I just benchmarked RC3 with compute intensive matrix computations and
compared against PreV2. Didn't yet get round to run on IBM.
Both perform well and visibly suffer from the small 256 KB Level-2 Cache
of the PentiumIII (as opposed to the UltraSparcII with 4 MB L2). Most
results show no significant difference. RC3 shows some speedup for
matrix-matrix mult ("zMult") and equation solving ("solve"), but perhaps
still not significant.
Note that these results are only representative for certain application
types and should not be overly generalized. They mainly benchmark
floating point calculations, invocation of small methods as well as the
speed of the memory hierarchy.
Here are the results:
http://nicewww.cern.ch/~hoschek/colt/V1.0Beta4/doc/cern/colt/matrix/doc-files/PerformanceLogFrame.html
and the sources:
http://nicewww.cern.ch/~hoschek/colt/V1.0Beta4/download/matrixBenchmark.zip
Wolfgang.
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
