Subject: Re: [linux-audio-dev] real-life 3dnow! results
From: est_AT_hyperreal.org
Date: la elo 28 1999 - 02:18:16 EDT
est_AT_hyperreal.org discourseth:
>
> I looked at Intel's online mmx manual and found an instruction that
> cleans up my code a lot. It doesn't seem to speed things up much, but
> it *does* add clipping. It also makes other code optimizations easier.
Such as gaining a further 1.5- to 2-fold speedup by morphing the inner
loop into:
myloop:
movq (%eax),%mm0
pfmul %mm2,%mm0
movq 8(%eax),%mm3
pf2id %mm0,%mm1
pfmul %mm2,%mm3
packssdw %mm1,%mm1
pf2id %mm3,%mm4
movd %mm1,(%ebp)
packssdw %mm4,%mm4
addl $16,%eax
movd %mm4,4(%ebp)
addl $8,%ebp
loop myloop
Further streamlining is possible..and scary. :)
Eric
This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:25:53 EST