Statements like a = b = c = d = e; store from right-to-left, so if
authorAlexander Strange <astrange@ithinksw.com>
Tue, 14 Aug 2007 22:28:09 +0000 (22:28 +0000)
committerGuillaume Poirier <gpoirier@mplayerhq.hu>
Tue, 14 Aug 2007 22:28:09 +0000 (22:28 +0000)
you write them in the right order it comes out backwards.
This removes them from fill_rectangle().
patch by Alexander Strange %astrange A ithinksw P com%
Original thread:
Date: Aug 14, 2007 5:36 AM
Subject: [FFmpeg-devel] [PATCH] two small h264 optimizations

Originally committed as revision 10118 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/h264.c

index 8f20bad..676ce51 100644 (file)
@@ -91,7 +91,7 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
         if(h==1) return;
         *(uint16_t*)(p + 1*stride)= v;
         if(h==2) return;
         if(h==1) return;
         *(uint16_t*)(p + 1*stride)= v;
         if(h==2) return;
-        *(uint16_t*)(p + 2*stride)=
+        *(uint16_t*)(p + 2*stride)= v;
         *(uint16_t*)(p + 3*stride)= v;
     }else if(w==4){
         const uint32_t v= size==4 ? val : val*0x01010101;
         *(uint16_t*)(p + 3*stride)= v;
     }else if(w==4){
         const uint32_t v= size==4 ? val : val*0x01010101;
@@ -99,7 +99,7 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
         if(h==1) return;
         *(uint32_t*)(p + 1*stride)= v;
         if(h==2) return;
         if(h==1) return;
         *(uint32_t*)(p + 1*stride)= v;
         if(h==2) return;
-        *(uint32_t*)(p + 2*stride)=
+        *(uint32_t*)(p + 2*stride)= v;
         *(uint32_t*)(p + 3*stride)= v;
     }else if(w==8){
     //gcc can't optimize 64bit math on x86_32
         *(uint32_t*)(p + 3*stride)= v;
     }else if(w==8){
     //gcc can't optimize 64bit math on x86_32
@@ -109,47 +109,47 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
         if(h==1) return;
         *(uint64_t*)(p + 1*stride)= v;
         if(h==2) return;
         if(h==1) return;
         *(uint64_t*)(p + 1*stride)= v;
         if(h==2) return;
-        *(uint64_t*)(p + 2*stride)=
+        *(uint64_t*)(p + 2*stride)= v;
         *(uint64_t*)(p + 3*stride)= v;
     }else if(w==16){
         const uint64_t v= val*0x0100000001ULL;
         *(uint64_t*)(p + 3*stride)= v;
     }else if(w==16){
         const uint64_t v= val*0x0100000001ULL;
-        *(uint64_t*)(p + 0+0*stride)=
-        *(uint64_t*)(p + 8+0*stride)=
-        *(uint64_t*)(p + 0+1*stride)=
+        *(uint64_t*)(p + 0+0*stride)= v;
+        *(uint64_t*)(p + 8+0*stride)= v;
+        *(uint64_t*)(p + 0+1*stride)= v;
         *(uint64_t*)(p + 8+1*stride)= v;
         if(h==2) return;
         *(uint64_t*)(p + 8+1*stride)= v;
         if(h==2) return;
-        *(uint64_t*)(p + 0+2*stride)=
-        *(uint64_t*)(p + 8+2*stride)=
-        *(uint64_t*)(p + 0+3*stride)=
+        *(uint64_t*)(p + 0+2*stride)= v;
+        *(uint64_t*)(p + 8+2*stride)= v;
+        *(uint64_t*)(p + 0+3*stride)= v;
         *(uint64_t*)(p + 8+3*stride)= v;
 #else
         *(uint64_t*)(p + 8+3*stride)= v;
 #else
-        *(uint32_t*)(p + 0+0*stride)=
+        *(uint32_t*)(p + 0+0*stride)= val;
         *(uint32_t*)(p + 4+0*stride)= val;
         if(h==1) return;
         *(uint32_t*)(p + 4+0*stride)= val;
         if(h==1) return;
-        *(uint32_t*)(p + 0+1*stride)=
+        *(uint32_t*)(p + 0+1*stride)= val;
         *(uint32_t*)(p + 4+1*stride)= val;
         if(h==2) return;
         *(uint32_t*)(p + 4+1*stride)= val;
         if(h==2) return;
-        *(uint32_t*)(p + 0+2*stride)=
-        *(uint32_t*)(p + 4+2*stride)=
-        *(uint32_t*)(p + 0+3*stride)=
+        *(uint32_t*)(p + 0+2*stride)= val;
+        *(uint32_t*)(p + 4+2*stride)= val;
+        *(uint32_t*)(p + 0+3*stride)= val;
         *(uint32_t*)(p + 4+3*stride)= val;
     }else if(w==16){
         *(uint32_t*)(p + 4+3*stride)= val;
     }else if(w==16){
-        *(uint32_t*)(p + 0+0*stride)=
-        *(uint32_t*)(p + 4+0*stride)=
-        *(uint32_t*)(p + 8+0*stride)=
-        *(uint32_t*)(p +12+0*stride)=
-        *(uint32_t*)(p + 0+1*stride)=
-        *(uint32_t*)(p + 4+1*stride)=
-        *(uint32_t*)(p + 8+1*stride)=
+        *(uint32_t*)(p + 0+0*stride)= val;
+        *(uint32_t*)(p + 4+0*stride)= val;
+        *(uint32_t*)(p + 8+0*stride)= val;
+        *(uint32_t*)(p +12+0*stride)= val;
+        *(uint32_t*)(p + 0+1*stride)= val;
+        *(uint32_t*)(p + 4+1*stride)= val;
+        *(uint32_t*)(p + 8+1*stride)= val;
         *(uint32_t*)(p +12+1*stride)= val;
         if(h==2) return;
         *(uint32_t*)(p +12+1*stride)= val;
         if(h==2) return;
-        *(uint32_t*)(p + 0+2*stride)=
-        *(uint32_t*)(p + 4+2*stride)=
-        *(uint32_t*)(p + 8+2*stride)=
-        *(uint32_t*)(p +12+2*stride)=
-        *(uint32_t*)(p + 0+3*stride)=
-        *(uint32_t*)(p + 4+3*stride)=
-        *(uint32_t*)(p + 8+3*stride)=
+        *(uint32_t*)(p + 0+2*stride)= val;
+        *(uint32_t*)(p + 4+2*stride)= val;
+        *(uint32_t*)(p + 8+2*stride)= val;
+        *(uint32_t*)(p +12+2*stride)= val;
+        *(uint32_t*)(p + 0+3*stride)= val;
+        *(uint32_t*)(p + 4+3*stride)= val;
+        *(uint32_t*)(p + 8+3*stride)= val;
         *(uint32_t*)(p +12+3*stride)= val;
 #endif
     }else
         *(uint32_t*)(p +12+3*stride)= val;
 #endif
     }else