diff x86/dsputil_mmx.c @ 10430:12c8175d6db5 libavcodec

simd add_hfyu_left_prediction 2.2x faster than C on conroe, 3.6x on penryn. 4-6% faster huffyuv decoding if using left or plane mode and yuv

--- a/x86/dsputil_mmx.c	Sun Oct 18 19:51:18 2009 +0000
+++ b/x86/dsputil_mmx.c	Sun Oct 18 20:10:10 2009 +0000
@@ -2385,6 +2385,8 @@
 void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
 void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len);
 void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top);
+int  ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, uint8_t *src, int w, int left);
+int  ff_add_hfyu_left_prediction_sse4(uint8_t *dst, uint8_t *src, int w, int left);
 void ff_x264_deblock_v_luma_sse2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
 void ff_x264_deblock_h_luma_sse2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
 void ff_x264_deblock_v8_luma_intra_mmxext(uint8_t *pix, int stride, int alpha, int beta);
@@ -2951,6 +2953,11 @@
             c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_ssse3;
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_ssse3;
             c->add_png_paeth_prediction= add_png_paeth_prediction_ssse3;
+#if HAVE_YASM
+            c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_ssse3;
+            if (mm_flags & FF_MM_SSE4) // not really sse4, just slow on Conroe
+                c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4;
+#endif
         }
 #endif