connection to the DDR3 EMIF • Provides CorePac access to coprocessors and IO peripherals • Memory protection and address extension to 64 GB (36 bits) • Provides multi-stream pre-fetching capability Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 SRAM EDMA • DDR3 External Memory Interface (EMIF) 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor • Support for 1x 16-bit, 1x 32-bit, and 1x 64-bit modes • Supports up to 1600 MHz • Supports power down of unused pins when using 16-bit or 32-bit width • Support for 8 GB memory address • Error detection and correction • EMIF-16 (Media Applications Only) • Three modes: • Synchronized SRAM • NAND flash • NOR flash • Can be used to connect asynchronous memory (e.g., NAND flash) up to 256 MB. Multicore Navigator Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF Memory Expansion Multicore Navigator MSMC Debug & Trace Boot ROM Semaphore Queue Manager and Packet DMA C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 SRAM EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T PCI e I2C x2 Queue Manager Ot hers CorePac & Memory Subsystem Packet DMA Security Accelerator Packet Accelerator Network Coprocessor • Low-overhead processing and routing of packet traffic • Simplified resource management • Effective inter-processor communications • Abstracts physical implementation from application host software • Virtualization to enable dynamic load balancing and provide seamless access to resources on different cores • 8K hardware queues and 16K descriptors • More descriptors can reside in any shared memory • 10 Gbps pre-fetching capability Network Coprocessor Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor • Packet Accelerator (PA) Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 SRAM EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor • Support for single or multiple IP addresses • 1 Gbps wire-speed throughput at 1.5 Mpps • UDP Checksum processing • IPSec ESP and AH tunnels with fast path fully offloaded • L2 support: Ethernet, Ethertype, and VLAN • L3/L4 Support: IPv4/IPv6 and UDP port-based raw Ethernet or IPv4/6 and SCTP port-based routing • Multicast to multiple queues • QoS capability: Per channel/flow to individual queue towards DSP cores and support for TX traffic shaping per device • Security Accelerator (SA) • Support for IPSec, SRTP, 3GPP and WiMAX Air Interface, and SSL/TLS security • Support for simultaneous wire-speed security processing on 1 Gbps Ethernet transmit and receive traffic. • Encryption Modes: ECB, CBC, CTR, F8, A5/3, CCM, GCM, HMAC, CMAC, and GMAC • Encryption Algorithms: AES, DES, 3DES, Kasumi, SNOW 3g, SHA-1, SHA-2, and MD5 External Interfaces Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 SRAM EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces • SGMII allows two 10/100/1000 Ethernet interfaces • Four high-bandwidth Serial RapidIO (SRIO) lanes for interDSP applications • SPI for boot operations • UART for development/testing • Two PCIe at 5 Gbps • I2C for EPROM at 400 Kbps • Application-specific Interfaces: – Antenna Interface 2 (AIF2) for wireless applications – Telecommunications Serial Port (TSIP) x2 for media applications TeraNet Switch Fabric Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric • TeraNet is a process controller – Channel Controller – Transfer Controller • TeraNet provides a configured way – within hardware – to manage traffic queues and ensure priority jobs are getting accomplished while minimizing the involvement of the DSP cores. • TeraNet facilitates high-bandwidth communications between CorePac cores, subsystems, peripherals, and memory. Diagnostic Enhancements Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric Diagnostic Enhancements • Embedded Trace Buffers (ETB) enhance the diagnostic capabilities of the CorePac. • CP Monitor enables diagnostic capabilities on data traffic through the TeraNet switch fabric. • Automatic statistics collection and exporting (non-intrusive) • Monitor individual events for better debugging • Monitor transactions to both memory end point and MMRs (memory mapped Regi) • Configurable monitor filtering capability based on address and transaction type HyperLink Bus Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric Diagnostic Enhancements HyperLink Bus • Provides the capability to expand the C66x to include hardware acceleration or other auxiliary processors • Four lanes with up to 12.5 Gbps per lane Miscellaneous Elements Application-Specific Coprocessors Memory Subsystem MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL L1 P-Cache x3 L1 D-Cache L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 TeraNet HyperLink Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O Ap pli catio nSpeci fic I/ O S PI UAR T x2 PCI e I2C Ot hers Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric Diagnostic Enhancements HyperLink Bus Miscellaneous • Semaphore2 provides atomic access to shared chip-level resources. • Boot ROM • Power Management • Eight 64-bit timers • Three on-chip PLLs: – PLL1 for CorePacs – PLL2 for DDR3 – PLL3 for Packet Acceleration • Three EDMA Device-Specific: Wireless Applications KeyStone Device Architecture for Wireless Applications Memory Subsystem 2MB MSM SRAM 64-Bit DDR3 EMIF Coprocessors MSMC RSA Debug & Trace RSA x2 VCP2 x4 Boot ROM Semaphore C66x™ CorePac Power Management TCP3d x2 TCP3e PLL 32KB L1 P-Cache x3 32KB L1 D-Cache FFTC x2 1024KB L2 Cache EDMA BCP 4 Cores @ 1.0 GHz / 1.2 GHz x3 HyperLink TeraNet Multicore Navigator Switc h E the rne t Switc h SGMI I 2 x4 SRI O x6 AI F2 S PI UAR T x2 PCI e I2C Ot he rs Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric Diagnostic Enhancements HyperLink Bus Miscellaneous Application-Specific Coprocessors Device-Specific (Wireless Apps) • Wireless-specific Coprocessors • FFTC • TCP3 Decoder/Encoder • VCP2 • BCP • Wireless-specific Interfaces: AIF2 x6 • Characteristics • Package Size: 24x24 • Process Node: 40nm • Pin Count: 841 • Core Voltage: 0.9-1.1 V • 2x Rake Search Accelerator (RSA) Device-Specific: Media Applications KeyStone Device Architecture for Media Applications Memory Subsystem 4MB MSM SRAM 64-Bit DDR3 EMIF MSMC Debug & Trace Boot ROM Semaphore C66x™ CorePac Power Management PLL 32KB L1 P-Cache x3 32KB L1 D-Cache 512KB L2 Cache EDMA 1 to 8 Cores @ up to 1.25 GHz x3 HyperLink TeraNet Multicore Navigator Switc h E the rne t Switc h SGMI I x2 x4 SRI O x2 TS IP S PI UAR T x2 PCI e I2C GPI O EMI F 1 6 Queue Manager Packet DMA Security Accelerator Packet Accelerator Network Coprocessor CorePac & Memory Subsystem Memory Expansion Multicore Navigator Network Coprocessor External Interfaces TeraNet Switch Fabric Diagnostic Enhancements HyperLink Bus Miscellaneous Application-Specific Coprocessors Device-Specific (Media Apps) • Media-specific Interfaces • TSIP x2 • EMIF 16 (EMIF-A) • Characteristics • Package Size: 24x24 • Process Node: 40nm • Pin Count: 841 • Core Voltage: 0.9-1.1 V KeyStone Overview • • • • • KeyStone Architecture CorePac & Memory Subsystem Interfaces and Peripherals Coprocessors and Accelerators Debug MSMC Block Diagram CorePac 0 CorePac 1 CorePac 2 CorePac 3 XMC XMC XMC XMC MPAX MPAX MPAX MPAX CorePac Slave Port Teranet 256 256 System Slave Port for shared SRAM ( SMS ) 256 System Slave Port for external memory ( SES ) 256 MSMC System Master Port 256 256 256 256 Memory Protection and Extension Unit ( MPAX ) CorePac Slave Port CorePac Slave Port 256 CorePac Slave Port MSMC Datapath Arbitration Memory Protection and Extension Unit ( MPAX ) 256 EDC MSMC Core events MSMC EMIF Master Port 256 TeraNet EMIF – 64 bit DDR 3 Shared RAM , 2048 KB C66 TeraNet Data Connections HyperLink S CPUCLK/2 256bit TeraNet HyperLink M TPCC TC0 M 16ch QDMA TC1 M EDMA_0 S DDR3 S Shared L2 S S S S L2 0-3 MM SS Core Core S S Core MM M M Network M Coprocessor TAC_FE M RAC_BE0,1 RAC_BE0,1 MM FFTC / PktDMA M FFTC / PktDMA M AIF / PktDMA M M M M SRIO S TCP3e_W/R S M PCIe M M S TCP3d TCP3d S TAC_BE S RAC_FE S RAC_FE S SVCP2 (x4) (x4) SVCP2 SVCP2 VCP2(x4) (x4) M QM_SS DebugSS S CPUCLK/3 128bit TeraNet TC2 M TPCC M TC6 TPCC TC3 64ch TC4TC7 M 64ch TC5TC8 QDMA M QDMA TC9 EDMA_1,2 MSMC M DDR3 • C6616 TeraNet facilitates high XMC SRIO M S QMSS S PCIe Bandwidth communication links between DSP cores, subsystems, peripherals, and memories. • TeraNet supports parallel orthogonal communication links • In order to evaluate the potential communication link throughput, consider the peripheral bit-width and the speed of TeraNet • Please note that while most of the communication links are possible, some of them are not, or are supported by particular Transfer Controllers. Details are provided in the C6616 Data Manual … TeraNet Switch Fabric Legend Bridge S Wireless Apps Only Media Apps Only HyperLink TPCC 16ch QDMA S M3_DDR CPU/2 256b TeraNet SCR M HyperLink M MSMC Subsystem S M3_SL2 EDMA_0 CONFIG XMC x1 to x8 x5 x8 x7 S CorePac M 1x to 8x SRIO M M PA/SA M TC2 M TPCC M TC6 TPCC TC3 64ch TC4 M 64ch TC5TC7 QDMA M TC8 QDMA TC9 S SRIO Bridge 13 M M M M Bridge 14 CPU/3 128b TeraNet SCR S M AIF / DMA M QM_SS M CPU / 3 128b SCR MPU S M M CPU/3 32b TeraNet SCR x2 x2 SCR CPU / 3 x4 SCR CPU / 3 x4 S S S S S S S x4 TPCC TPTC TPCC TPTC S SRIO S CP Tracer (x5) S CP Tracer (x8) S CP Tracer (x7) S TSIP S AIF2 S VCP2 S TCP3D S TCP3E S FFTC TPCC TPTC S MPU S Semaphore MPU S QMSS PA/SA S DebugSS S SEC_CTL QM_SS S PLL_CTL PCIe S Bootcfg S Timer S GPIO S I2C S EMIF16 CP Tracer (x5) M S Boot ROM CP Tracer (x8) M CP Tracer (x7) M Preliminary Information S under SPI NDA - subject to change STM TETB CPU/3 32b Write-only TeraNet SCR DebugSS S STM S TETB CPU/6 32b TeraNet SCR x8 / x16 S INTC S UART … CPU / 3 32b TeraNet SCR CPU / 6 32b TeraNet SCR TCP3E CPU/3 32b TeraNet SCR SCR CPU /2 S VCP2 (x4) S DAP (DebugSS) M TSIP0,1 x2 MPU S TCP3e_W/R FFTC / DMA S TETB Bridge 12 EDMA_1,2 PCIe DDR3 M S TC0 M TC1 M Global Timestamp Multicore Navigator Overview • Multicore Navigator – Purpose - seamless inter-core communications between cores, IP and peripherals. “Fire and forget” – Supports synchronization between cores, move data between cores, move data to and from peripherals – Consists of a Queue Manager and multiple, dedicated Packet DMA engines – Data transfer architecture designed to minimize host interaction while maximizing memory and bus efficiency – Move Descriptors and buffers (or pointers to) between different parts of the Chip • Navigator hardware: – Queue Manager Subsystem (QMSS) – Multiple Packet DMA (PKTDMA) Navigator Architecture Queue Interrupts Link RAM Host (App SW) Buffer Memory Queue Man register I/F PKTDMA register I/F Accumulator command I/F L2 or DDR Descriptor RAMs Accumulation Memory VBUS Hardware Block PKTDMA Rx Coh Unit QMSS Rx Core Tx Core Timer Timer PKTDMA Tx Scheduling Control (internal) APDSP APDSP (Accum) (Monitor) Config RAM Interrupt Distributor Register I/F Rx Channel Ctrl / Fifos Tx Channel Ctrl / Fifos Tx DMA Scheduler Queue Interrupts queue pend Rx Streaming I/F Tx Streaming I/F Output (egress) Input (ingress) PKTDMA Control Tx Scheduling I/F (AIF2 only) Queue Manager queue pend Config RAM Register I/F Link RAM (internal) Queue Manager Subsystem (QMSS) • Features: – 8192 total hardware queues – Up to 20 Memory regions for descriptor storage (LL2, MSMC, DDR) – Up to 2 Linking RAMs for queue linking/management • Up to 16K descriptors can be handled by internal Link RAM. • Second Link RAM can be placed in L2 or DDR. – Up to 512K descriptors supported in total. – Can copy descriptor pointers of transferred data to destination core’s local memory to reduce access latency • Major hardware components: – Queue Manager – PKTDMA (Infrastructure DMA) – 2 PDSPs (Packed Data Structure Processors) for: • Descriptor Accumulation / Queue Monitoring • Load Balancing and Traffic Shaping – Interrupt Distributor (INTD) module Packet DMA Topology FFTC (B) FFTC (A) Queue Manager Subsystem Queue Manager PKTDMA PKTDMA 0 1 2 3 4 5 SRIO PKTDMA ... 8192 AIF PKTDMA Network Coprocessor PKTDMA PKTDMA • Multiple Packet DMA instances in KeyStone devices: — QMSS, PA and SRIO instances for all KeyStone devices. — AIF2 and FFTC (A and B) instances are only in KeyStone devices for wireless applications. • Transfer engine interface between peripherals/accelerators and QMSS • Autonomously determines memory buffers to fill and queues to post based on initial setup and buffer descriptor Queues/Descriptors/Packets Queue PTR PTR PTR Host Packet Descriptor PTR Length ... Data ... PTR Host Buffer Descriptor PTR Length ... Monolithic Descriptor Length ... Data Data XMC – External Memory Controller The XMC is responsible for: 1. 2. 3. 4. Address extension/translation Memory protection for addresses outside C66x Shared memory access path Cache and Pre-fetch support User Control of XMC: 1. 2. MPAX registers – Memory Protection and Extension Registers MAR registers – Memory attributes registers Each core has its own set of MPAX and MAR registers ! The MPAX Registers • Translate between physical and logical address • 16 registers (64 bits each) control (up to) 16 memory segments. • Each register translates logical memory into physical memory for the segment. • Segment definition in the MPAX registers: – – – – Segment size – 5 bits – power of 2, smallest segment size 4K, up to 4GB Logical base address – Physical (replacement address) base Permission – access type allowed in this address range The MAR Registers • MAR = Memory Attributes Registers • 256 registers (32 bits each) control 256 memory segments. – Each segment size is 4M Bytes, from logical address 0x00000000 to address 0xffffffff – The first 16 registers are read-only. They control the internal memory of the core. • Each register controls the cache-ability of the segment (bit 0) and the prefetch-ability (bit 3). All other bits are reserved and set to 0. • All MAR bits are set to zero after reset. KeyStone Overview • • • • • KeyStone Architecture CorePac & Memory Subsystem Interfaces and Peripherals Coprocessors and Accelerators Debug EDMA 3 EDMA Channel Controllers • • • • 1 controller in CPU/2 domain – 2 transfer controllers/queues with 1KB channel buffer – 8 QDMA channels – 16 interrupt channels – 128 PaRAM entries 2 controllers in CPU/3 domain each with – 4 transfer controllers/queues with 1KB or 512B channel buffer – 8 QDMA channels – 64 interrupt channels – 512 PaRAM entries Flexible transfer definition – Linking mechanism allows automatic PaRAM set update – Chaining allows multiple transfers to execute with one event Interrupt generation – Transfer completion – Error conditions 510 511 Interfaces Overview Common Interfaces • One PCI Express (PCIe) Gen II port – – – – Two lanes running at 5G Baud Support for root complex (host) mode and end point mode Single Virtual Channel (VC) and up to eight Traffic Classes (TC) Hot plug • Universal Asynchronous Receiver/Transmitter (UART) – 2.4, 4.8, 9.6, 19.2, 38.4, 56, and 128 K baud rate • Serial Port Interface (SPI) – Operate at up to 66 MHz – Two-chip select – Master mode • Inter IC Control Module (I2C) – One for connecting EPROM (up to 4Mbit) – 400 Kbps throughput – Full 7-bit address field • General Purpose IO (GPIO) module – 16-bit operation – Can be configured as interrupt pin – Interrupt can select either rising edge or falling edge • Serial RapidIO (SRIO) – RapidIO 2.1 compliant – Four lanes @ 5 Gbps • 1.25/2.5/3.125/5 Gbps operation per lane • Configurable as four 1x, two 2x, or one 4x – Direct I/O and message passing (VBUSM slave) – Packet forwarding – Improved support for dual-ring daisy-chain – Reset isolation – Upgrades for inter-operation with packet accelerator • Two SGMII ports with embedded switch – – – – – – – – Supports IEEE1588 timing over Ethernet Supports 1G/100 Mbps full duplex Supports 10/100 Mbps half duplex Inter-working with RapidIO message Integrated with packet accelerator for efficient IPv6 support Supports jumbo packets (9 Kb) Three-port embedded Ethernet switch with packet forwarding Reset isolation with SGMII ports and embedded ETH switch • HyperLink bus – Hardware hooks for analog device or customer ASIC Application-Specific Interfaces For Wireless Applications • Antenna Interface 2 (AIF2) – Multiple-standard support (WCDMA, LTE, WiMAX, GSM/Edge) – Generic packet interface (~12Gbits/sec ingress & egress) – Frame Sync module (adapted for WiMAX, LTE & GSM slots/frames/symbols boundaries) – Reset Isolation For Media Gateway Applications • Telecommunications Serial Port (TSIP) – Two TSIP ports for interfacing TDM applications – Supports 2/4/8 lanes at 32.768/16.384/8.192 Mbps per lane & up to 1024 DS0s Ethernet Switch: Overview • 3 Port Ethernet Switch – Port 0: CPPI port – Port 1: SGMII 0 Port – Port 2: SGMII 1 Port • Ethernet Switch Modules – – – – 2 EMAC modules Address Lookup Engine (ALE) module 2 Statistics modules CPTS (connect Port TS) module • The PA will be discussed later Serial RapidIO (SRIO) • • • SRIO or RapidIO provides a 3-Layered Architecture – Physical defines electrical characteristics, link flow control (CRC) – Transport defines addressing scheme (8b/16b device IDs) – Logical defines packet format and operational protocol 2 Basic Modes of Logical Layer Operation – DirectIO • Transmit Device needs knowledge of memory map of Receiving Device • Includes NREAD, NWRITE_R, NWRITE, SWRITE • Functional units: LSU, MAU, AMU – Message Passing • Transmit Device does not need knowledge of memory map of Receiving Device • Includes Type 11 Messages and Type 9 Packets • Functional units: TXU, RXU Gen 2 Implementation – Supporting up to 5 Gbps PCIe Interface • KeyStone incorporates a single PCIe interface with the following characteristics: – Two SERDES lanes running at 5 GBaud/2.5GBaud – Gen2 compliant – Three different operational modes (default defined by pin inputs at power up; can be overwritten by software): • Root Complex (RC) • End Point (EP) • Legacy End Point – Single Virtual Channel (VC) – Single Traffic Class (TC) – Maximum Payloads • Egress – 128 bytes • Ingress – 256 bytes – Configurable BAR filtering, IO filtering, and configuration filtering HyperLink Bus HyperLink TeraNet Switch Fabric Device #1 TeraNet Switch Fabric • Provides a high-speed interface between device interfaces through the TeraNet switch fabric. • A single 4x bus operating at up to 12.5 Gbps per lane • Connections are point-to-point. Device #2 AIF 2.0 • AIF2 is a peripheral module that supports data transfers between uplink and downlink baseband processors through a high-speed serial interface. AIF2 directly supports the following: • • • • • • WCDMA/FDD LTE FDD LTE TDD WiMax TD-SCDMA GSM/Edge (OBSAI only)Autonomous DMA • PKTDMA or AIF VBUS Master • More efficient data transfer for OFDM standards • FIFO-based buffer provides flexible support for various sampling frequencies. Other Peripherals & System Elements (1/3) • TSIP – Supports 1024 DS0s per TSIP – Supports 2/4/8 lanes at 32.768/16.384/8.192 Mbps per lane • UART Interface – Operates at up to 128,000 baud • I2C Interface – Supports 400Kbps throughput – Supports full 7-bit address field – Supports EEPROM size of 4 Mbit • SPI Interface – Operates at up to 66 MHz – Supports two chip selects – Support master mode • GPIO Interface – 16 GPIO pins – Can be configured as interrupt pins – Interrupt can select either rising edge or falling edge Other Peripherals & System Elements (2/3) • EMIF16 – – – – Used for booting, logging, announcement, etc. Supports NAND flash memory, up to 256MB Supports NOR flash up to 16MB Supports asynchronous SRAM mode, up to 1MB • 64-Bit Timers – Total of 16 64-bit timers • One 64-bit timer per core is dedicated to serve as a watchdog (or may be used as a general purpose timer) • Eight 64-bit timers are shared for general purpose timers – Each 64-bit timer can be configured as two individual 32-bit timers – Timer Input/Output pins • • • • Two timer Input pins Two timer Output pins Timer input pins can be used as GPI Timer output pins can be used as GPO • On-Chip PLLs – Core – Packet & Security CoProcessors – DDR Other Peripherals & System Elements (3/3) • Hardware Semaphores • Power Management • Support to assert NMI input for each core – separate hardware pins for NMI and core selector • Support for local reset for each core – separate hardware pins for local reset and core selector KeyStone Overview • • • • • KeyStone Architecture CorePac & Memory Subsystem Interfaces and Peripherals Coprocessors and Accelerators Debug Network and Security Coprocessor Overview • Packet Accelerator (PA) – Deciphers and adds protocol headers to (and from) packets. – Standard protocols and limited user’s defined protocol routing • Security Accelerator (SA) – Encrypts and decrypts packages Network Coprocessor (Logical) PKTDMA Queue QMSS FIFO Queue Lookup Engine (IPSEC16 entries, 32 IP, 16 Ethernet) Packet Accelerator SRIO message RX Ethernet RX MAC Classify Pass 1 RX PKTDMA Security Accelerator Ingress Path (cp_ace) Egress Path TX PKTDMA Classify Pass 2 Modify Modify Ethernet TX Ethernet MAC TX MAC SRIO message TX CorePac 0 DSP 0 DSP0 0 DSP Session Identification • Hardware lookup identifies the session. • First-pass lookup: – IPv4, IPv6, or Ethernet only – 64 entries (16 Ethernet, 32 up to IPv6, 16 up to IPSec) – IP with ESP or AH as next protocol and SPI • Second-pass lookup: – 8192 entries – UDP, SCTP, etc. or proprietary up to 32-bit identifier within the first 128 bytes of the packet IP/UDP or Raw Ethernet/Flow ID PKTDMA Queue Lookup Engine (IPSEC16 entries, 32 IP, 16 Ethernet) QMSS FIFO Queue 1. No IP Sec detected, IPv6 address matched. Packet Accelerator 2. UDP port or proprietary session ID number matched. SRIO message RX Ethernet RX MAC Classify Pass 1 RX PKTDMA Security Accelerator Ingress Path (cp_ace) Egress Path TX PKTDMA Classify Pass 2 3. UDP checksum verified and result set in descriptor. Modify Modify Ethernet TX Ethernet MAC TX MAC SRIO message TX CorePac 0 DSP 0 DSP0 0 DSP IPSec Flow (IP/UDP in IP/ESP) 1. IP Sec detected and SPI matched against configured security contexts. Lookup Engine (IPSEC16 entries, 32 IP, 16 Ethernet) 2. Authenticate, decrypt, and replay protection. SRIO message RX Ethernet RX MAC Classify Pass 1 PKTDMA Queue QMSS FIFO Queue Packet Accelerator 4. UDP checksum verified and result set in descriptor. Classify Pass 2 RX PKTDMA Security Accelerator Ingress Path (cp_ace) Egress Path TX PKTDMA 3. IPv6 address and UDP port or proprietary session ID number matched. Modify Modify Ethernet TX Ethernet MAC TX MAC SRIO message TX CorePac 0 DSP 0 DSP0 0 DSP IPSec Transmit Flow PKTDMA Queue QMSS FIFO Queue Lookup Engine (IPSEC16 entries, 32 IP, 16 Ethernet) SRIO message RX Ethernet RX MAC Classify Pass 1 2. UDP checksum calculated and result stored in IP header. Classify Pass 2 RX PKTDMA Security Accelerator Ingress Path (cp_ace) Egress Path TX PKTDMA Packet Accelerator 3. Payload is encrypted and authentication tag is computed and stored in trailer. Modify Modify Ethernet TX Ethernet MAC TX MAC SRIO message TX 1. Host SW builds payload and IP/ESP header. CorePac 0 DSP 0 DSP0 0 DSP What is FFTC? • The FFTC is an accelerator that can be used to perform FFT and Inverse FFT (IFFT) on data. • The FFTC has been designed to be compatible with various OFDM-based wireless standards like WiMax and LTE. • The Packet DMA (PKTDMA) is used to move data in and out of the FFTC module. • The FFTC supports four input (Tx) queues that are serviced in a round-robin fashion. • Using the FFTC to perform computations that otherwise would have been done in software frees up CPU cycles for other tasks. FFTC Features • Provides algorithms for both FFT and IFFT • Multiple block sizes: – Maximum 8192 – All LTE DFT (Long Term Evolution Discrete Fourier Transform) sizes • LTE 7.5 kHz frequency shift • 16 bits I/ 16 bits Q input and output – block floating point output • Dynamic and programmable scaling modes – Dynamic scaling mode returns block exponent • Support for left-right FFT shift (switch the left/right halves) • Support for variable FFT shift – For OFDM (Orthogonal Frequency Division Multiplexing) downlink, supports data format with DC subcarrier in the middle of the subcarriers • Support for cyclic prefix – Addition and removal – Any length supported • Three-buffer design allows for back-to-back computations • 128-bit, CPU/3, full-duplex VBUS connection • Input data scaling with shift eliminates the need for front-end digital AGC (Automatic Gain Control) • Output data scaling Turbo CoProcessor 3 Decoder (TCP3D) • TCP3D is a programmable peripheral for decoding of 3GPP (WCDMA, HSUPA, HSUPA+, TD_SCDMA), LTE, and WiMax turbo codes. • Turbo decoding is a part of bit processing. LTE Bit Processing Per Transport Block Soft Bits De-Scrambling Per Code Block Channel De-interleaver LLR combining LLR Data • Systematic • Parity 0 • Parity 1 Hard decision Decoded bits TB CRC De-Rate Matching TCP3D TCP3D Key Features (1/2) • • • • • • • • • • • • Supports 3GPP Rel-7 and older (WCDMA), LTE, and WiMAX turbo decoding Native Code Rate: 1/3 Radix 4 Binary and Duo-Binary MAP Decoders Dual MAP decoders for non-contentious interleavers Split decoder mode: TCP3D works as two independent, single MAP decoders Max Star and Max log-map algorithms Double Buffer input memory for lower latency transfers (except in split mode) 128-bit data bus for reduced latency transfers Input data bit width: 6 bits Programmable hard decision bit ordering within a 128-bit word: 0-127 or 127-0 Soft output information for systematic and parity bits: 8 bits Extrinsic scaling per MAP for up to eight iterations (Both Max and Max Star) TCP3D Key Features (2/2) • • • • • • • • • • • • Block sizes supported: 40 to 8192 Programmable sliding window sizes {16, 32, 48, 64, 96, 128} Max number of iterations: 1 to 15 Min number of iterations: 1 to 15 SNR stopping criterion: 0 to 20 dB threshold LTE CRC stopping criterion LTE, WCDMA and WiMAX Hardware Interleaver Generators Channel Quality Indication Emulation support Low DSP pre-processing load Runs in parallel with CorePac Targets base station environment Turbo CoProcessor 3 Encoder (TCP3E) • TCP3E = Turbo CoProcessor 3 Encoder – No previous versions, but came out at same time as third version of decoder co-processor (TCP3D) – Runs in parallel with DSP • Performs Turbo Encoding for forward error correction of transmitted information (downlink for basestation) – Adds redundant data to transmitted message – Turbo Decoder in handset uses the redundant data to correct errors – Often avoids retransmission due to a noisy channel Turbo Encoder (TCP3E) Downlink Turbo Decoder in Handset TCP3E Features Supported • 3GPP, WiMAX and LTE encoding – 3GPP includes: WCDMA, HSDPA, and TD-SCDMA • • • • • • • • Code rate: 1/3 Can achieve throughput of 250 Mbps in all three modes On-the-fly interleaver table generation Dual-encode engines with input and output memories for increased throughput Programmable input and output format within a 32-bit word Block sizes supported: 40 to 8192 Tail biting for WiMAX CRC encoding for LTE TCP3E Block Diagram Config Registers Input Memory Encode Engine Output Memory Config Registers Ping Input Memory Output Memory Config Registers Input Memory Encode Engine Output Memory Pong • • • • Internally, TCP3E has dual (ping/pong) encode engines, config registers, input and output memories Externally, TCP3E looks like a single set of config regs and input / output buffers Routing to ping/pong is handled internally Alternates between ping and pong from one code block to the next Bit Rate Coprocessor (BCP) The Bit Rate Coprocessor (BCP) is a programmable peripheral for baseband bit processing. Integrated into the Texas Instruments DSP, it supports FDD LTE, TDD LTE, WCDMA, TD-SCDMA, HSPA, HSPA+, WiMAX 802.16-2009 (802.16e), and monitoring/planning for LTE-A. Primary functionalities of the BCP peripheral include the following: • • • • • • • • • • • • • CRC Turbo / convolutional encoding Rate Matching (hard and soft) / rate de-matching LLR combining Modulation (hard and soft) Interleaving / de-interleaving Scrambling / de-scrambling Correlation (final de-spreading for WCDMA RX and PUCCH correlation) Soft slicing (soft demodulation) 128-bit Navigator interface Two 128-bit direct I/O interfaces Runs in parallel with DSP Internal debug logging Viterbi Decoder Coprocessor (VCP2) KeyStone Overview • • • • • KeyStone Architecture CorePac & Memory Subsystem Interfaces and Peripherals Coprocessors and Accelerators Debug Emulation Features (1/2) • Host tooling can halt any or all of the cores on the device. – Each core supports a direct connection to the JTAG interface. – Emulation has full visibility of the CorePac memory map. • Real-Time Emulation allows the user to debug application code while interrupts designated as real-time continue to be serviced. – Normal code execution runs code in the absence of a debug event halting execution with the peripheral operating in a continuous fashion. – Secondary code execution runs code related to the service of a real-time interrupt after a debug event has halted code execution. – No code execution does not run code because a debug event halts code execution, and no real-time interrupt is serviced after code execution is halted. Emulation Features (2/2) • Advanced Event Triggering (AET) allows the user to identify events of interest: – Utilize instruction and data bus comparators, auxiliary event detection, sequencers/state machines, and event counters – Manage breakpoints, trace acquisition, data collection via an interrupt, timing measurement, and generate external triggers – Control a state machine and the counters used to create the intermediate events (loop counts and state machines) – Allow event combining to create simple or complex triggers using modules call trigger builders • AET logic is provided for monitoring program, memory bus, system event activity, remembering event sequences, counting event occurrences, or measuring the interval between events. – Perform range and identity comparisons – Detect exact transactions – Detect touching of a byte or range of bytes by memory references • External event detectors allow monitoring of external triggers or internal states of interest (i.e., cache miss). – Enables four states for the identification of a sequence of triggers – Allow specific system activity to generate breakpoints, an interrupt used for the collection of system data, or the identification of program activity that is observed through trace • Any system event routed to a C66x core can be routed (through software selection) to the AET. Trace Subsystem (Simplified) Trace Stream(s) Optionally Exported ETB0 One Embedded Trace Buffer per CorePac CorePac 0 DRM ETBn-1 S0 DMA Switch Fabric CorePac n Sm Other Masters Other Slaves CP_MONITOR 0 One CP _MONITOR per monitored slave endpoint CP_MONITOR_M TeraNet VBUS command signals exported to CP _ MONITORs Trace Logs generated through dedicated SCR STM ETBn One Embedded Trace for System Trace Trace Features • Trace Pin Support for XDS560T Trace • On-Chip Embedded Trace Buffers – 4 KB (Core) /32 KB (STM) on-chip receiver – One ETB per core for Trace and one for STM – Snapshot and circular buffer mode – Simultaneous write (sink) and read (drain) capability – Can be used in CoreSight ETB mode • C66x CPU Trace: – Trace targets the debug of unstable code: • Provides for the recording of program flow, memory references, cache statistics, and application specific data with a time stamp, performance analysis, and quality assurance. • Bus snoopers to collect and export trace data using hardware dedicated to the trace function. • All or a percentage of the debug port pins can be allocated to trace for any of the cores (or a mix). – Program flow and timing can be traced at the same rate generated by the CPU. – Event trace provides a log of user-selectable system events. Event trace can also be used in conjunction with profiling tools. – Data references must be restricted however as the export mechanism is limited to a number of pins, which is insufficient to sustain tracing of all memory references. • The Advanced Event Triggering facilities provide a means to restrict the trace data exported to data of interest to maintain the non-intrusive aspect of trace. • Error indications are embedded in the debug stream in the event the export logic is unable to keep up with the data rate generated by the collection logic. • The user can optionally select the export of all specified trace data. – In this case, the CPU is stalled to avoid the loss of trace data – The user is notified that trace stalls have occurred although the number of stalls and their location is not recorded. For more information on these features, please refer to Debug/Trace User Guide for your selected KeyStone device. KeyStone CP Tracer Modules Legend Bridge S Wireless Apps Only Media Apps Only VUSR TPCC 16ch QDMA S M3_DDR CPU/2 256b TeraNet SCR CP Tracer M VUSR CPT TC0 M TC1 M EDMA_0 M MSMC_SS S M3_SL2 S for EMIF_DDR3 (36b) CPT CPT CPT CPT 4 CPTs for SRAM (36b) CONFIG XMC X 4/ x 8 DDR3 M x5 x8 S S CP Tracer (x5) S CP Tracer (x8) x7 x4 for Wireless x8 for Media SRIO M M PA/SA M TC2 M TPCC M TC6 TPCC TC3 64ch TC4 M 64ch TC5TC7 QDMA M TC8 QDMA TC9 EDMA_1,2 S RAC_BE0,1 M FFTC / DMA M AIF / DMA M QMSS M PCIe M M M M Bridge 14 CPU/3 128b TeraNet SCR S TCP3e_W/R CPT S CPT CPU / 3 128b SCR MPU CPT TCP3d x2 CPU/3 32b TeraNet SCR S S SCR CPU /2 x2 SCR CPU / 3 x4 SCR CPU / 3 x4 S S S S S S TPCC TPTC TPCC TPTC x4 S CP Tracer (x7) S TSIP S AIF2 S VCP2 S TCP3D S TCP3E S FFTC TPCC TPTC S MPU CPT S Semaphore MPU CPT S QMSS PA/SA S DebugSS S SEC_CTL QMSS S PLL_CTL PCIe S Bootcfg S Timer S GPIO S I2C S EMIF16 CP Tracer (x5) M S Boot ROM CP Tracer (x8) M STM TETB CPU/3 32b TeraNet Write-only SCR CP Tracer (x7) M Preliminary Information S SPI under NDA - subject to change DebugSS S STM S TETB CPU/6 32b TeraNet SCR X8 / x16 S INTC S UART … CPU / 3 32b TeraNet SCR CPU / 6 32b TeraNet SCR CPU/3 32b TeraNet SCR S VCP2 (x4) S M M x2 MPU CPT x2 DAP (DebugSS) M TSIP0,1 SRIO Bridge 13 Monitors transactions from AIF, TCs M S TETB Bridge 12 Monitors transactions from AIF,SRIO, Core, TCs TAC_FE S CorePac M SRIO Global Timestamp CP Tracer Module Features (1/2) • Transaction trace (output to STM) • Ability to 'see' the transactions for each master to selected slave interfaces through tracing of key transaction points: – Arbitration Won (Event B) – Transaction Complete (Event C, E) • Two filtering functions for transaction traces to bring out the specific transactions: – Transaction-qualifier-filtering: read/write – Address-range-based filtering • Statistics counters: – Throughput counts represent the total number of bytes forwarded to the target slave during a specified time duration. • Counter accumulates the byte-count presented at the initiation of a new transfer. • Can be used to calculate the effective throughput in terms of bytes-per-second at a given memory slave interface. • Can be used to track the bandwidth consumed by the system masters. (#bytes/time) – Each CP Tracer provides two independent throughput counters. • Each can be used to track the total number of bytes forwarded from a group of masters. • Each system master can be assigned to either / both /none of the two masters groups for throughput collection. • CP Tracer also provides address range based filtering and transaction qualifier based filtering functions to further narrow the interested transactions. – Accumulated Wait time counter • Provides an indication of how busy the bus is and how many cycles elapsed with at least one bus master waiting for access to the bus – Num Grant counter • Provides an indication of the number of bus grants. The average transaction size can be determined by looking at throughput / num Grant CP Tracer Module Features (2/2) • Sliding Time Window: – Specifies the measurement interval for all the CBA statistic counters implemented in the CP Tracer module. – When the sliding window timer expires, the counter values are loaded into the respective registers and the count starts again. – If enabled, an interrupt is also generated when the sliding time window expires. – The host CPU and/or EDMA can read the statistics counters upon assertion of the interrupt. – If enabled, the counter values can also be exported to STM automatically after the sliding time window is expired. • Cross-trigger generation: can assert EMU0/1 when a qualified event occurs – External trigger to start/stop monitoring. – The EMU0 trigger line is coupled to trace start. The EMU1 trigger line is coupled to trace stop. – Both EMU0 and EMU1 are sourced from any of the CorePac cores. – It can also be controlled from an external source via the EMU0 and EMU1 pins on the device. – The EMU0 trigger enables the EMU01_TraceEnableStatus bit of the Transaction Qualifier register, the EMU1 trigger disables this bit. • STM Trace Export Enables – Status message – Event message – Statistics message For More Information • For more information, refer to the C66x Getting Started page to locate the data manual for your KeyStone device. • View the complete C66x Multicore SOC Online Training for KeyStone Devices, including details on the individual modules. • For questions regarding topics covered in this training, visit the support forums at the TI E2E Community website.