[FFmpeg-devel] [PATCH] motion_est_mmx: prefer xmm registers below xmm6 when they are available

Ronald S. Bultje rsbultje
Sun Oct 31 12:59:02 CET 2010


Hi,

On Sat, Oct 30, 2010 at 3:41 PM, Ramiro Polla <ramiro.polla at gmail.com> wrote:
> $subj
[..]
--- a/libavcodec/x86/motion_est_mmx.c
+++ b/libavcodec/x86/motion_est_mmx.c
> @@ -111,9 +111,9 @@ static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)
>          : "r" ((x86_reg)stride)
>      );
>      __asm__ volatile(
> -        "movhlps %%xmm6, %%xmm0         \n\t"
> -        "paddw   %%xmm0, %%xmm6         \n\t"
> -        "movd    %%xmm6, %0             \n\t"
> +        "movhlps %%xmm2, %%xmm0         \n\t"
> +        "paddw   %%xmm0, %%xmm2         \n\t"
> +        "movd    %%xmm2, %0             \n\t"
>          : "=r"(ret)
>      );
>      return ret;

The blocks should probably be merged.

Ronald



More information about the ffmpeg-devel mailing list