论文部分内容阅读
在过去的28年中,超级计算机经历了设计上的三次革命:向量超级计算;多CPU上的并行超级计算;分层组织的、带有高速缓存的微处理器群集上的超级计算。 最后一次革命至今仍在进行当中,所以它带来的结果尚不能确定。可以说,超级计算机设计上的变化推动了计算方法及其实现程序的重大转变。 向量超级计算 向量超级计算的概念是由CDC STAR超级计算机的设计者引入的,并在1967年的时候,被LawrenceLivermore国家实验室认可。这种想法是为了克服当时磁芯存储器存在的延时问题而出现的。1971年前后,Frank McMahon将这种想法成功地应用于
Over the past 28 years, supercomputers have undergone three revolutions in design: vector supercomputing; parallel supercomputing on multiple CPUs; and supercomputing in hierarchical, microprocessor-based caches with caches. The last revolution is still in progress, so the result it brings is yet to be determined. It can be said that the supercomputer design changes have driven a major shift in computing methods and their implementation processes. Vector Supercomputing Vector The concept of supercomputing was introduced by the designers of the CDC STAR supercomputer and was recognized by Lawrence Livermore National Laboratory in 1967. This idea is to overcome the existence of the core memory delay problems arise. Around 1971 Frank McMahon successfully applied this idea