rv40dsp: implement prescaled versions for biweight.

Quite often, the original weights are multiple of 512. By prescaling them
by 1/512 when they are computed (once per frame), no intermediate shifting
is needed, and no prescaling on each call either.

The x86 code already used that trick.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

1 files changed, 20 insertions, 10 deletions

diff --git a/libavcodec/x86/rv40dsp_init.c b/libavcodec/x86/rv40dsp_init.c
index 79c70f78c3..df468aa9e5 100644
--- a/libavcodec/x86/rv40dsp_init.c
+++ b/libavcodec/x86/rv40dsp_init.c
@@ -41,10 +41,14 @@ void ff_avg_rv40_chroma_mc4_3dnow(uint8_t *dst, uint8_t *src,
                                   int stride, int h, int x, int y);
 
 #define DECLARE_WEIGHT(opt) \
-void ff_rv40_weight_func_16_##opt(uint8_t *dst, uint8_t *src1, uint8_t *src2, \
-                                  int w1, int w2, ptrdiff_t stride); \
-void ff_rv40_weight_func_8_##opt (uint8_t *dst, uint8_t *src1, uint8_t *src2, \
-                                  int w1, int w2, ptrdiff_t stride);
+void ff_rv40_weight_func_rnd_16_##opt(uint8_t *dst, uint8_t *src1, uint8_t *src2, \
+                                      int w1, int w2, ptrdiff_t stride); \
+void ff_rv40_weight_func_rnd_8_##opt (uint8_t *dst, uint8_t *src1, uint8_t *src2, \
+                                      int w1, int w2, ptrdiff_t stride); \
+void ff_rv40_weight_func_nornd_16_##opt(uint8_t *dst, uint8_t *src1, uint8_t *src2, \
+                                        int w1, int w2, ptrdiff_t stride); \
+void ff_rv40_weight_func_nornd_8_##opt (uint8_t *dst, uint8_t *src1, uint8_t *src2, \
+                                        int w1, int w2, ptrdiff_t stride);
 DECLARE_WEIGHT(mmx)
 DECLARE_WEIGHT(sse2)
 DECLARE_WEIGHT(ssse3)
@@ -57,8 +61,10 @@ void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
     if (mm_flags & AV_CPU_FLAG_MMX) {
         c->put_chroma_pixels_tab[0] = ff_put_rv40_chroma_mc8_mmx;
         c->put_chroma_pixels_tab[1] = ff_put_rv40_chroma_mc4_mmx;
-        c->rv40_weight_pixels_tab[0] = ff_rv40_weight_func_16_mmx;
-        c->rv40_weight_pixels_tab[1] = ff_rv40_weight_func_8_mmx;
+        c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmx;
+        c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_mmx;
+        c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_mmx;
+        c->rv40_weight_pixels_tab[1][1] = ff_rv40_weight_func_nornd_8_mmx;
     }
     if (mm_flags & AV_CPU_FLAG_MMX2) {
         c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmx2;
@@ -68,12 +74,16 @@ void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
         c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_3dnow;
     }
     if (mm_flags & AV_CPU_FLAG_SSE2) {
-        c->rv40_weight_pixels_tab[0] = ff_rv40_weight_func_16_sse2;
-        c->rv40_weight_pixels_tab[1] = ff_rv40_weight_func_8_sse2;
+        c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_sse2;
+        c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_sse2;
+        c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_sse2;
+        c->rv40_weight_pixels_tab[1][1] = ff_rv40_weight_func_nornd_8_sse2;
     }
     if (mm_flags & AV_CPU_FLAG_SSSE3) {
-        c->rv40_weight_pixels_tab[0] = ff_rv40_weight_func_16_ssse3;
-        c->rv40_weight_pixels_tab[1] = ff_rv40_weight_func_8_ssse3;
+        c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_ssse3;
+        c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_ssse3;
+        c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_ssse3;
+        c->rv40_weight_pixels_tab[1][1] = ff_rv40_weight_func_nornd_8_ssse3;
     }
 #endif
 }