[Ffmpeg-cvslog] CVS: ffmpeg/libavcodec dsputil.c, 1.136, 1.137 dsputil.h, 1.130, 1.131 h264.c, 1.199, 1.200

[Loren Merritt CVS]

Update of /cvsroot/ffmpeg/ffmpeg/libavcodec
In directory mail:/var2/tmp/cvs-serv31939

Modified Files:
	dsputil.c dsputil.h h264.c 
Log Message:
prefetch pixels for future motion compensation. 2-5% faster h264.



Index: dsputil.c
===================================================================
RCS file: /cvsroot/ffmpeg/ffmpeg/libavcodec/dsputil.c,v
retrieving revision 1.136
retrieving revision 1.137
diff -u -d -r1.136 -r1.137
--- dsputil.c	17 Mar 2006 12:08:46 -0000	1.136
+++ dsputil.c	23 Mar 2006 20:16:35 -0000	1.137
@@ -3773,6 +3773,8 @@
     dest[0] = cm[dest[0] + ((block[0] + 4)>>3)];
 }
 
+static void just_return() { return; }
+
 /* init static data */
 void dsputil_static_init(void)
 {
@@ -4054,6 +4056,8 @@
     c->inner_add_yblock = ff_snow_inner_add_yblock;
 #endif
 
+    c->prefetch= just_return;
+
 #ifdef HAVE_MMX
     dsputil_init_mmx(c, avctx);
 #endif

Index: dsputil.h
===================================================================
RCS file: /cvsroot/ffmpeg/ffmpeg/libavcodec/dsputil.h,v
retrieving revision 1.130
retrieving revision 1.131
diff -u -d -r1.130 -r1.131
--- dsputil.h	16 Mar 2006 19:18:18 -0000	1.130
+++ dsputil.h	23 Mar 2006 20:16:36 -0000	1.131
@@ -343,6 +343,8 @@
     void (*vertical_compose97i)(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
     void (*horizontal_compose97i)(DWTELEM *b, int width);
     void (*inner_add_yblock)(uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
+
+    void (*prefetch)(void *mem, int stride, int h);
 } DSPContext;
 
 void dsputil_static_init(void);

Index: h264.c
===================================================================
RCS file: /cvsroot/ffmpeg/ffmpeg/libavcodec/h264.c,v
retrieving revision 1.199
retrieving revision 1.200
diff -u -d -r1.199 -r1.200
--- h264.c	22 Mar 2006 12:41:02 -0000	1.199
+++ h264.c	23 Mar 2006 20:16:36 -0000	1.200
@@ -2752,6 +2752,22 @@
                     x_offset, y_offset, qpix_put, chroma_put, qpix_avg, chroma_avg, list0, list1);
 }
 
+static inline void prefetch_motion(H264Context *h, int list){
+    /* fetch pixels for estimated mv 4 macroblocks ahead
+     * optimized for 64byte cache lines */
+    MpegEncContext * const s = &h->s;
+    const int refn = h->ref_cache[list][scan8[0]];
+    if(refn >= 0){
+        const int mx= (h->mv_cache[list][scan8[0]][0]>>2) + 16*s->mb_x + 8;
+        const int my= (h->mv_cache[list][scan8[0]][1]>>2) + 16*s->mb_y;
+        uint8_t **src= h->ref_list[list][refn].data;
+        int off= mx + (my + (s->mb_x&3)*4)*s->linesize + 64;
+        s->dsp.prefetch(src[0]+off, s->linesize, 4);
+        off= (mx>>1) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + 64;
+        s->dsp.prefetch(src[1]+off, src[2]-src[1], 2);
+    }
+}
+
 static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                       qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
                       qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
@@ -2762,6 +2778,8 @@
 
     assert(IS_INTER(mb_type));
 
+    prefetch_motion(h, 0);
+
     if(IS_16X16(mb_type)){
         mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
                 qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
@@ -2833,6 +2851,8 @@
             }
         }
     }
+
+    prefetch_motion(h, 1);
 }
 
 static void decode_init_vlc(H264Context *h){