On Fri, Oct 04, 2013 at 10:53:40AM +0200, Diego Biurrun wrote:
> On Tue, Oct 01, 2013 at 02:49:21PM -0400, Justin Ruggles wrote:
> > On 10/01/2013 01:44 PM, Diego Biurrun wrote:
> > >The mmxext optimizations should be faster if available and should
> > >thus override the amd3dnow optimizations, not the other way around.
> > 
> > Should be faster? or are faster?
> 
> 17:34 <@DonDiego> pengvado, Skyler_, BBB, can there be cases where amd3dnow 
> is faster than mmxext?
> 13:20 <@pengvado> DonDiego: any instruction that has both a 3dnow version and 
> a mmxext version will be equally fast on both. 3dnow could be faster only if 
> you're using it for float on a cpu that doesn't have sse1.
> 13:23 <@pengvado> there is some 3dnow float (e.g. fft), but float functions 
> don't have mmx ext versions.
> 13:24 <@DonDiego> pengvado: ok, thx

OKed by Justin on IRC.

Diego
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to