vp3/5: move put_no_rnd_pixels_l2 from dsputil to VP3DSPContext.

The function is only used in VP3 and VP5, so no need to have it in
DSPContext.

7 files changed, 39 insertions, 11 deletions

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 251fb0c2d9..ea40d2e8ed 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2765,7 +2765,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->clear_blocks                  = FUNCC(clear_blocks ## dct   , depth);\
     c->add_pixels8                   = FUNCC(add_pixels8  ## dct   , depth);\
     c->add_pixels4                   = FUNCC(add_pixels4  ## dct   , depth);\
-    c->put_no_rnd_pixels_l2          = FUNCC(put_no_rnd_pixels8_l2 , depth);\
 \
     c->put_h264_chroma_pixels_tab[0] = FUNCC(put_h264_chroma_mc8   , depth);\
     c->put_h264_chroma_pixels_tab[1] = FUNCC(put_h264_chroma_mc4   , depth);\
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index c0645a1eba..ddbbbe7539 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -291,8 +291,6 @@ typedef struct DSPContext {
      */
     op_pixels_func avg_no_rnd_pixels_tab[4][4];
 
-    void (*put_no_rnd_pixels_l2)(uint8_t *block/*align 8*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
-
     /**
      * Thirdpel motion compensation with rounding (a+b+1)>>1.
      * this is an array[12] of motion compensation functions for the 9 thirdpe
diff --git a/libavcodec/dsputil_template.c b/libavcodec/dsputil_template.c
index b9d5e97bc0..bd5c48b1a8 100644
--- a/libavcodec/dsputil_template.c
+++ b/libavcodec/dsputil_template.c
@@ -582,10 +582,6 @@ PIXOP2(put, op_put)
 #define put_no_rnd_pixels8_c  put_pixels8_c
 #define put_no_rnd_pixels16_c put_pixels16_c
 
-static void FUNCC(put_no_rnd_pixels8_l2)(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){
-    FUNC(put_no_rnd_pixels8_l2)(dst, a, b, stride, stride, stride, h);
-}
-
 #define H264_CHROMA_MC(OPNAME, OP)\
 static void FUNCC(OPNAME ## h264_chroma_mc2)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\
     pixel *dst = (pixel*)_dst;\
diff --git a/libavcodec/vp3.c b/libavcodec/vp3.c
index 58db8908db..33cfc8c204 100644
--- a/libavcodec/vp3.c
+++ b/libavcodec/vp3.c
@@ -1564,7 +1564,7 @@ static void render_slice(Vp3DecodeContext *s, int slice)
                                 motion_source, stride, 8);
                         }else{
                             int d= (motion_x ^ motion_y)>>31; // d is 0 if motion_x and _y have the same sign, else -1
-                            s->dsp.put_no_rnd_pixels_l2(
+                            s->vp3dsp.put_no_rnd_pixels_l2(
                                 output_plane + first_pixel,
                                 motion_source - d,
                                 motion_source + stride + 1 + d,
diff --git a/libavcodec/vp3dsp.c b/libavcodec/vp3dsp.c
index 9e6209dfdd..1883099002 100644
--- a/libavcodec/vp3dsp.c
+++ b/libavcodec/vp3dsp.c
@@ -274,8 +274,27 @@ static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
     }
 }
 
+static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
+                                 const uint8_t *src2, ptrdiff_t stride, int h)
+{
+    int i;
+
+    for (i = 0; i < h; i++) {
+        uint32_t a, b;
+
+        a = AV_RN32A(&src1[i * stride]);
+        b = AV_RN32A(&src2[i * stride]);
+        AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
+        a = AV_RN32A(&src1[i * stride + 4]);
+        b = AV_RN32A(&src2[i * stride + 4]);
+        AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
+    }
+}
+
 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
 {
+    c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
+
     c->idct_put      = vp3_idct_put_c;
     c->idct_add      = vp3_idct_add_c;
     c->idct_dc_add   = vp3_idct_dc_add_c;
diff --git a/libavcodec/vp3dsp.h b/libavcodec/vp3dsp.h
index feb300017a..d28c8478f0 100644
--- a/libavcodec/vp3dsp.h
+++ b/libavcodec/vp3dsp.h
@@ -19,10 +19,26 @@
 #ifndef AVCODEC_VP3DSP_H
 #define AVCODEC_VP3DSP_H
 
+#include <stddef.h>
 #include <stdint.h>
 #include "dsputil.h"
 
 typedef struct VP3DSPContext {
+    /**
+     * Copy 8xH pixels from source to destination buffer using a bilinear
+     * filter with no rounding (i.e. *dst = (*a + *b) >> 1).
+     *
+     * @param dst destination buffer, aligned by 8
+     * @param a first source buffer, no alignment
+     * @param b second source buffer, no alignment
+     * @param stride distance between two lines in source/dest buffers
+     * @param h height
+     */
+    void (*put_no_rnd_pixels_l2)(uint8_t *dst,
+                                 const uint8_t *a,
+                                 const uint8_t *b,
+                                 ptrdiff_t stride, int h);
+
     void (*idct_put)(uint8_t *dest, int line_size, DCTELEM *block);
     void (*idct_add)(uint8_t *dest, int line_size, DCTELEM *block);
     void (*idct_dc_add)(uint8_t *dest, int line_size, DCTELEM *block);
diff --git a/libavcodec/vp56.c b/libavcodec/vp56.c
index c05d38ea10..64210881a4 100644
--- a/libavcodec/vp56.c
+++ b/libavcodec/vp56.c
@@ -373,9 +373,9 @@ static void vp56_mc(VP56Context *s, int b, int plane, uint8_t *src,
             s->filter(s, dst, src_block, src_offset, src_offset+overlap_offset,
                       stride, s->mv[b], mask, s->filter_selection, b<4);
         else
-            s->dsp.put_no_rnd_pixels_l2(dst, src_block+src_offset,
-                                        src_block+src_offset+overlap_offset,
-                                        stride, 8);
+            s->vp3dsp.put_no_rnd_pixels_l2(dst, src_block+src_offset,
+                                           src_block+src_offset+overlap_offset,
+                                           stride, 8);
     } else {
         s->dsp.put_pixels_tab[1][0](dst, src_block+src_offset, stride, 8);
     }