On Mar 15, 1:16 pm, Jeff Gilchrist <jeff.gilchr...@gmail.com> wrote:
> So I was unable to run the latest speed tests on the Linux 64bit live
> CD that works on my Q9550 since it only contains Python 2.5 but I did
> get a chance to run some benchmarks one of my GMP programs so we have
> some real-world tests on a subset of MPIR. You can see a very nice
> progression from MPIR 0.9.0 to the almost 1.0 code (SVN 1728). I was
> expecting a big change in Core2 code but I was also surprised to see
> the AMD64 code got a nice speed jump since 0.9.0 as well.
>
> MPIR Benchmarks
> ===============
> mpz_mul(t, r, r);
> mpz_sub_ui(t, t, 2);
> mpz_mod(r, t, p);
>
> Benchmarked values are: 83338 bits (25088 digits) and 125002 bits (37630
> digits)
>
> Intel Core2 Q9550 @ 3.4GHz (45nm, Family 6, Model 7, Stepping 7)
> 83338 bits (25088 digits)
> Library Time1 Time2
> ------------------------- --------- ---------
> MPIR SVN1728 (gcc 64bit) 1m45.864s 1m45.722s
> MPIR SVN1728 (MSVC 64bit) 1m47.063s 1m47.095s
> GMP 4.2.4 (MSVC 64bit) 1m53.472s 1m53.505s
> MPIR 0.9.0 (MSVC 64bit) 1m53.709s 1m53.709s
> MPIR 0.9.0 (gcc 64bit) 2m04.794s 2m04.859s
> GMP 4.2.4 (gcc 64bit) 3m19.200s 3m19.089s
> GMP 4.2.4 (MSVC 32bit) 6m02.295s 6m02.280s
> MPIR SVN1728 (MSVC 32bit) 6m02.435s 6m02.482s
> MPIR 0.9.0 (MSVC 32bit) 6m02.545s 6m02.405s
>
> 125002 bits (37630 digits)
> Library Time1 Time2
> ------------------------- --------- ---------
> MPIR SVN1728 (gcc 64bit) 4m51.502s 4m51.395s
> MPIR SVN1728 (MSVC 64bit) 4m57.288s 4m57.345s
> MPIR 0.9.0 (MSVC 64bit) 5m14.861s 5m15.095s
> GMP 4.2.4 (MSVC 64bit) 5m17.398s 5m17.237s
> MPIR 0.9.0 (gcc 64bit) 5m49.962s 5m49.823s
> GMP 4.2.4 (gcc 64bit) 9m23.591s 9m23.568s
>
> AMD Opteron 248 @ 2.2GHz (Family 15, Model 37, Stepping 1)
> 83338 bits (25088 digits)
> Library Time1 Time2
> ------------------------- --------- ---------
> MPIR SVN1728 (gcc 64bit) 1m37.092s 1m36.801s
> MPIR 0.9.0 (gcc 64bit) 2m13.795s 2m13.481s
> GMP 4.2.4 (gcc 64bit) 2m56.692s 2m56.340s
>
> 125002 bits (37630 digits)
> Library Time1 Time2
> ------------------------- --------- ---------
> MPIR SVN1728 (gcc 64bit) 4m29.804s 4m30.865s
> MPIR 0.9.0 (gcc 64bit) 6m09.441s 6m08.634s
> GMP 4.2.4 (gcc 64bit) 8m11.460s 8m08.912s
i Jeff,
Thank you for your efforts in obtaining these figures, which are very
informative.
GMP 4.2.4 and MPIR 0.9.0 are essentially the same code on Windows so
its not a surprise to find that their performance is very close.
Moreover there has not been (and won't be, at least by me) any
development of the 32-bit Windows assembler code so speed improvements
won't be seen here.
Brian
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"mpir-devel" group.
To post to this group, send email to mpir-devel@googlegroups.com
To unsubscribe from this group, send email to
mpir-devel+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/mpir-devel?hl=en
-~----------~----~----~----~------~----~------~--~---
