Hello Mike,
Am 26.06.2008, 03:34 Uhr, schrieb Stoner, Michael
<
michael.stoner@...>:
> Hello,
> I'm doing some performance analysis for Intel and I'd like to know if
> most users are building LAME with the Takehiro magic float hack enabled?
I would say it's in 99 percent the case. The others are those users
that own something different from a x86 system.
> I found that disabling the hack improved my encode time by about 15%. I
> used the VC++ 2005 compiler with /arch:SSE2 and /fp:fast switches.
My own speed tests were some time ago, but if I remember correct,
enabling SSE2 was a good way to make LAME running slower (using VC2005).
> Here's my data running on Nehalem 2.8 Ghz and Merom 2.93 Ghz cores:
>
> Workload: lame.exe -h -b160 --nores file.wav
>
> NHM MRM
> Baseline w/ Takehiro hack - 50 sec 51 sec
> Hack + /fp:fast /arch:SSE2 - 46 sec 47 sec
> w/o Hack + /fp:fast /arch:SSE2 - 41 sec 40 sec
>
>
> Thanks,
> Mike Stoner
> Intel Software Solutions Group
> Folsom, CA
> (916) 356-9414
Ciao Robert
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php_______________________________________________
Lame-dev mailing list
Lame-dev@...
https://lists.sourceforge.net/lists/listinfo/lame-dev
