Parallel Computer Architecture

A.5 PARKBENCH 967

Because the NAS benchmarks, particularly the kernels, are relatively easy to

implement and represent an interesting range of computations for scientific comput-

ing, they have been widely embraced by multiprocessor vendors. With the portabil-

ity provided by the Message Passing Interface (MPI) standard, a recent follow-on

effort called NPB2 utilizes fixed applications written in MPI rather than pencil-and-

paper benchmarks, just like traditional benchmark suites. The benchmarks and

their documentation can be obtained from NAS, as described in Table A.l.

PARKBENCH

The PARKBENCH (PARallel Kernels and BENCHmarks) effort (PARKBENC ...