FINAL Am29050 RISC Microprocessor with On-Chip Floating-Point Unit Advanced Micro Devices 64-entry Memory Management Unit (MMU) with region mapping DISTINCTIVE CHARACTERISTICS Full 32-bit, three-bus architecture 1024-byte branch target cache 55 million instructions per second (MIPS) sustained at 40 MHz 4-entry physical address cache On-chip double-precision floating-point arithmetic unit Demultiplexed and pipelined address, instruction, and data buses 80-megaflop peak floating-point execution at 40 MHz 192 general-purpose registers 40-, 33-, 25-, and 20-MHz operating frequencies On-chip byte-alignment support allows optional byte/half-word accesses Three-address instruction architecture CMOS technology/TTL compatible 4-Gbyte virtual address space with demand paging Pin and bus compatibility with Am29000 and Am29005 microprocessors Concurrent instruction and data accesses Binary compatibility with all 29K microprocessors and microcontrollers Burst-mode access support Advanced debugging support SIMPLIFIED BLOCK DIAGRAM Address Am29050 RISC Microprocessor Data 32 32 32 Instruction ROM Instruction Instruction Memory Data Memory Publication# 15039 Rev. B Amendment /0 Issue Date: December 1994. WWW: 5/4/95 AMD ADVANCE INFORMATION DISTINCTIVE CHARACTERISTICS 1 SIMPLIFIED BLOCK DIAGRAM 1 GENERAL DESCRIPTION 3 RELATED AMD PRODUCTS 3 29K FAMILY DEVELOPMENT SUPPORT PRODUCTS 3 THIRD-PARTY DEVELOPMENT SUPPORT PRODUCTS 3 ORDERING INFORMATION 4 LOGIC SYMBOL 5 KEY FEATURES AND BENEFITS 6 PERFORMANCE OVERVIEW 8 CONNECTION DIAGRAM 12 169-LEAD PGA PGA Pin Designations by Pin Number PGA Pin Designations by Pin Name 12 13 14 PIN DESCRIPTIONS 15 ABSOLUTE MAXIMUM RATINGS 19 OPERATING RANGES 19 DC CHARACTERISTICS over COMMERCIAL and INDUSTRIAL Operating Ranges 19 CAPACITANCE 19 SWITCHING CHARACTERISTICS over COMMERCIAL Operating Range 21 SWITCHING CHARACTERISTICS over INDUSTRIAL Operating Range 22 SWITCHING WAVEFORMS 25 CAPACITIVE OUTPUT DELAYS 28 SWITCHING TEST CIRCUIT 28 THERMAL CHARACTERISTICS 29 PHYSICAL DIMENSIONS 30 CGX 169 2 30 Am29050 Microprocessor AMD GENERAL DESCRIPTION The Am29050 RISC microprocessor is a high-performance, general-purpose, 32-bit microprocessor implemented in CMOS technology. It supports a variety of applications by virtue of a flexible architecture and rapid execution of simple instructions that are common to a wide range of tasks. The Am29050 microprocessor meets the demanding requirements of floating-point intensive embedded applications such as X terminals, signal processing, and digital communications. Because it excels at complex math operations required for fast matrix transformations, the Am29050 microprocessor provides graphics and imaging applications with fast 3D performance and gives printers the highest performance page description language (PDL) possible. The Am29050 microprocessor instruction set has been influenced by the results of high-level language, optimizing compiler research. It is appropriate for a variety of languages because it efficiently executes operations that are common to all languages. Consequently, the Am29050 microprocessor is an ideal target for high-level languages such as C, FORTRAN, Pascal, Ada, and COBOL. The processor is available in a 169-lead pin grid array (PGA) package. The PGA has 141 signal pins, 27 power and ground pins, and 1 alignment pin. RELATED AMD PRODUCTS 29K Family Devices Part No. Am29000 Description 32-bit RISC microprocessor Am29005 Low-cost 32-bit RISC microprocessor with no MMU and no branch target cache Am29030 32-bit RISC microprocessor with 8-Kbyte instruction cache Am29035 32-bit RISC microprocessor with 4-Kbyte instruction cache Am29040 32-bit RISC microprocessor with 8-Kbyte instruction cache and 4-Kbyte data cache Am29200 32-bit RISC microcontroller Am29205 Low-cost 32-bit RISC microcontroller with 16-bit bus interface Am29240 32-bit RISC microcontroller with 4-Kbyte instruction cache and 2-Kbyte data cache Am29243 32-bit RISC data microcontroller with instruction and data caches and DRAM parity Am29245 Low-cost 32-bit RISC microcontroller with 4-Kbyte instruction cache 29K FAMILY DEVELOPMENT SUPPORT PRODUCTS Contact your local AMD representative for information on the complete set of development support tools. The following software and hardware development products are available on several hosts: Optimizing compilers for common high-level languages Assembler and utility packages Source- and assembly-level software debuggers Target-resident development monitors Simulators Demonstration and evaluation systems THIRD-PARTY DEVELOPMENT SUPPORT PRODUCTS The Fusion29K Program of Partnerships for Application Solutions provides the user with a vast array of products designed to meet critical time-to-market needs. Products and solutions available from the AMD Fusion29K Partners include Modeling/simulation tools Software development tools Real-time operating systems (RTOS) Application-specific hardware and software Silicon products Board-level products Emulators Manufacturing and prototyping support Hardware and software debuggers Custom support and training Am29050 Microprocessor 3 AMD ORDERING INFORMATION Standard Products AMD standard products are available in several packages and operating ranges. Valid order numbers are formed by a combination of the elements below. AM29050 –25 C G SHIPPING OPTION Blank = Standard Processing B = Burn-in TEMPERATURE RANGE C = Commercial (TC = 0°C to +85°C) I = Industrial (TC = –40°C to +125°C *) * TC = TJ = 125°C max PACKAGE TYPE G = 169-Lead Pin Grid Array (CG169) SPEED OPTION –40 = 40 MHz –33 = 33 MHz –25 = 25 MHz –20 = 20 MHz DEVICE NUMBER/DESCRIPTION Am29050 RISC Microprocessor with On-Chip Floating-Point Unit Valid Combinations 4 Comments AM29050–40 AM29050–33 AM29050–25 AM29050–20 GC With heat sink With heat sink Without heat sink Without heat sink AM29050–33 AM29050–25 AM29050–20 GI With heat sink Without heat sink Without heat sink Valid Combinations Valid Combinations list configurations planned to be supported in volume for this device. Consult the local AMD sales office to confirm availability of specific valid combinations and to check on newly released combinations. Am29050 Microprocessor AMD LOGIC SYMBOL BREQ BGRT PEN BINV IRDY R/W IERR SUP/US IBACK LOCK DRDY MPGM1–MPGM0 DERR IREQ DBACK PDA DBREQ CDA WARN 4 2 DREQT1–DREQT0 CNTL1–CNTL0 TEST OPT2–OPT0 3 STAT2–STAT0 3 INCLK 32 2 MSERR INTR2–INTR0 RESET 2 2 IREQT TRAP1–TRAP0 PIA IBREQ I31–I0 PWRCLK A31–A0 SYSCLK 32 D31–D0 Am29050 Microprocessor 5 AMD KEY FEATURES AND BENEFITS The Am29050 microprocessor extends the 29K Family of processors with a high-performance, pipelined, onchip floating-point unit. The floating-point unit performs IEEE-compatible, single-precision and double-precision arithmetic at a peak rate of 80 million floating-point operations per second (MFLOPS) at 40 MHz. The Am29050 microprocessor also has features to improve the performance of loads and branches, allowing sustained integer performance of 55 million instructions per second (MIPS) at 40 MHz. The Am29050 microprocessor provides a powerful upgrade to the Am29000 microprocessor. It can be used in existing Am29000 processor applications without hardware or software modifications, bringing a dramatic increase in performance to floating-point-intensive applications, particularly graphics and laser-printer applications. Added features include a pipelined floating-point arithmetic unit, region mapping for virtual-to-physical address translation, a monitor mode for debugging supervisor code, and instruction breakpoints for enhanced debugging. Specific performance enhancements to the Am29000 microprocessor include a larger branch target cache, a physical address cache, an early address generator, and instruction forwarding logic. On-Chip Floating-Point Arithmetic Unit An on-chip floating-point unit performs single- and double-precision floating-point operations in accordance with the IEEE Standard for Binary Floating-Point Arithmetic (ANSI/IEEE Standard 754-1985). This unit also performs integer multiplications. An independent divide/square root unit interactively computes floatingpoint divisions and square roots in parallel with other operations. A 64-bit register file port is dedicated to the writing of floating-point results. These results can be written without interfering with integer operations. Multiple concurrent operations can be supported, the principal limitation being resource contention. Most of the time, floating-point operations are performed in parallel with integer operations and other floating-point operations. Memory Management The Am29050 microprocessor incorporates a Memory Management Unit (MMU) that accepts a 32-bit virtual byte address and translates it to a 32-bit physical byte address in a single cycle. The MMU is not dedicated to any particular address-translation architecture. Address translation is performed either by the Translation Look-Aside Buffer (TLB) or by one of the two region mapping units. The TLB maps virtual regions of fixed size into physical regions of the same size. The TLB is an associative 6 table that contains the most recently used address translations for the processor. TLB entries are modified directly by processor instructions. A TLB entry consists of 64 bits and appears as two word-length TLB registers that can be inspected and modified by instructions. In addition to page-by-page address translation, the Am29050 microprocessor supports translation for variable-sized regions. The region mapping units map virtual regions of variable size ranging from 64 Kbyte to 2 Gbyte into regions of physical memory. Each region mapping unit consists of two protected special-purpose registers. Any virtual address not mapped by the region mapping units is translated by the TLB. Branch Target Cache The branch target cache on the Am29050 microprocessor allows fast access to instructions fetched nonsequentially. This keeps the instruction pipeline full until the processor can establish a new instruction-prefetch stream from the external instruction memory. A branch instruction can execute in a single cycle if the branch target is in the branch target cache. The branch target cache is a 1-Kbyte storage array that contains blocks of instructions from recently taken branches. To improve the proportion of successful searches, the branch target cache is organized as a two-way set-associative memory. The branch target cache can be configured under software control to cache either two or four instructions for each branch. Each of the two sets in the branch target cache contains 128 instructions, and the 128 instructions are further divided either into 32 blocks of four instructions each or into 64 blocks of two instructions each. System Interface The Am29050 microprocessor channel consists of the following 32-bit buses and related controls: An instruction bus that transfers instructions into the processor. A data bus that transfers data to and from the processor. An address bus that provides addresses for both instruction and data accesses. The address bus also is used to transfer data to a coprocessor. The channel performs accesses and data transfers to all external devices and memories, including instruction and data memories, instruction caches, data caches, input/output devices, bus converters, and coprocessors. The channel defines three different access protocols: simple, pipelined, and burst-mode. For simple accesses, the Am29050 microprocessor holds the address valid throughout the entire access. This is appropriate for high-speed devices that can complete Am29050 Microprocessor AMD an access in one cycle, and for low-cost devices that are accessed infrequently (such as read-only memories containing initialization routines). Pipelined and burstmode accesses provide high performance with other types of devices and memories. circuits. It is possible to connect a second processor in parallel with the first, where the second processor has its outputs disabled due to the Test mode. The second processor detects open-circuit signals, as well as provides a check of the outputs of the first processor. The Am29050 microprocessor determines whether an access is simple, pipelined, or burst-mode on a transferby-transfer (i.e., generally device-by-device) basis. However, an access that begins as a simple access may be converted to a pipelined or burst-mode access at any time during the transfer. This relaxes the timing constraints on the channel-protocol implementation, since addressed devices do not have to respond immediately to a pipelined or burst-mode request. Debugging and Testing Except for the shared address bus, the channel maintains a strict division between instruction and data accesses. In the most common situation, the system supplies the processor with instructions using burstmode accesses, with instruction addresses transmitted to the system only when a branch occurs. Data accesses can occur simultaneously without interfering with instruction transfer. The Am29050 microprocessor contains arbitration logic to support other masters on the channel. A single external master can arbitrate directly for the channel, while multiple masters can arbitrate using a daisy chain or other method that requires no additional arbitration logic. However, to increase arbitration performance in a multiple-master configuration, an external channel arbiter should be used. This arbiter works in conjunction with the processor’s arbitration logic. Clocks The Am29050 microprocessor generates and distributes a system clock at its operating frequency. This clock is specially designed to reduce skews between the system clock and the processor’s internal clocks. The internal clock-generation circuitry requires a single-phase oscillator signal at twice the processor operating frequency. For systems in which processor-generated clocks are not appropriate, the Am29050 microprocessor also can accept a clock from an external clock generator. The processor decides between these two clocking arrangements based on whether the power supply to the clock-output driver is tied to +5 V or to Ground. Master/Slave Operation Each Am29050 microprocessor output has associated logic that compares the signal on the output with the signal that the processor is providing internally to the output driver. The processor signals situations where the output of any enabled driver does not agree with its input. For a single processor, the output comparison detects short circuits in output signals, but does not detect open The Am29050 microprocessor provides debugging and testing features for both software and hardware. Software debugging on the Am29050 microprocessor is supported by the trace facility, hardware breakpoints, and the monitor mode. The trace facility guarantees exactly one trap after the execution of any instruction in a program being tested. The trace trap allows a debug routine to follow the execution of instructions, and to determine the state of the processor and system at the end of each instruction. The Am29050 microprocessor provides two hardware instruction breakpoints that can suspend execution of the current program on a specified instruction access. Suspension either forces a trace trap or forces a halt if the system is under emulator control. Monitor mode allows debugging of operating-system routines and interrupt and trap handlers. Monitor mode can also be used by an external hardware debugger. The Am29050 microprocessor also supports the direct attachment of a hardware-development system such as an in-circuit emulator. This attachment is made directly to the processor in the system under development, without removing the processor from the system. A test/development interface makes it possible for the hardware-development system to inspect and modify the internal state (e.g., general-purpose register contents, TLB entries, etc.) of the Am29050 microprocessor. In addition, the Am29050 microprocessor can be used to access other system devices and memories on behalf of the hardware-development system. The test/development interface is composed of a group of pins that indicate the state of the processor and control the operation of the processor. The Halt, Step, Reset, and Load Test Instruction modes allow the hardware-development system to control the operation of the Am29050 microprocessor. The hardware-development system can supply the processor with instructions on the instruction bus using the Load Test Instruction mode. Internal processor state can be inspected and modified via the data bus. Coprocessor Attachment A coprocessor for the Am29050 microprocessor attaches directly to the processor channel. However, this attachment has features that are different than those of other channel devices. The coprocessor interface on the Am29050 microprocessor supports a high operand- Am29050 Microprocessor 7 AMD transfer rate and the overlap of coprocessor operations with other processor operations, including other external accesses. The coprocessor is assigned a special address space on the channel. This permits the transfer of operands and other information on the address bus without interfering with normal addressing functions. Since both the address bus and data bus are used for data transfer, the Am29050 microprocessor can transfer 64 bits of information to the coprocessor in one cycle. Pin, Bus, and Binary Compatibility Compatibility within a processor family is critical for achieving a rational, easy upgrade path. The Am29050 microprocessor provides compatibility on several levels. The processor is pin, bus, and binary compatible with the Am29000 and Am29005 processors. Pin and bus compatibility ensures a convenient upgrade path for embedded applications. In addition, the processor is binary compatible with the other members of the 29K family (the Am29030, Am29035, and Am29040 microprocessors, and the Am29200, Am29205, Am29240, Am29243, and Am29245 microcontrollers). Complete Development and Support Environment A complete development and support environment is vital for reducing a product’s time-to-market. Advanced Micro Devices has created a standard development environment for the 29K Family of processors. In addition, the Fusion29K program’s third-party support organization provides the most comprehensive customer/partner program in the embedded processor market. Advanced Micro Devices offers a complete set of hardware and software tools for design, integration, debugging, and benchmarking. These tools, which are available now for the 29K Family, include the following: High C 29K optimizing C compiler with assembler, linker, ANSI library functions, and 29K Family architectural simulator XRAY29K source-level debugger MiniMON29K debug monitor A complete family of demonstration and development systems In addition, Advanced Micro Devices has developed a standard host interface (HIF) specification for operating system services, the Universal Debug Interface (UDI) for seamless connection of debuggers to ICEs and target hardware, and extensions for the UNIX common object file format (COFF). This support is augmented by a staff of factory support and field application engineers, an engineering hotline, and an on-line technical bulletin board. 8 PERFORMANCE OVERVIEW The Am29050 microprocessor provides a significant margin of performance over other processors in its class, since the majority of processor features were defined with the maximum achievable performance in mind. This section describes the features of the Am29050 microprocessor from the point of view of system performance. Instruction Set Overview All 29K Family members employ a three-address instruction set architecture. The compiler or assemblylanguage programmer is given complete freedom to allocate register usage. There are 192 general-purpose registers, allowing the retention of intermediate calculations and avoiding needless data destruction. Instruction operands can be contained in any of the general-purpose registers, and the results can be stored into any of the general-purpose registers. The Am29050 microprocessor also provides four double-precision floating-point accumulator registers for use with the floating-point multiply-accumulate and multiply-sum operations. Instructions are provided for writing and reading the accumulator registers directly. The Am29050 microprocessor instruction set contains 125 instructions that are divided into nine classes. These classes are integer arithmetic, compare, logical, shift, data movement, constant, floating point, branch, and miscellaneous. All instructions are capable of executing in one processor cycle, with the exception of interrupt returns, loads, stores, and certain floating-point instructions. Instruction Execution The Am29050 microprocessor uses an arithmetic/logic unit, a field shift unit, and a prioritizer to execute most instructions. Each of these is organized to operate on 32-bit operands, and provide a 32-bit result. Floating-point operations are performed in an on-chip floating-point unit. The floating-point unit performs 32-bit, single-precision and 64-bit, double-precision computations. The performance degradation of load and store operations is minimized in the Am29050 microprocessor by overlapping them with instruction execution, by taking advantage of pipelining, and by organizing the flow of external data onto the processor so that the impact of external accesses is minimized. Early Loads The early load feature of the Am29050 microprocessor speeds up the execution of load operations by making the physical address of the load instruction available at the end of the decode cycle of the load instruction. At the Am29050 Microprocessor AMD beginning of the next cycle, when the load enters the execute stage, the physical address appears on the channel. In effect, early loads reduce the memory access time by one cycle. Early loads can occur in two different ways. Either the physical address of the load is available in the four-entry physical address cache, or, when an address computation immediately precedes the load instruction, the computed physical address can be forwarded directly to the channel. Pipelining The Am29050 microprocessor utilizes a four-stage pipeline for integer operations, allowing it to execute one integer instruction every clock cycle. The processor can complete an instruction on every cycle, even though four cycles are required from the beginning of an instruction to its completion. Floating-point operations are pipelined to a depth determined by the operation latency and are overlapped with integer operations. A floating-point operation and an integer operation can complete at the same time without stalling the pipeline. Instruction operations are overlapped with operand fetch and result write-back to the register file. Pipeline forwarding logic detects pipeline dependencies and routes data as required, avoiding delays which might arise from these dependencies. Pipeline interlocks are implemented by processor hardware, including those required for floating-point operations. Except for a few special cases, it is not necessary to rearrange programs to avoid pipeline dependencies, although this is sometimes desirable for performance. Register File The on-chip register file containing 192 general-purpose registers allows most instruction operands to be fetched without the delay of an external access. The register file incorporates several features that aid the retention of data required by an executing program. Because of the number of general-purpose registers, the frequency of external references for the Am29050 microprocessor is significantly lower than the frequency of references in processors having only 16 or 32 registers. Four-port access to the register file allows two 64-bit source-operands to be fetched in one cycle while two previously computed results are written; one write port is for integer operations, and the other port is for floatingpoint operations. Four 64-bit internal buses prevent contention in the routing of operands. All operand fetches and result write-backs for instruction execution can be performed in a single cycle. procedure calls and returns can be executed 5 to 10 times faster (on a cycle-by-cycle basis) than in processors that require the implementation of a run-time stack in external memory (with the attendant loading and storing of registers on procedure call and return). Protection The Am29050 microprocessor offers three mutually exclusive modes of execution that restrict or permit accesses to certain processor registers and external storage locations. Supervisor and user modes are for normal program execution, and monitor mode is used for debugging. All system-protection features of the Am29050 microprocessor are based on the difference between these modes. Memory access protection is provided by the MMU. Six protection bits determine whether or not an access is permitted to the page associated with the entry. For the same virtual page, the access authority of programs executing in supervisor mode can be different from the authority of programs executing in user mode. The register file can be configured to restrict accesses to supervisor-mode programs on a bank-by-bank basis using a register bank protect register. Data Formats The Am29050 microprocessor defines a word as 32 bits of data, a half-word as 16 bits, and a byte as 8 bits. The hardware provides direct support for single- and doubleprecision floating-point, word-integer (signed and unsigned), word-logical, word-boolean, half-word integer (signed and unsigned), and byte data (signed and unsigned). Word-boolean data is based on the value contained in the most significant bit of the word. The values TRUE and FALSE are represented by the most significant bit values 1 and 0, respectively. Other data formats, such as character strings, are supported by instruction sequences. Separate Address, Instruction, and Data Buses The Am29050 microprocessor incorporates two 32-bit buses for instruction and data transfers, and a third address bus that is shared between instruction and data accesses. This bus structure allows simultaneous instruction and data transfers, even though the address bus is shared. The channel achieves the performance of four separate 32-bit buses at a much reduced pin count. The address bus is pipelined, so that it can be released before an instruction or data transfer is completed. This allows a subsequent access to begin before the first has completed and allows the processor to have two accesses in progress simultaneously. The registers allow efficient procedure linkage by caching a portion of a compiler’s run-time stack. On the average, Am29050 Microprocessor 9 AMD Support of Burst Devices and Memories Burst-mode accesses provide high transfer rates for instructions and data at sequential addresses. For such accesses, the address of the first instruction or datum is sent, and subsequent requests for instructions or data at sequential addresses do not require additional address transfers. These instructions or data are transferred until either party involved in the transfer terminates the access. Burst-mode accesses can occur at the rate of one access per cycle after the first address has been processed. At 40 MHz, the maximum achievable transfer bandwidth for either instructions or data is 160 Mbyte/s. Burst-mode accesses may occur to input/output devices, if the system design permits. Interface to Fast Devices and Memories The processor can be interfaced to devices and memories that complete accesses within one cycle. The channel protocol takes maximum advantage of such devices and memories by allowing data to be returned to the processor during the cycle in which the address is transmitted. This allows a full range of memory-speed trade-offs to be made within a particular system. Branch Target Cache Memory In general, the Am29050 microprocessor meets its instruction bandwidth requirements via instruction prefetching. However, instruction prefetching is ineffective when a branch occurs. The Am29050 microprocessor therefore incorporates a 64- or 128-entry (configurable at run time) branch target cache array to supply instructions for a branch—if this branch has been taken previously—while a new prefetch stream is established. If branch-target instructions are in the branch target cache, branches execute in a single cycle. This has a very positive effect on processor performance, due to the amount of time the processor could otherwise be idle waiting for the new instruction stream. The branch target cache in the Am29050 microprocessor eliminates the branch latency for 80% of all successful branches on the average. Branching Branch conditions in the Am29050 microprocessor can be based on Boolean data contained in general-purpose registers, as well as on arithmetic condition codes. has a dramatic effect on performance, since successful branches typically represent 15% to 25% of a processor’s instruction mix. The techniques used to achieve single-cycle branching also minimize the execution time of branches in those cases where the target is not in the branch target cache. To keep the pipeline operating at the maximum rate, the instruction following the branch, referred to as the delay instruction, is executed regardless of the outcome of the branch. An optimizing compiler can define a useful instruction for the delay instruction in approximately 90% of branch instructions, thereby increasing the performance of branches. Memory Management A 64-entry Translation Look-Aside Buffer (TLB) and two Region Mapping registers on the Am29050 microprocessor perform virtual-to-physical address translation, avoiding the cycle that would be required to transfer the virtual address to an external TLB. A number of enhancements improve the performance of address translation: Pipelining—The operation of the TLB is pipelined with other processor operations. Early Address Translation—Address translations for load, store, and branch instructions occur during the cycle in which these instructions are executed. This allows the physical address to be transferred externally in the next cycle. Region Mapping—The region mapping registers permit efficient mapping of large, contiguous regions of memory. This is useful for code libraries and large data structures; these can appear in a virtual address space without paging overhead. Task Identifiers—Task identifiers allow TLB entries to be matched to different processes, so that TLB invalidation is not required during task switches. Least-Recently Used Hardware—This hardware allows immediate selection of a TLB set to be replaced. Software Reload—Software reload allows the operating system to use a page-mapping scheme that is best matched to its environment. Paged-segmented, one-level-page mapping, two-level-page mapping, or any other user-defined page-mapping scheme can be supported. Because Am29050 microprocessor instructions execute at an average rate of nearly one instruction per cycle, software reload has a performance approaching that of hardware TLB reload. The Am29050 microprocessor executes branches in a single cycle, for those cases where the target of the branch is in the branch target cache. The single-cycle branch is unusual for a pipelined processor, and is due to processor hardware which allows much of the branch instruction operation to be performed early in the execution of the branch. Single-cycle branching 10 Am29050 Microprocessor AMD Interrupts and Traps When the Am29050 microprocessor takes an interrupt or trap, it does not automatically save its current state information in memory. This feature greatly improves the performance of temporary interruptions such as TLB reload or other simple operating-system calls or interrupts that require no saving of state information. In cases where the processor state must be saved, the saving and restoring of state information is under the control of software. The methods and data structures used to handle interrupts—and the amount of state saved—can be tailored to the needs of a particular system. Interrupts and traps are dispatched through a 256-entry vector area that directs the processor to a routine to handle a given interrupt or trap. The vector area can be relocated in memory by the modification of a processor register. There can be multiple vector areas in the system, though only one is active at any given time. The vector area is either a table of pointers to the interrupt and trap handlers, or a segment of instruction memory (possibly read-only memory) containing the handlers themselves. The choice between the two possible vector area definitions is determined by the cost/ performance trade-offs made for a particular system. Am29050 Microprocessor 11 AMD CONNECTION DIAGRAM 169-Lead PGA Bottom View A B C D E F G H J K L M N P R T U 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Note: Pinout observed from pin side of package. 12 Am29050 Microprocessor AMD PGA PIN DESIGNATIONS (Sorted by Pin Number) Pin No. Pin Name Pin No. Pin Name Pin No. Pin Name Pin No. Pin Name A1 GND C10 GND J16 A16 R12 STAT2 A2 I1 C11 GND J17 A14 R13 GND A3 I0 C12 D22 K1 I26 R14 OPT0 A4 D2 C13 D26 K2 I25 R15 A2 A5 D4 C14 VCC K3 GND R16 A6 A6 D6 C15 D30 K15 VCC R17 A7 A7 D9 C16 D31 K16 A12 T1 INCLK A8 D11 C17 A29 K17 A13 T2 BREQ A9 D12 D1 I11 L1 I27 T3 DERR A10 D14 D2 I10 L2 I28 T4 IRDY A11 D16 D3 I7 L3 VCC T5 WARN A12 D18 D4♦ PIN169 L15 VCC T6 INTR2 A13 D20 D15 A31 L16 A10 T7 INTR0 A14 D21 D16 A28 L17 A11 T8 BINV A15 D25 D17 A26 M1 I29 T9 BGRT A16 D27 E1 I13 M2 I30 T10 DREQ A17 GND E2 I12 M3 GND T11 LOCK B1 I6 E3 VCC M15 GND T12 MSERR B2 I5 E15 GND M16 A0 T13 STAT0 B3 I3 E16 A27 M17 A1 T14 SUP/ US B4 D0 E17 A23 N1 I31 T15 OPT1 B5 D1 F1 I16 N2 TEST T16 A3 B6 D5 F2 I15 N3 SYSCLK T17 A4 B7 D8 F3 I14 N15 GND U1 GND B8 D10 F15 A25 N16 MPGM1 U2 PEN B9 D13 F16 A24 N17 MPGM0 U3 IERR B10 D15 F17 A21 P1 CNTL1 U4 IBACK B11 D17 G1 I19 P2 CNTL0 U5 INTR3 B12 D19 G2 I18 P3 PWRCLK U6 INTR1 B13 D23 G3 I17 P15 A5 U7 TRAP0 B14 D24 G15 A22 P16 A8 U8 IBREQ B15 D28 G16 A20 P17 A9 U9 IREQ B16 D29 G17 A19 R1 RESET U10 PIA B17 A30 H1 I20 R2 CDA U11 R/W C1 I9 H2 I22 R3 DRDY U12 DREQT1 C2 I8 H3 I21 R4 DBACK U13 DREQT0 C3 I4 H15 GND R5 GND U14 STAT1 C4 I2 H16 A18 R6 VCC U15 IREQT C5 GND H17 A17 R7 TRAP1 U16 OPT2 C6 D3 J1 I23 R8 GND U17 GND C7 D7 J2 I24 R9 DBREQ C8 VCC J3 GND R10 PDA C9 VCC J15 A15 R11 VCC ♦Note: Pin Number D4 is the alignment pin and is electrically connected to the package lid. Am29050 Microprocessor 13 AMD PGA PIN DESIGNATIONS (Sorted by Pin Name) Pin Name Pin No. Pin Name Pin No. Pin Name Pin No. Pin Name Pin No. A0 M16 D5 B6 GND K3 INCLK T1 A1 M17 D6 A6 GND N15 INTR0 T7 A2 R15 D7 C7 GND R5 INTR1 U6 A3 T16 D8 B7 GND U1 INTR2 T6 A4 T17 D9 A7 GND R13 INTR3 U5 A5 P15 D10 B8 GND R8 IRDY T4 A6 R16 D11 A8 GND M3 IREQ U9 A7 R17 D12 A9 GND U17 IREQT U15 A8 P16 D13 B9 I0 A3 LOCK T11 A9 P17 D14 A10 I1 A2 MPGM0 N17 A10 L16 D15 B10 I2 C4 MPGM1 N16 A11 L17 D16 A11 I3 B3 MSERR T12 A12 K16 D17 B11 I4 C3 OPT0 R14 A13 K17 D18 A12 I5 B2 OPT1 T15 A14 J17 D19 B12 I6 B1 OPT2 U16 A15 J15 D20 A13 I7 D3 PDA R10 A16 J16 D21 A14 I8 C2 PEN U2 A17 H17 D22 C12 I9 C1 PIA U10 A18 H16 D23 B13 I10 D2 PIN169 D4♦ A19 G17 D24 B14 I11 D1 PWRCLK P3 A20 G16 D25 A15 I12 E2 R/W U11 A21 F17 D26 C13 I13 E1 RESET R1 A22 G15 D27 A16 I14 F3 STAT0 T13 A23 E17 D28 B15 I15 F2 STAT1 U14 A24 F16 D29 B16 I16 F1 STAT2 R12 A25 F15 D30 C15 I17 G3 SUP/US T14 A26 D17 D31 C16 I18 G2 SYSCLK N3 A27 E16 DBACK R4 I19 G1 TEST N2 A28 D16 DBREQ R9 I20 H1 TRAP0 U7 A29 C17 DERR T3 I21 H3 TRAP1 R7 A30 B17 DRDY R3 I22 H2 VCC C14 A31 D15 DREQ T10 I23 J1 VCC L15 BGRT T9 DREQT0 U13 I24 J2 VCC C8 BINV T8 DREQT1 U12 I25 K2 VCC C9 BREQ T2 GND E15 I26 K1 VCC E3 CDA R2 GND H15 I27 L1 VCC K15 CNTL0 P2 GND M15 I28 L2 VCC L3 CNTL1 P1 GND C10 I29 M1 VCC R6 D0 B4 GND A1 I30 M2 VCC R11 D1 B5 GND A17 I31 N1 WARN T5 D2 A4 GND C5 IBACK U4 D3 C6 GND C11 IBREQ U8 D4 A5 GND J3 IERR U3 ♦Note: 14 Pin Number D4 is the alignment pin and is electrically connected to the package lid. Am29050 Microprocessor AMD PIN DESCRIPTIONS Although certain outputs are described as being threestate or bidirectional outputs, all outputs (except MSERR) may be placed in a high-impedance state by the Test mode. The three-state and bidirectional terminology in this section is for those outputs (except SYSCLK) that are disabled when the processor grants the channel to another master. D31–D0 A31–A0 This input is active whenever a burst-mode data access has been established. It may be active even though no data are currently being accessed. Address Bus (three-state outputs, synchronous) The address bus transfers the byte address for all accesses except burst-mode accesses. For burst-mode accesses, it transfers the address for the first access in the sequence. Data Bus (bidirectional, synchronous) The data bus transfers data to and from the processor for load and store operations. DBACK Data Burst Acknowledge (input, synchronous) DBREQ Data Burst Request (three-state output, synchronous) This output signals to an external master that the processor is relinquishing control of the channel in response to BREQ. This signal is used to establish a burst-mode data access and to request data transfers during a burst-mode data access. DBREQ may be active even though the address bus is being used for an instruction access. This signal becomes valid late in the cycle, with respect to DREQ. BINV DERR BGRT Bus Grant (output, synchronous) Bus Invalid (output, synchronous) Data Error (input, synchronous) This output indicates that the address bus and related controls are invalid. It defines an idle cycle for the channel. BREQ Bus Request (input, synchronous) This input allows other masters to arbitrate for control of the processor channel. This input indicates that an error occurred during the current data access. For a load, the processor ignores the content of the data bus. For a store, the access is terminated. In either case, a Data Access Exception trap occurs. The processor ignores this signal if there is no pending data access. DRDY CDA Coprocessor Data Accept (input, synchronous) This signal allows the coprocessor to indicate the acceptance of operands or operation codes. For transfers to the coprocessor, the processor does not expect a DRDY response; an active level on CDA performs the function normally performed by DRDY. CDA may be active whenever the coprocessor is able to accept transfers. Data Ready (input, synchronous) For loads, this input indicates that valid data is on the data bus. For stores, it indicates that the access is complete, and that data need no longer be driven on the data bus. The processor ignores this signal if there is no pending data access. DREQ Data Request (three-state output, synchronous) CNTL1–CNTL0 This signal requests a data access. When it is active, the address for the access appears on the address bus. CPU Control (inputs, asynchronous) These inputs control the processor mode: DREQT1–DREQT0 Data Request Type (three-state outputs, synchronous) CNTL1 CNTL0 Mode 0 0 Load Test Instruction 0 1 Step 1 0 Halt 1 1 Normal These signals specify the address space of a data access, as follows (the value “x” is a “don’t care”): DREQT1 DREQT0 0 0 Instruction/data memory access 0 1 Input/output access 1 x Coprocessor transfer Am29050 Microprocessor Mode 15 AMD An interrupt/trap vector request is indicated as a data memory read. If required, the system can identify the vector fetch by the STAT2–STAT0 outputs. DREQT1–DREQT0 are valid only when DREQ is active. I31–I0 IRDY Instruction Ready (input, synchronous) This input indicates that a valid instruction is on the instruction bus. The processor ignores this signal if there is no pending instruction access. Instruction Bus (inputs, synchronous) IREQ The instruction bus transfers instructions to the processor. Instruction Request (three-state output, synchronous) IBACK Instruction Burst Acknowledge (input, synchronous) This input is active whenever a burst-mode instruction access has been established. It may be active even though no instructions are currently being accessed. IBREQ This signal requests an instruction access. When it is active, the address for the access appears on the address bus. IREQT Instruction Request Type (three-state output, synchronous) This signal specifies the address space of an instruction request when IREQ is active: Instruction Burst Request (three-state output, synchronous) This signal is used to establish a burst-mode instruction access and to request instruction transfers during a burst-mode instruction access. IBREQ may be active even though the address bus is being used for a data access. This signal becomes valid late in the cycle with respect to IREQ. IREQT Mode 0 Instruction/data memory access 1 Instruction read-only memory access LOCK IERR Lock (three-state output, synchronous) Instruction Error (input, synchronous) This input indicates that an error occurred during the current instruction access. The processor ignores the content of the instruction bus, and an Instruction Access Exception trap occurs if the processor attempts to execute the invalid instruction. The processor ignores this signal if there is no pending instruction access. INCLK This output allows the implementation of various channel and device interlocks. It may be active only for the duration of an access, or active for an extended period of time under control of the Lock bit in the Current Processor Status. The processor does not relinquish the channel (in response to BREQ) when LOCK is active. Input Clock (input) MPGM1–MPGM0 When the processor generates the clock for the system, this is an oscillator input to the processor at twice the processor’s operating frequency. In systems where the clock is not generated by the processor, this signal must be tied High or Low, except in certain master/slave configurations. MMU Programmable (three-state outputs, synchronous) INTR3–INTR0 MSERR Interrupt Request (inputs, asynchronous) Master/Slave Error (output, synchronous) This output shows the result of the comparison of processor outputs with the signals provided internally to the off-chip drivers. If there is a difference for any enabled driver, this line is asserted. These inputs generate prioritized interrupt requests. The interrupt caused by INTR0 has the highest priority, and the interrupt caused by INTR3 has the lowest priority. The interrupt requests are masked in prioritized order by the Interrupt Mask field in the Current Processor Status Register. INTR0 cannot be masked with the Interrupt Mask field. These outputs reflect the value of two PGM bits in the Translation Look-Aside Buffer or Region Mapping Unit entry associated with the access. If no address translation is performed, these signals are both Low. OPT2–OPT0 Option Control (three-state outputs, synchronous) These outputs reflect the value of bits 18–16 of the load or store instruction that begins an access. Bit 18 of the 16 Am29050 Microprocessor AMD instruction is reflected on OPT2, bit 17 on OPT1, and bit 16 on OPT0. R/W The standard definitions of these signals (based on DREQT) are as follows (the value “x” is a “don’t care”): This signal indicates whether data is being transferred from the processor to the system, or from the system to the processor. R/W is valid only when the address bus is valid. R/W will be High when IREQ is active. DREQT1 DREQT0 OPT2 0 x 0 OPT1 OPT0 Meaning 0 0 Wordlength access 0 x 0 0 1 Byte access 0 x 0 1 0 Half-word access 0 Instruction ROM access (as data) 0 0 0 0 1 1 0 1 0 —All Others— In-circuit emulator access Reserved During an interrupt/trap vector fetch, the OPT2–OPT0 signals indicate a word-length access (000). Also, the system should return an entire aligned word for a read, regardless of the indicated data length. The Am29050 microprocessor does not explicitly prevent a store to the instruction ROM. OPT3–OPT0 are valid only when DREQ is active. Read/Write (three-state output, synchronous) RESET Reset (input, asynchronous) This input places the processor in the Reset mode. STAT2–STAT0 CPU Status (outputs, synchronous) These outputs indicate the state of the processor’s execution stage on the previous cycle. They are encoded as follows: STAT2 STAT1 STAT0 Condition 0 0 0 Halt or Step Modes 0 0 1 Pipeline Hold Mode 0 1 0 Load Test Instruction Mode, Halt/Freeze 0 1 1 Wait Mode 1 0 0 Interrupt Return 1 0 1 Taking Interrupt or Trap 1 1 0 Non-sequential Instruction Fetch 1 1 1 Executing Mode SUP/US PDA Supervisor/User Mode (three-state output, synchronous) Pipelined Data Access (three-state output, synchronous) If DREQ is not active, this output indicates that a data access is pipelined with another in-progress data access. The indicated access cannot be completed until the first access is complete. The completion of the first access is signaled by the assertion of DREQ. PEN Pipeline Enable (input, synchronous) This signal allows devices that can support pipelined accesses (i.e., that have input latches for the address and required controls) to signal that a second access may begin while the first is being completed. PIA This output indicates the program mode for an access. SYSCLK System Clock (bidirectional) This is either a clock output with a frequency that is half that of INCLK, or an input from an external clock generator at the processor’s operating frequency. TEST Test Mode (input, asynchronous) When this input is active, the processor is in Test mode. All outputs and bidirectional lines, except MSERR, are forced to the high impedance state. TRAP1–TRAP0 Pipelined Instruction Access (three-state output, synchronous) Trap Request (inputs, asynchronous) If IREQ is not active, this output indicates that an instruction access is pipelined with another in-progress instruction access. The indicated access cannot be completed until the first access is complete. The completion of the first access is signaled by the assertion of IREQ. These inputs generate prioritized trap requests. The trap caused by TRAP0 has the highest priority. These trap requests are disabled by the DA bit of the Current Processor Status Register. Am29050 Microprocessor 17 AMD WARN SPECIAL PINS Warn (input, asynchronous, edge-sensitive) The following pins are not signal pins, but are named in Am29050 microprocessor documentation because of their special role in the processor and system. A high-to-low transition on this input causes a nonmaskable WARN trap to occur. This trap bypasses the normal trap vector fetch sequence and is useful in situations where the vector fetch may not work (e.g., when data memory is faulty). PWRCLK Power Supply for SYSCLK Driver This pin is a power supply for the SYSCLK output driver. It isolates the SYSCLK driver and is used to determine whether or not the Am29050 microprocessor generates the clock for the system. If power (+5 V) is applied to this pin, the Am29050 microprocessor generates a clock on the SYSCLK output. If this pin is grounded, the Am29050 microprocessor accepts a clock generated by the system on the SYSCLK input. PIN169 Alignment pin In the PGA package, this pin is used to indicate proper pin-alignment of the Am29050 microprocessor and is used by an in-circuit emulator to communicate its presence to the system. 18 Am29050 Microprocessor AMD ABSOLUTE MAXIMUM RATINGS OPERATING RANGES Storage Temperature . . . . . . . . . . . . –65°C to +150°C Voltage on any Pin with Respect to GND . . . . . . –0.5 V to VCC +0.5 V Maximum VCC . . . . . . . . . . . . . . . . . . . . . . . . . 6.0 V DC Commercial (C) and Industrial (I) Devices Stresses above those listed under ABSOLUTE MAXIMUM RATINGS may cause permanent device failure. Functionality at or above these limits is not implied. Exposure to absolute maximum ratings for extended periods may affect device reliability. * measured “instant on” (TC=TJ) Case Temperature (TC) . . . . . . . . . . 0°C to +85°C (C) Case Temperature (TC) * . . . . . . . –40°C to +125°C (I) Supply Voltage (VCC) . . . . . . . . . . . +4.75 V to +5.25 V Operating ranges define those limits between which the functionality of the device is guaranteed. DC CHARACTERISTICS over COMMERCIAL and INDUSTRIAL Operating Ranges Symbol Parameter Description Test Conditions Min Max Unit VIL Input Low Voltage –0.5 0.8 V VIH Input High Voltage 2.0 VCC +0.5 V VILINCLK INCLK Input Low Voltage –0.5 0.8 V VIHINCLK INCLK Input High Voltage 2.0 VCC +0.5 V VILSYSCLK SYSCLK Input Low Voltage –0.5 0.8 V VIHSYSCLK SYSCLK Input High Voltage VCC–0.8 VCC +0.5 V VOL Output Low Voltage for All Outputs except SYSCLK IOL = 3.2 mA 0.45 V VOH Output High Voltage for All Outputs except SYSCLK IOH = –400 µA ILI Input Leakage Current 0.45 V ≤ VIN ≤ VCC –0.45 V ±10 µA ILO Output Leakage Current 0.45 V ≤ VOUT ≤ VCC –0.45 V ±10 µA ICCOP Operating Power Supply Current VCC = 5.25 V, Outputs Floating; Holding RESET active with externally supplied SYSCLK 38 mA/MHz VOLC SYSCLK Output Low Voltage IOLC = 20 mA 0.6 V VOHC SYSCLK Output High Voltage IOHC = –20 mA 2.4 V VCC –0.6 V CAPACITANCE Symbol CIN CINCLK CSYSCLK Parameter Description Max Unit Input Capacitance 15 pF INCLK Input Capacitance 20 pF 90 pF SYSCLK Capacitance Test Conditions fC = 1 MHz (see Note) Min COUT Output Capacitance 20 pF CI/O I/O Pin Capacitance 20 pF Note: Not 100% tested. Am29050 Microprocessor 19 AMD SWITCHING CHARACTERISTICS over COMMERCIAL Operating Range 40 MHz Parameter Description Test Conditions1 System Clock (SYSCLK) Period (T) 1A No. 33 MHz Min Max Min Max Unit Note 2 25 100 30 100 ns SYSCLK at 1.5V to SYSCLK at 1.5V when used as an output Note 14 0.5T–1 0.5T+1 0.5T–1 0.5T+1 ns 2 SYSCLK High Time when used as input Notes 14,17 15 ns 3 SYSCLK Low Time when used as input Notes 14, 17 14 ns 4 SYSCLK Rise Time Note 3 3 4 ns 5 SYSCLK Fall Time Note 3 3 4 ns 6 Synchronous SYSCLK Output Valid Delay Notes 4, 13 2 12 2 13 ns Synchronous SYSCLK Output Valid Delay for D31–D0 Notes 13, 19 2 13 2 14 ns 7 Three-State Synchronous SYSCLK Output Invalid Delay Notes 5, 15, 16 1 20 1 20 ns 8 Synchronous SYSCLK Output Valid Delay Notes 6, 13 0 8 2 10 ns Three-State SYSCLK Synchronous Output Invalid Delay Notes 6, 15, 16 1 20 1 20 ns 9 Synchronous Input Setup Time Note 8 8 10 ns 9A Synchronous Input Setup Time for D31–D0, I31–I0 3.5 4 ns 9B Synchronous Input Setup Time for DRDY 8 11 ns 10 Synchronous Input Hold Time Note 7 2.5 3.0 ns 11 Asynchronous Input Minimum Pulse Width Note 9 T+10 T+10 ns 12 INCLK Period 12.5 50 15 50 ns 12A INCLK to SYSCLK Delay 1.5 7 2 8 ns 12B INCLK to SYSCLK Delay 1.5 7 2 8 ns 1 6A 8A 13 INCLK Low Time Note 18 3 4.25 ns 14 INCLK High Time Note 18 3 4.25 ns 15 INCLK Rise Time Note 18 3.25 3.25 ns 16 INCLK Fall Time Note 18 3.25 3.25 ns 17 INCLK to Deassertion of RESET (for phase synchronization of SYSCLK) Note 10 0 5 ns 18 WARN Asynchronous Deassertion Hold Minimum Pulse Width Note 11 4T 19 BINV Synchronous Output Valid Delay from SYSCLK Note 13 0 6.0 0 6 ns 20 Three-State Synchronous SYSCLK Output Invalid Delay for D31–D0 Notes 12, 15, 16 3 20 3 20 ns 20 Am29050 Microprocessor 4 1 4T ns AMD SWITCHING CHARACTERISTICS over COMMERCIAL Operating Range 25 MHz Parameter Description Test Conditions1 System Clock (SYSCLK) Period (T) 1A No. 20 MHz Min Max Min Max Unit Note 2 40 100 50 100 ns SYSCLK at 1.5V to SYSCLK at 1.5V when used as an output Note 14 0.5T–1 0.5T+1 0.5T–1 0.5T+1 ns 2 SYSCLK High Time when used as input Note 14 19 22 ns 3 SYSCLK Low Time when used as input Note 14 17 19 ns 4 SYSCLK Rise Time Note 3 5 5 ns 5 SYSCLK Fall Time Note 3 5 5 ns 6 Synchronous SYSCLK Output Valid Delay Notes 4, 13 2 14 2 16 ns Synchronous SYSCLK Output Valid Delay for D31–D0 Notes 13, 19 2 18 2 20 ns 7 Three-State Synchronous SYSCLK Output Invalid Delay Notes 5, 15, 16 3 30 3 30 ns 8 Synchronous SYSCLK Output Valid Delay Notes 6, 13 2 14 2 16 ns Three-State SYSCLK Synchronous Output Invalid Delay Notes 6, 15, 16 3 30 3 30 ns 9 Synchronous Input Setup Time Note 8 12 15 ns 9A Synchronous Input Setup Time for D31–D0, I31–I0 6 8 ns 9B Synchronous Input Setup Time for DRDY 13 16 ns 10 Synchronous Input Hold Time Note 7 3.5 4 ns 11 Asynchronous Input Minimum Pulse Width Note 9 T+10 T+10 ns 12 INCLK Period 20 50 25 50 ns 12A INCLK to SYSCLK Delay 2 10 2 12 ns 12B INCLK to SYSCLK Delay 2 10 2 12 ns 1 6A 8A 13 INCLK Low Time 6.75 9.25 ns 14 INCLK High Time 6.75 9.25 ns 15 INCLK Rise Time 3.25 3.25 ns 16 INCLK Fall Time 3.25 3.25 ns 17 INCLK to Deassertion of RESET (for phase synchronization of SYSCLK) Note 10 1 8 ns 18 WARN Asynchronous Deassertion Hold Minimum Pulse Width Note 11 4T 19 BINV Synchronous Output Valid Delay from SYSCLK Note 13 0 7 0 9 ns 20 Three-State Synchronous SYSCLK Output Invalid Delay for D31–D0 Notes 12, 15, 16 3 20 3 25 ns Am29050 Microprocessor 7 1 4T ns 21 AMD SWITCHING CHARACTERISTICS over INDUSTRIAL Operating Range 33 MHz No. Parameter Description Test Conditions1 System Clock (SYSCLK) Period (T) Min Max Unit Note 2 30 100 ns SYSCLK at 1.5V to SYSCLK at 1.5V when used as an output Note 14 0.5T–1 0.5T+1 ns 2 SYSCLK High Time when used as input Note 14 15 ns 3 SYSCLK Low Time when used as input Note 14 14 ns 4 SYSCLK Rise Time Note 3 3 ns 5 SYSCLK Fall Time Note 3 3 ns 6 Synchronous SYSCLK Output Valid Delay Notes 4, 13 2 13 ns Synchronous SYSCLK Output Valid Delay for D31–D0 Notes 13, 19 2 14 ns 7 Three-State Synchronous SYSCLK Output Invalid Delay Notes 5, 15, 16 1 20 ns 8 Synchronous SYSCLK Output Valid Delay Notes 6, 13 2 10 ns Three-State SYSCLK Synchronous Output Invalid Delay Notes 6, 15, 16 1 20 ns Synchronous Input Setup Time Note 8 10 ns 1 1A 6A 8A 9 9A Synchronous Input Setup Time for D31–D0, I31–I0 4 ns 9B Synchronous Input Setup Time for DRDY 11 ns 10 Synchronous Input Hold Time Note 7 3 ns 11 Asynchronous Input Minimum Pulse Width Note 9 T+10 ns 12 INCLK Period 15 50 ns 12A INCLK to SYSCLK Delay 2 8 ns 12B INCLK to SYSCLK Delay 2 8 ns 13 INCLK Low Time 4.25 ns 14 INCLK High Time 4.25 ns 15 INCLK Rise Time 3.25 ns 16 INCLK Fall Time 3.25 ns 17 INCLK to Deassertion of RESET (for phase synchronization of SYSCLK) Note 10 1 5 ns 18 WARN Asynchronous Deassertion Hold Minimum Pulse Width Note 11 4T 19 BINV Synchronous Output Valid Delay from SYSCLK Note 13 0 6 ns 20 Three-State Synchronous SYSCLK Output Invalid Delay for D31–D0 Notes 12, 15, 16 3 20 ns 22 Am29050 Microprocessor ns AMD SWITCHING CHARACTERISTICS over INDUSTRIAL Operating Range 25 MHz Parameter Description Test Conditions1 System Clock (SYSCLK) Period (T) 1A No. 20 MHz Min Max Min Max Unit Note 2 40 100 50 100 ns SYSCLK at 1.5V to SYSCLK at 1.5V when used as an output Note 14 0.5T–1 0.5T+1 0.5T–1 0.5T+1 ns 2 SYSCLK High Time when used as input Note 14 19 22 ns 3 SYSCLK Low Time when used as input Note 14 17 19 ns 4 SYSCLK Rise Time Note 3 4 4 ns 5 SYSCLK Fall Time Note 3 4 4 ns 6 Synchronous SYSCLK Output Valid Delay Notes 4, 13 2 14 2 16 ns Synchronous SYSCLK Output Valid Delay for D31–D0 Notes 13, 19 2 18 2 20 ns 7 Three-State Synchronous SYSCLK Output Invalid Delay Notes 5, 15, 16 3 30 3 30 ns 8 Synchronous SYSCLK Output Valid Delay Notes 6, 13 2 14 2 16 ns Three-State SYSCLK Synchronous Output Invalid Delay Notes 6, 15, 16 3 30 3 30 ns 9 Synchronous Input Setup Time Note 8 12 15 ns 9A Synchronous Input Setup Time for D31–D0, I31–I0 6 8 ns 9B Synchronous Input Setup Time for DRDY 13 16 ns 10 Synchronous Input Hold Time Note 7 3.5 4 ns 11 Asynchronous Input Minimum Pulse Width Note 9 T+10 T+10 ns 12 INCLK Period 20 50 25 50 ns 12A INCLK to SYSCLK Delay 2 10 2 12 ns 12B INCLK to SYSCLK Delay 2 10 2 12 ns 1 6A 8A 13 INCLK Low Time 6.75 9.25 ns 14 INCLK High Time 6.75 9.25 ns 15 INCLK Rise Time 3.25 3.25 ns 16 INCLK Fall Time 3.25 3.25 ns 17 INCLK to Deassertion of RESET (for phase synchronization of SYSCLK) Note 10 1 8 ns 18 WARN Asynchronous Deassertion Hold Minimum Pulse Width Note 11 4T 19 BINV Synchronous Output Valid Delay from SYSCLK Note 13 0 7 0 9 ns 20 Three-State Synchronous SYSCLK Output Invalid Delay for D31–D0 Notes 12, 14, 15 3 20 3 25 ns Am29050 Microprocessor 7 1 4T ns 23 AMD Notes: 1. Test conditions: All inputs/outputs are TTL compatible for VIH , VIL , VOH , and VOL unless otherwise noted. All output timing specifications are for 80 pF of loading. All setup, hold, and delay times are measured relative to SYSCLK or INCLK unless otherwise noted. All input Low levels must be driven to 0.45 V and all input High levels must be driven to 2.4 V except SYSCLK. 2. AC measurements made relative to 1.5 V, except where noted. 3. SYSCLK rise and fall times measured between 0.8 V and (VCC – 1.0 V). 4. Synchronous Outputs relative to SYSCLK rising edge include: A31–A0, BGRT, R/W, SUP/US, LOCK, MPGM1–MPGM0, IREQ, IREQT, PIA, DREQ, DREQT1–DREQT0, PDA, OPT2–OPT0, STAT2–STAT0, and MSERR. 5. Three-state Synchronous Outputs relative to SYSCLK rising edge include: A31–A0, R/W, SUP/US, LOCK, MPGM1–MPGM0, IREQ, IREQT, PIA, DREQ, DREQT1–DREQT0, PDA, and OPT2–OPT0. 6. Synchronous Outputs relative to SYSCLK falling edge (SYSCLK): IBREQ, DBREQ. 7. Synchronous Inputs include: BREQ, PEN, IRDY, IERR, IBACK, DERR, DBACK, CDA, I31–I0, DRDY, and D31–D0. 8. Synchronous Inputs include: BREQ, PEN, IRDY, IERR, IBACK, DERR, DBACK, and CDA. 9. Asynchronous Inputs include: WARN, INTR3–INTR0, TRAP3–TRAP0, and CNTL1–CNTL0. 10. RESET is an asynchronous input on assertion/deassertion. As an option to the user, RESET deassertion can be used to force the state of the internal divide-by-two flip-flop to synchronize the phase of SYSCLK (if internally generated) relative to RESET/ INCLK. 11. WARN has a minimum pulse width requirement upon deassertion. 12. To guarantee Store/Load with one-cycle memories, D31–D0 must be asserted relative to SYSCLK falling edge from an external drive source. 13. Refer to Capacitive Output Delay table when capacitive loads exceed 80 pF. 14. When used as an input, SYSCLK presents a 90-pF max load to the external driver. When SYSCLK is used as an output, timing is specified with an external load capacitance of ≤ 200 pF. For frequencies ≤ 33 MHz, the measurement point is 1.5 V. For frequencies > 33 MHz, the measurement point is 2.5 V. 15. Three-State Output Inactive Test Load. Three-State Synchronous Output Invalid Delay is measured as the time to a ±500 mV change from prior output level. 16. When a three-state output makes a synchronous transition from a valid logic level to a high-impedance state, data are guaranteed to be held valid for an amount of time equal to the lesser of the minimum Three-State Synchronous Output Invalid Delay and the minimum Synchronous Output Valid Delay. 17. SYSCLK is not available as an input at 40 MHz. 18. Performance guaranteed by design for 40 MHz part. 19. All unused outputs should be terminated. 24 Am29050 Microprocessor AMD SWITCHING WAVEFORMS SYSCLK 1.5 V 1.5 V 1.5 V 1.5 V 8 8A SYSCLK Synchronous Outputs 1.5 V 6& 6A SYSCLK Synchronous Outputs 7& 20 1.5 V 19 BINV 1.5 V 9,9A & 9B 10 Synchronous Inputs 1.5 V 1.5 V Relative to SYSCLK Am29050 Microprocessor 25 AMD SWITCHING WAVEFORMS (continued) 13 14 16 15 INCLK 2.0 V 1.5 V 0.8 V 2.0 V 1.5 V 0.8 V 1.5 V 0.8 V 12 17 RESET 1.5 V 18 WARN 1.5 V 1.5 V 11 Asynchronous Inputs 1.5 V INCLK and Asynchronous Inputs 26 Am29050 Microprocessor 1.5 V AMD SWITCHING WAVEFORMS (continued) 1 1A 3 2 VCC –1.0 V VCC –1.0 V 1.5 V 1.5 V 1.5 V 0.8 V 0.8 V 4 5 SYSCLK Definition for 33 MHz and Below 12B 12A 1.5 V 1.5 V SYSCLK 2.0 V 2.0 V 1.5 V INCLK 1.5 V 0.8 V 0.8 V 16 15 12 INCLK to SYSCLK Delay Am29050 Microprocessor 27 AMD CAPACITIVE OUTPUT DELAYS For loads greater than 80 pF This table describes the additional output delays for capacitive loads greater than 80 pF. Values in the Maximum Additional Delay column should be added to the value listed in the Switching Characteristics table. For loads less than or equal to 80 pF, refer to the delays listed in the Switching Characteristics table. Delay values are based on model simulation and are not guaranteed. No. Total External Capacitance Parameter Description 6 Synchronous SYSCLK Output Valid Delay 100 pF 150 pF 200 pF 250 pF 300 pF +1 ns +2 ns +4 ns +6 ns +8 ns 8 Synchronous SYSCLK Output Valid Delay 100 pF 150 pF 200 pF 250 pF 300 pF +1 ns +2 ns +4 ns +6 ns +8 ns 19 BINV Synchronous Output Valid Delay from SYSCLK 100 pF 150 pF 200 pF 250 pF 300 pF +1 ns +3 ns +4 ns +6 ns +7 ns SWITCHING TEST CIRCUIT VL IOL = 3.2 mA CL V Am29050 Pin Under Test VREF = 1.5 V IOH = 400 µA VH CL is guaranteed to 80 pF. For capacitive loading greater than 80 pF, refer to the Capacitive Output Delay table. 28 Maximum Additional Delay Am29050 Microprocessor AMD THERMAL CHARACTERISTICS Pin Grid Array (PGA) Package θJA θCA θJC ÉÉÉÉ θJA = θJC + θCA Thermal Resistance — °C/Watt Airflow — ft./min. (m/sec) Parameter θJC Junction-to-Case θCA Case-to-Ambient (no Heatsink) θCA Case-to-Ambient (with omnidirectional 3-fin Heatsink) 0 (0) 150 (0.76) 300 (1.53) 480 (2.45) 700 (3.58) 2.4 2.4 2.4 2.4 2.4 14.7 12.9 11.4 10.0 8.8 8.6 6.6 3.7 2.1 1.2 Am29050 Microprocessor 29 AMD PHYSICAL DIMENSIONS CGX 169 Ceramic Pin Grid Array 1.600 BSC 1.740 1.780 –A– –B– b1= 0.080 Max 1.740 1.780 1.600 BSC Lid Outline Cavity Down PGA 0.000 (4x) BSC Index Corner 0.003 Min (4x) View of PGA Pin Matrix for Both Small and Large Outline Base Plane –C– Heatsink Option Lid 0.420 Max 0.120 0.140 0.025 0.055 0.080 0.140 0.105 0.195 Side View of a Cavity-Down PGA Notes: 1. All measurements are in inches unless otherwise noted. 2. Not to scale. For reference only. 30 Am29050 Microprocessor AMD Trademarks AMD, Am29000, and Fusion29K are registered trademarks; and Am29005, Am29030, Am29035, Am29040, Am29050, Am29200, Am29205, Am29240, Am29243, Am29245, 29K, MiniMON29K, and XRAY29K are trademarks of Advanced Micro Devices, Inc. High C is a registered trademark of MetaWare, Inc. Product names used in this publication are for identification purposes only and may be trademarks of their respective companies. Copyright 1994 Advanced Micro Devices. All rights reserved. Am29050 Microprocessor 31