作者: Jack J. Dongarra , Piotr Luszczek
DOI: 10.21236/ADA439315
关键词:
摘要: Abstract : The HPCChallenge suite of benchmarks will examine the performance HPC architectures using kernels with memory access patterns more challenging than those High Performance Linpack (HPL) benchmark used in Top5OO list. is being designed to augment list, provide that bound many real applications as a function characteristics, e.g., spatial and temporal locality, framework for including additional benchmarks. are scalable size data sets largest HPL matrix system. has been released by DARPA HPCS program help define boundaries future Petascale computing systems. composed several well known computational (STREAM, Linpack, multiply - DGEMM, transpose, FFT, RandomAccess, bandwidth/latency tests) attempt span high low locality space.