Performance Evaluation of Massively Parallel Systems Using SPECOMP Suite

Performance analysis plays an essential role in achieving a scalable performance of applications on massively parallel supercomputers equipped with thousands processors. This paper is empirical investigation to study, depth, the two most common High-Performance Computing architectures world. IBM has developed three generations Blue Gene supercomputers—Blue Gene/L, P, and Q—that use, at large scale, low-power processors achieve high performance. Better CPU core efficiency been empowered by higher level integration gain more parallelism per processing element. On other hand, Intel Xeon Phi coprocessor armed 61 on-chip x86 cores, provides theoretical peak performance, as well software development flexibility existing high-level programming tools. We present extensive evaluation study peaks scalability these modern using SPEC OMP benchmarks.