store the number of runs to avoid storing the last run value

about 10% lower bitrate for -qscale 32 (forman & some music video)
worst case bitrate increase <0.1% (lossless or low qscale)
and now the bad news, even though this just adds a single subtraction and an if() into the medium sized unpack_coeffs() loop and the if() will only be false once per unpac_coeff() call, gcc produces 50% slower code, i didnt look at the generated asm yet, not sure if i want to ...

Originally committed as revision 4131 to svn://svn.ffmpeg.org/ffmpeg/trunk

3 files changed, 22 insertions, 13 deletions

diff --git a/libavcodec/snow.c b/libavcodec/snow.c
index ffabafbc65..7020b39ba0 100644
--- a/libavcodec/snow.c
+++ b/libavcodec/snow.c
@@ -1657,6 +1657,7 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE
         int run=0;
         int runs[w*h];
         int run_index=0;
+        int max_index;
                 
         for(y=0; y<h; y++){
             for(x=0; x<w; x++){
@@ -1696,11 +1697,14 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE
                 }
             }
         }
+        max_index= run_index;
         runs[run_index++]= run;
         run_index=0;
         run= runs[run_index++];
 
-        put_symbol2(&s->c, b->state[1], run, 3);
+        put_symbol2(&s->c, b->state[30], max_index, 0);
+        if(run_index <= max_index)
+            put_symbol2(&s->c, b->state[1], run, 3);
         
         for(y=0; y<h; y++){
             if(s->c.bytestream_end - s->c.bytestream < w*40){
@@ -1742,7 +1746,8 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE
                     if(!run){
                         run= runs[run_index++];
 
-                        put_symbol2(&s->c, b->state[1], run, 3);
+                        if(run_index <= max_index)
+                            put_symbol2(&s->c, b->state[1], run, 3);
                         assert(v);
                     }else{
                         run--;
@@ -1776,14 +1781,17 @@ static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, i
     int x,y;
     
     if(1){
-        int run;
+        int run, runs;
         x_and_coeff *xc= b->x_coeff;
         x_and_coeff *prev_xc= NULL;
         x_and_coeff *prev2_xc= xc;
         x_and_coeff *parent_xc= parent ? parent->x_coeff : NULL;
         x_and_coeff *prev_parent_xc= parent_xc;
 
-        run= get_symbol2(&s->c, b->state[1], 3);
+        runs= get_symbol2(&s->c, b->state[30], 0);
+        if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
+        else           run= INT_MAX;
+
         for(y=0; y<h; y++){
             int v=0;
             int lt=0, t=0, rt=0;
@@ -1826,7 +1834,8 @@ static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, i
                     }
                 }else{
                     if(!run){
-                        run= get_symbol2(&s->c, b->state[1], 3);
+                        if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
+                        else           run= INT_MAX;
                         v= 2*(get_symbol2(&s->c, b->state[0 + 2], 0-4) + 1);
                         v+=get_rac(&s->c, &b->state[0][16 + 1 + 3]);
                         
diff --git a/tests/ffmpeg.regression.ref b/tests/ffmpeg.regression.ref
index 81e4745804..1f920a97e6 100644
--- a/tests/ffmpeg.regression.ref
+++ b/tests/ffmpeg.regression.ref
@@ -119,12 +119,12 @@ a7ef4746f27be309138c188e327d3ebe *./data/a-ffv1.avi
 2653642 ./data/a-ffv1.avi
 799d3db687f6cdd7a837ec156efc171f *./data/out.yuv
 stddev:  0.00 PSNR:99.99 bytes:7602176
-98f6913d98bccb4ba4829ccde7286e8f *./data/a-snow.avi
-1266766 ./data/a-snow.avi
+24c1fcbcdb08052359679700056ef4e5 *./data/a-snow.avi
+1267526 ./data/a-snow.avi
 575fdd879119902a8289c825c2389dca *./data/out.yuv
 stddev:  2.96 PSNR:38.67 bytes:7602176
-28b6a82fdd8058e2df35778c9a5edbbf *./data/a-snow53.avi
-3537074 ./data/a-snow53.avi
+0c570833b736a8d4c8c428bf6e82272f *./data/a-snow53.avi
+3537490 ./data/a-snow53.avi
 799d3db687f6cdd7a837ec156efc171f *./data/out.yuv
 stddev:  0.00 PSNR:99.99 bytes:7602176
 b5b6275f58f012de73644bbaa9080097 *./data/a-svq1.mov
diff --git a/tests/rotozoom.regression.ref b/tests/rotozoom.regression.ref
index 078df0a0d7..540f4176fc 100644
--- a/tests/rotozoom.regression.ref
+++ b/tests/rotozoom.regression.ref
@@ -119,12 +119,12 @@ d0831a8339491fd680b650f05262e5d9 *./data/a-ffv1.avi
 3524768 ./data/a-ffv1.avi
 dde5895817ad9d219f79a52d0bdfb001 *./data/out.yuv
 stddev:  0.00 PSNR:99.99 bytes:7602176
-d1546d2176c4cd3913ccb0920e5e556f *./data/a-snow.avi
-326454 ./data/a-snow.avi
+2d665a41c46d8e581cf5898e85a2b55f *./data/a-snow.avi
+326532 ./data/a-snow.avi
 c0b1b63dc45141c76d538edfd618ac19 *./data/out.yuv
 stddev:  2.40 PSNR:40.51 bytes:7602176
-8b510a8e4a26d6d469bd80b46922d5e6 *./data/a-snow53.avi
-2724602 ./data/a-snow53.avi
+d0a8bdb25d7ed641eab0debcac396eb7 *./data/a-snow53.avi
+2725842 ./data/a-snow53.avi
 dde5895817ad9d219f79a52d0bdfb001 *./data/out.yuv
 stddev:  0.00 PSNR:99.99 bytes:7602176
 920c610ec324b772d882b0717e375943 *./data/a-svq1.mov