[MPlayer-dev-eng] [PATCH]: af_resample SIMD optimization (MMX/SSE2/3DNOW!/SSE/SSE3)

Zhou Zongyi zhouzongyi at pset.suntec.net
Wed Apr 1 12:37:36 CEST 2009


Hi,

Even if s->w is 16-byte aligned, &x[xi] and &w[wi*L] is not. Thus aligned SSE instruction still crashes.

During the "UP" phase x is loaded outside the inner-loop, so multiplication from memory is not necesary.

Zhou Zongyi

On Wed, 1 Apr 2009 13:35:15 +0800
Zuxy Meng <zuxy.meng at gmail.com> wrote:

> Hi,
> 
> Weights are malloc()'ed in s->w at af_resample.c:control(). You can
> force a 16-byte alignment there. memalign() isn't available on all
> platforms and you can use av_malloc() instead.
> 
> -- 
> Zuxy
> Beauty is truth,
> While truth is beauty.
> PGP KeyID: E8555ED6



More information about the MPlayer-dev-eng mailing list