LargeIntegers plugin affects benchmark
Dwight Hughes
dwighth at ipa.net
Sun Jun 18 16:50:40 UTC 2000
#tinyBenchmarks is just running #benchmark and #benchFib - neither of
these creates LargeIntegers. For my system (a PII/400) I'm getting:
42049934 bytecodes/sec; 1146059 sends/sec
Checking the arguments #tinyBenchmarks is passing on my system (it
scales the values to make each benchmark run for at least 1 second on
any machine): 128 for benchmark; 29 for benchFib. A larger receiver for
#benchmark simply increases the number of iterations through the seive;
29 benchFib is only 1664079 and no larger values are produced during
execution.
I assume you are on a Windows VM -- Andreas changed over to a GCC based
set of tools to generate the most recent VMs for Win32, which allowed
some crucial optimizations for Pentiums with a quite significant
performance increase (putting things more on par with the PowerPC based
VMs).
-- Dwight
Henrik Gedenryd wrote:
>
> 2.8 reports a bytecode rate almost 50% higher than before for
> Integer>>tinyBenchmark, and I'm guessing that this is caused by the
> LargeIntegers plugin.
>
> In any case, if this is true I'm not well enough initiated to fix this.
>
> Henrik
More information about the Squeak-dev
mailing list
|