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, 17 insertions, 3 deletions

diff --git a/libavcodec/rv40dsp.c b/libavcodec/rv40dsp.c
index c12958a89c..19a18d37a5 100644
--- a/libavcodec/rv40dsp.c
+++ b/libavcodec/rv40dsp.c
@@ -278,7 +278,7 @@ RV40_CHROMA_MC(put_, op_put)
 RV40_CHROMA_MC(avg_, op_avg)
 
 #define RV40_WEIGHT_FUNC(size) \
-static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
+static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
 {\
     int i, j;\
 \
@@ -289,6 +289,18 @@ static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src
         src2 += stride;\
         dst  += stride;\
     }\
+}\
+static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
+{\
+    int i, j;\
+\
+    for (j = 0; j < size; j++) {\
+        for (i = 0; i < size; i++)\
+            dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\
+        src1 += stride;\
+        src2 += stride;\
+        dst  += stride;\
+    }\
 }
 
 RV40_WEIGHT_FUNC(16)
@@ -578,8 +590,10 @@ av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
     c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
     c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
 
-    c->rv40_weight_pixels_tab[0] = rv40_weight_func_16;
-    c->rv40_weight_pixels_tab[1] = rv40_weight_func_8;
+    c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16;
+    c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8;
+    c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16;
+    c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8;
 
     c->rv40_weak_loop_filter[0]     = rv40_h_weak_loop_filter;
     c->rv40_weak_loop_filter[1]     = rv40_v_weak_loop_filter;