论文部分内容阅读
1989年2月,Intel 公司推出了最新的80860(或S8860)芯片,它是一个含有100万只晶体管的64位RISC CPU,工作速率为50MHz.目前大多数RISC 处理机都使用若干块芯片;例如,SPARC CPU 使用5块不同的芯片,Motorola 的88000由3块芯片组成,包括指令和数据高速缓存.但是,80860把所有功能集中在一块芯片上.该CPU 的1/3是数据和指令缓存;还有1/3处理整数指令.最后1/3容纳浮点加法器和乘法器单元.这种片载缓存可以使数据传送速率达到近1000兆字节/秒,比使用外部缓存芯片快得多.3个独立的FPU(浮点处理单元)可以并行工作,在40MHz 时每秒可完成多达1.2亿次操作.FPU 模仿Cray 的流水线向量结构.在对一组数据执行
In February 1989, Intel introduced the latest 80860 (or S8860) chip, a 64-bit RISC CPU with 1 million transistors operating at 50 MHz, which is currently used by most RISC processors; for example, , The SPARC CPU uses five different chips and the Motorola 88000 consists of three chips, including instruction and data caches, but the 80860 has all the features on a single chip, one-third of which is data and instruction cache, One-third deal with integer instructions and the last one-third with floating-point adder and multiplier units. This on-chip cache enables data transfer rates to nearly 1,000 megabytes per second, much faster than using external cache chips Three independent FPUs (floating-point processing units) can work in parallel to achieve up to 120 million operations per second at 40 MHz FPU mimics Cray’s pipeline vector structure When executing a set of data