大家来聊聊这三款CPU.它们性能如何,能否做个比较.下面是性能比较列表。<br />这三种中,OR1K,LEON,都是开源的,MIPS用的较多,是否开源?我没找到。<br /><b>MIPS32 4KE </b>:<br />32-bit MIPS32 privileged resource architecture <br />32-bit address and data paths <br />5-stage pipeline <br />32 general purpose registers <br />Supervisor mode operation <br />MIPS32-compatible instruction set <br />MIPS16e code compression <br />Special PC-relative instructions for efficient loading of addresses and constants <br />SAVE & RESTORE macro instructions for setting up and tearing down stack frames within subroutines <br />Programmable cache sizes <br />Individually configurable instruction and data caches, sizes from 0 - 64KB <br />Direct mapped, 2-, 3-, or 4-way set associative <br />Loads block only until critical word is available <br />Non-blocking prefetches <br />Write-back support <br />16-byte cache line size <br />Memory management unit <br />16 dual entry JTLB with variable page size <br />Fixed or TLB-based MMU, dependent on family member <br />Scratchpad RAM support <br />Can optionally replace 1 way of the I- and/or D-cache with a fast scratchpad RAM <br />20 index address bits allow access of arrays up to 1MB <br />Interface allows back-stalling the core <br />Simple bus interface unit (BIU) <br />All I/Os fully registered <br />Separate unidirectional 32-bit address and data buses <br />Two 16-byte collapsing write buffers <br />Designed to allow easy conversion to other bus protocols <br />Integer multiply/divide unit e <br />Fast or area efficient MDU, dependent on family member <br />Maximum issue rate of one 32x16 multiply per clock (fast MDU) <br />Maximum issue rate of one 32x32 multiply every other clock (fast MDU) <br />General purpose coprocessor (COP2) interface <br />32-bit interface to an external coprocessor <br />Power control <br />Minimum frequency: 0 MHz <br />Power-down mode (triggered by WAIT instruction) <br />Support for software-controlled clock divider <br />Support for extensive use of local gated clocks <br />EJTAG debug <br />Support for single stepping <br />Virtual instruction and data address breakpoints <br />PC and data tracing <br /><br />Benefits : <br /> Process: 0.13um G <br /> Frequency Typical: 260-300MHz <br /> Frequency Worst-case: 210-255MHz <br /> Typical Performance 405 Dhrystone MIPS <br /> Power Consumption: 0.12 - 0.37 mW/MHz at 1.2V (Core Only) <br /> Core Size: 0.4 - 1.9 mm(2) (core only) <br /> Note: Frequency, power consumption and size depend upon configuration options, synthesis, silicon vendor, process and cell libraries. <br /> <br /><br /><b>:LEON3</b>:<br />The LEON3 is a synthesisable VHDL model of a 32-bit processor compliant with the SPARC V8 architecture. The model is highly configurable, and particularly suitable for system-on-a-chip (SOC) designs. The full source code is available under the GNU GPL license, allowing free and unlimited use for research and education. LEON3 is also available under a low-cost commercial license, allowing it to be used in any commercial application to a fraction of the cost of comparable IP cores. The LEON3 processor has the following features:<br /><br />SPARC V8 instruction set with V8e extensions<br />Advanced 7-stage pipeline <br />Hardware multiply, divide and MAC units <br />High-performance, fully pipelined IEEE-754 FPU <br />Separate instruction and data cache (Harvard architecture) with snooping <br />Configurable caches: 1 - 4 sets, 1 - 256 kbytes/set. Random, LRR or LRU replacement<br />Local instruction and data scratch pad rams<br />SPARC Reference MMU (SRMMU) with configurable TLB<br />AMBA-2.0 AHB bus interface<br />Advanced on-chip debug support with instruction and data trace buffer <br />Symmetric Multi-processor support (SMP)<br />Power-down mode <br />Robust and fully synchronous single-edge clock design <br />Up to 125 MHz in FPGA and 400 MHz on 0.13 um ASIC technologies<br />Fault-tolerant and SEU-proof version available for space applications<br />Extensively configurable <br />Large range of software tools: compilers, kernels, simulators and debug monitors<br /><br />The LEON3 processor is distributed as part of the GRLIB IP library, allowing simple integration into complex SOC designs. GRLIB also includes a configurable LEON3 multi-processor design, with up to 4 CPU's and a large range of on-chip peripheral blocks.<br /><br /><br /><b>OpenRISC 1000 </b>:<br /><br />The OpenRISC 1000 architecture is the latest in<br />the development of modern open architectures<br />and the base for a family of 32- and 64-bit<br />RISC/DSP processors. Open architecture allows<br />a spectrum of chip and system implementations<br />at a variety of price/performance points for a<br />range of applications. Designed with emphasis<br />on performance, simplicity, low power<br />consumption, scalability, and versatility, it<br />targets medium and high performance<br />networking, portable, embedded, and automotive<br />applications.<br />High Performance 32-Bit CPU/DSP<br />· 32-bit architecture implementing ORBIS32<br />instruction set<br />· Scalar, single-issue 5-stage pipeline<br />delivering sustained throughput<br />· Single-cycle instruction execution on most<br />instructions<br />· 250 MIPS performance @ 250MHz worstcase<br />conditions<br />· Predictable execution rate for hard real-time<br />applications<br />· Fast and deterministic internal interrupt<br />response<br />· Thirty-two, 32-bit general-purpose registers<br />· DSP MAC 32x32<br />· Custom user instructions<br />L1 Caches<br />· Harvard model with split instruction and<br />data cache<br />· Instruction/data cache size scalable from<br />1KB to 64KB<br />· Physically tagged and addressed<br />· Cache management special-purpose<br />registers<br />Memory Management Unit<br />· Harvard model with split instruction and<br />data MMU<br />· Instruction/data TLB size scalable from 16<br />to 256 entries<br />· Direct-mapped hash-based TLB<br />· Linear address space with 32-bit virtual<br />address and physical address from 24 to 32<br />bits<br />· Page size 8KB with per-page attributes<br />Sophisticated Power Management Unit<br />· Power reduction from 2x to 100x<br />· Software controlled clock frequency in slow<br />and idle modes<br />· Interrupt wake-up in doze and sleep modes<br />· Dynamic clock gating for individual units<br />Advanced Debug Unit<br />· Conventional target-debug agent with a<br />debug exception handler<br />· Non-intrusive debug/trace for both RISC<br />and system<br />· Real-time trace of RISC and system<br />· Access and control of debug unit from RISC<br />or via development interface<br />· Complex chained watchpoint and breakpoint<br />conditions<br />Integrated Tick Timer<br />· Task scheduling and precise time measuring<br />· Maximum timer range of 2^32 clock cycles<br />· Maskable tick-timer interrupt<br />· Single-run, restartable or continuous mode<br />Programmable Interrupt Controller<br />· 2 non-maskable interrupt sources<br />· 30 maskable interrupt sources<br />· two interrupt priorities<br />Custom and Optional Units<br />· Additional units such as a floating-point unit<br />can be added as standard units<br />· 8 custom units can be added and controlled<br />through special-purpose registers or<br />customer instructions<br />Specifications<br />· 250 MHz in worst-case 0.18u 6LM<br />The OpenRISC 1200 Processor Core is ideally<br />suited for applications that require 32-bit<br />performance compared to performance of 16-bit<br />processors and need low cost and low power<br />consumption advantage compared to 64-bit<br />processors.<br />· 250 MIPS Dhrystone 2.1 @ 250MHz wc<br />· 250 MMAC operations @ 250MHz wc<br />· <1W @ 250MHz, 0.18u, full throttle (est)<br />· <500mW @ 250MHz, 0.18u, half throttle<br />(est)<br />· Area <0.5 sqmm @ 0.18u 6LM (cache<br />memories not included)<br />Target Applications<br />· Internet, networking and telecom<br />applications<br />· Embedded applications<br />· Portable and wireless applications<br />· Home entertainment consumer electronics<br />· Automotive applications |
|