Implement a SIMD version of emulated_edge_mc() for x86.

[ffmpeg] / libavcodec / snow.c

diff --git a/libavcodec/snow.c b/libavcodec/snow.c
index a4758b69ebb0a4761ce54b8c854dca436208524e..9ab9fed71f22955a869d66f57d55824e4b011fbc 100644 (file)
--- a/libavcodec/snow.c
+++ b/libavcodec/snow.c
@@ -920,7 +920,7 @@ static void decode_blocks(SnowContext *s){
     }
 }
 
-static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, uint8_t *tmp, int stride, int b_w, int b_h, int dx, int dy){
+static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int b_w, int b_h, int dx, int dy){
     static const uint8_t weight[64]={
     8,7,6,5,4,3,2,1,
     7,7,0,0,0,0,0,1,
@@ -1106,9 +1106,8 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, uint8_t *tmp, i
 
 #define mca(dx,dy,b_w)\
 static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, const uint8_t *src, int stride, int h){\
-    uint8_t tmp[stride*(b_w+HTAPS_MAX-1)];\
     assert(h==b_w);\
-    mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, tmp, stride, b_w, b_w, dx, dy);\
+    mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, stride, b_w, b_w, dx, dy);\
 }
 
 mca( 0, 0,16)
@@ -1172,7 +1171,7 @@ static void pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride, i
         src += sx + sy*stride;
         if(   (unsigned)sx >= w - b_w - (HTAPS_MAX-2)
            || (unsigned)sy >= h - b_h - (HTAPS_MAX-2)){
-            ff_emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+HTAPS_MAX-1, b_h+HTAPS_MAX-1, sx, sy, w, h);
+            s->dsp.emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+HTAPS_MAX-1, b_h+HTAPS_MAX-1, sx, sy, w, h);
             src= tmp + MB_SIZE;
         }
 //        assert(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h);
@@ -1180,7 +1179,7 @@ static void pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride, i
         assert(b_w>1 && b_h>1);
         assert((tab_index>=0 && tab_index<4) || b_w==32);
         if((dx&3) || (dy&3) || !(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h) || (b_w&(b_w-1)) || !s->plane[plane_index].fast_mc )
-            mc_block(&s->plane[plane_index], dst, src, tmp, stride, b_w, b_h, dx, dy);
+            mc_block(&s->plane[plane_index], dst, src, stride, b_w, b_h, dx, dy);
         else if(b_w==32){
             int y;
             for(y=0; y<b_h; y+=16){
@@ -2081,8 +2080,8 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPac
     common_init_after_header(avctx);
 
     // realloc slice buffer for the case that spatial_decomposition_count changed
-    slice_buffer_destroy(&s->sb);
-    slice_buffer_init(&s->sb, s->plane[0].height, (MB_SIZE >> s->block_max_depth) + s->spatial_decomposition_count * 8 + 1, s->plane[0].width, s->spatial_idwt_buffer);
+    ff_slice_buffer_destroy(&s->sb);
+    ff_slice_buffer_init(&s->sb, s->plane[0].height, (MB_SIZE >> s->block_max_depth) + s->spatial_decomposition_count * 8 + 1, s->plane[0].width, s->spatial_idwt_buffer);
 
     for(plane_index=0; plane_index<3; plane_index++){
         Plane *p= &s->plane[plane_index];
@@ -2199,10 +2198,10 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPac
             y = FFMIN(p->height, slice_starty);
             end_y = FFMIN(p->height, slice_h);
             while(y < end_y)
-                slice_buffer_release(&s->sb, y++);
+                ff_slice_buffer_release(&s->sb, y++);
         }
 
-        slice_buffer_flush(&s->sb);
+        ff_slice_buffer_flush(&s->sb);
         }
 
     }
@@ -2228,16 +2227,16 @@ static av_cold int decode_end(AVCodecContext *avctx)
 {
     SnowContext *s = avctx->priv_data;
 
-    slice_buffer_destroy(&s->sb);
+    ff_slice_buffer_destroy(&s->sb);
 
     common_end(s);
 
     return 0;
 }
 
-AVCodec snow_decoder = {
+AVCodec ff_snow_decoder = {
     "snow",
-    CODEC_TYPE_VIDEO,
+    AVMEDIA_TYPE_VIDEO,
     CODEC_ID_SNOW,
     sizeof(SnowContext),
     decode_init,
@@ -2829,9 +2828,9 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, con
          * to improve the score of the whole frame, thus iterative motion
          * estimation does not always converge. */
         if(s->avctx->me_cmp == FF_CMP_W97)
-            distortion = w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
+            distortion = ff_w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
         else if(s->avctx->me_cmp == FF_CMP_W53)
-            distortion = w53_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
+            distortion = ff_w53_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
         else{
             distortion = 0;
             for(i=0; i<4; i++){
@@ -3986,9 +3985,9 @@ static av_cold int encode_end(AVCodecContext *avctx)
     return 0;
 }
 
-AVCodec snow_encoder = {
+AVCodec ff_snow_encoder = {
     "snow",
-    CODEC_TYPE_VIDEO,
+    AVMEDIA_TYPE_VIDEO,
     CODEC_ID_SNOW,
     sizeof(SnowContext),
     encode_init,