README
1
2 SciMark 2.0 Java Numerical Benchmark
3
4 Roldan Pozo, Bruce Miller
5
6 NIST
7
8SciMark 2.0 is a composite Java benchmark measuring the performance of
9numerical kernels occurring in scientific and engineering applications.
10It consists of five kernels which typify computational routines
11commonly found in numeric codes: Fast Fourier Transforms (FFTs),
12Jacobi Successive Over-relaxation (SOR), Sparse matrix-multiply,
13Monte Carlo integration, and dense LU matrix factorization.
14
15(See http://www.math.nist.gov/scimark for further information
16and latest updates.)
17
18
191) INSTALLATION
20
21Unpack the contents of archive into a subdirectory on your
22CLASSPATH. Be sure to keep the directory structure of the
23file contents.
24
252) COMPILING THE BENCHMARKS (optional)
26
27From the directory above this one, issue the command:
28
29 >javac -O commandline.java
30
31This should compile main benchmark driver and dependent files.
32
333) RUNNING THE BENCHMARKS
34
35From the directory above this one, issue the command:
36
37 >java jnt.scimark2.commandline
38
39or
40 >java jnt.scimark2.commandline -large
41
42to run the large problem size version. (Note that this one
43takes considerably longer to run.)
44
45After a few minutes, the program should respond with
46the benchmark results, e.g.
47
48 >javac jnt.scimark2.commandline
49
50 SciMark 2.0a
51
52 Composite Score: 20.791595999749727
53 FFT (4096): 30.260047144878346
54 Jacobi SOR (100x100): 33.074935359763934
55 Monte Carlo (25000): 11.510791361970528
56 Sparse matmult (nz=25000), 10 iterations: 8.007507030681996
57 LU (100x100): 21.104699101453836
58
59 java.vendor: Sun Microsystems Inc.
60 java.version: 1.2
61 os.arch: x86
62 os.name: Windows NT
63 os.version: 4.0
64
65One can send these results to "pozo@nist.gov".
66
67/* ---------------------- END OF README -----------------------------*/
68