|scimark2_1c.zip: compressed source code.||scimark4_c.zip: compressed source code.|
This is an ANSI C version of the SciMark2 benchmark, translated from the original Java sources. The intent in making this benchmark available in C is mainly for performance comparisons.
Version 4.0 of the software has added check-sum counters to adequately test newer optimizing compilers.
Version 5.0 will provide option for single and double precision benchmarks.
Results of this benchmark can be sent to firstname.lastname@example.org.
The program is split up into the main driver (scimark2.c) and kernel routines. A sample makefile is included; however, one could simply write
> cc -o scimark2 -O *.cand then run
> scimark2This produces an output similar to
** ** ** SciMark2 Numeric Benchmark, see http://math.nist.gov/scimark ** ** for details. (Results can be submitted to email@example.com) ** ** ** Using 2.00 seconds min time per kenel. Composite Score: 65.56 FFT Mflops: 63.38 (N=1024) SOR Mflops: 124.80 (100 x 100) MonteCarlo: Mflops: 16.05 Sparse matmult Mflops: 59.15 (N=1000, nz=5000) LU Mflops: 64.40 (M=100, N=100) 0:29.62 Elapsed, 29.620 user sec, 0.010 sys sec, 100.0% utilization.
The first SciMark number reported is the composite score, followed by the an approximate Mflop rate for each kernel.
To run the "large" version of this benchmark (with data structures that typically do not fit in cache) use
Commercial equipment and software is referred to on these pages in order to precisely document experimental measurements. Mention of these products does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the products so identified are necessarily the best available for that purpose.