comparison i386/dsputil_mmx.c @ 296:c1a8a1b4a24b libavcodec

sizeof(s->block) isnt 64*6*2 anymore bugfix mpeg12 decoding optimization

295:6622b0fd036c

   pix += line_size;
   p +=   8 ;
  } while(--h);
}

static void just_return() { return; }

void dsputil_init_mmx(void)
{
    mm_flags = mm_support();

    if (mm_flags & MM_MMX) {
        get_pixels = get_pixels_mmx;
        put_pixels_clamped = put_pixels_clamped_mmx;
        add_pixels_clamped = add_pixels_clamped_mmx;
        
        pix_abs16x16     = pix_abs16x16_mmx;
        pix_abs16x16_x2  = pix_abs16x16_x2_mmx;
        pix_abs16x16_y2  = pix_abs16x16_y2_mmx;
        pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx;
        pix_abs8x8    = pix_abs8x8_mmx;

296:c1a8a1b4a24b

   pix += line_size;
   p +=   8 ;
  } while(--h);
}

static void clear_blocks_mmx(DCTELEM *blocks)
{
        asm volatile(
                "pxor %%mm7, %%mm7		\n\t"
                "movl $-128*6, %%eax		\n\t"
                "1:				\n\t"
                "movq %%mm7, (%0, %%eax)	\n\t"
                "movq %%mm7, 8(%0, %%eax)	\n\t"
                "movq %%mm7, 16(%0, %%eax)	\n\t"
                "movq %%mm7, 24(%0, %%eax)	\n\t"
                "addl $32, %%eax		\n\t"
                " js 1b				\n\t"
                : : "r" (((int)blocks)+128*6)
                : "%eax"
        );
}

static void just_return() { return; }

void dsputil_init_mmx(void)
{
    mm_flags = mm_support();

    if (mm_flags & MM_MMX) {
        get_pixels = get_pixels_mmx;
        put_pixels_clamped = put_pixels_clamped_mmx;
        add_pixels_clamped = add_pixels_clamped_mmx;
        clear_blocks= clear_blocks_mmx;
       
        pix_abs16x16     = pix_abs16x16_mmx;
        pix_abs16x16_x2  = pix_abs16x16_x2_mmx;
        pix_abs16x16_y2  = pix_abs16x16_y2_mmx;
        pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx;
        pix_abs8x8    = pix_abs8x8_mmx;