On Mon, 11 Nov 2002 14:13:34 +1100, "Sisyphus" <[EMAIL PROTECTED]> wrote:
>> Yes, because gcc 2.95.2 is really a worse compiler compared to VC++ 6. It >> runs a lot slower, but more importantly, the code it generates is also >> about 20% slower than the code that VC++ generates. > >This is not what I just found. >I have 3 perls currently: >1) 5.6.1 built from AS build 626 source with MSVC++ 6.0. >2) 5.6.1 built from AS build 626 source with dmake and mingw. >3) 5.8.0 built from cpan source with MSVC++ 6.0. > >I timed, using Benchmark, 300 iterations of the subroutine given below. > >1) took 4.64 seconds >2) took 3.12 seconds >3) took 5.31 seconds Interesting. I did some digging, but only found my notes from testing Perl 5.005_53 with EGCS (gcc 2.91.57). I thought I did some later testing with Perl 5.6 and gcc 2.95, but can't find any logs anymore. Maybe I should rerun some tests again... BTW, I was using the perlbench tests for comparison, and I think running pod2html on perlfunc. >So the mingw build was about 25% faster than the comparable msvc build. Are you sure you used the same compilation options? Especially, did you set all 3 of USE_MULTI, USE_ITHREADS and USE_IMP_SYS? >Am I missing something ? >Is my msvc-built perl suffering from having been built with the Standard >Edition (no Service Packs added, afaik) ? Service packs shouldn't make a difference. I cannot remember if the "Standard Edition" of VC++ 6 supports optimization or not. Did you get any warning from using the -O2 switch? Cheers, -Jan _______________________________________________ Perl-Win32-Users mailing list [EMAIL PROTECTED] To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs