Commit 1ad26d5c authored by gpoirier's avatar gpoirier

Statements like a = b = c = d = e; store from right-to-left, so if

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


git-svn-id: file:///var/local/repositories/ffmpeg/trunk@10118 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b
parent bd791378
...@@ -91,7 +91,7 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride, ...@@ -91,7 +91,7 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
if(h==1) return; if(h==1) return;
*(uint16_t*)(p + 1*stride)= v; *(uint16_t*)(p + 1*stride)= v;
if(h==2) return; if(h==2) return;
*(uint16_t*)(p + 2*stride)= *(uint16_t*)(p + 2*stride)= v;
*(uint16_t*)(p + 3*stride)= v; *(uint16_t*)(p + 3*stride)= v;
}else if(w==4){ }else if(w==4){
const uint32_t v= size==4 ? val : val*0x01010101; 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, ...@@ -99,7 +99,7 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
if(h==1) return; if(h==1) return;
*(uint32_t*)(p + 1*stride)= v; *(uint32_t*)(p + 1*stride)= v;
if(h==2) return; if(h==2) return;
*(uint32_t*)(p + 2*stride)= *(uint32_t*)(p + 2*stride)= v;
*(uint32_t*)(p + 3*stride)= v; *(uint32_t*)(p + 3*stride)= v;
}else if(w==8){ }else if(w==8){
//gcc can't optimize 64bit math on x86_32 //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, ...@@ -109,47 +109,47 @@ static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride,
if(h==1) return; if(h==1) return;
*(uint64_t*)(p + 1*stride)= v; *(uint64_t*)(p + 1*stride)= v;
if(h==2) return; if(h==2) return;
*(uint64_t*)(p + 2*stride)= *(uint64_t*)(p + 2*stride)= v;
*(uint64_t*)(p + 3*stride)= v; *(uint64_t*)(p + 3*stride)= v;
}else if(w==16){ }else if(w==16){
const uint64_t v= val*0x0100000001ULL; const uint64_t v= val*0x0100000001ULL;
*(uint64_t*)(p + 0+0*stride)= *(uint64_t*)(p + 0+0*stride)= v;
*(uint64_t*)(p + 8+0*stride)= *(uint64_t*)(p + 8+0*stride)= v;
*(uint64_t*)(p + 0+1*stride)= *(uint64_t*)(p + 0+1*stride)= v;
*(uint64_t*)(p + 8+1*stride)= v; *(uint64_t*)(p + 8+1*stride)= v;
if(h==2) return; if(h==2) return;
*(uint64_t*)(p + 0+2*stride)= *(uint64_t*)(p + 0+2*stride)= v;
*(uint64_t*)(p + 8+2*stride)= *(uint64_t*)(p + 8+2*stride)= v;
*(uint64_t*)(p + 0+3*stride)= *(uint64_t*)(p + 0+3*stride)= v;
*(uint64_t*)(p + 8+3*stride)= v; *(uint64_t*)(p + 8+3*stride)= v;
#else #else
*(uint32_t*)(p + 0+0*stride)= *(uint32_t*)(p + 0+0*stride)= val;
*(uint32_t*)(p + 4+0*stride)= val; *(uint32_t*)(p + 4+0*stride)= val;
if(h==1) return; if(h==1) return;
*(uint32_t*)(p + 0+1*stride)= *(uint32_t*)(p + 0+1*stride)= val;
*(uint32_t*)(p + 4+1*stride)= val; *(uint32_t*)(p + 4+1*stride)= val;
if(h==2) return; if(h==2) return;
*(uint32_t*)(p + 0+2*stride)= *(uint32_t*)(p + 0+2*stride)= val;
*(uint32_t*)(p + 4+2*stride)= *(uint32_t*)(p + 4+2*stride)= val;
*(uint32_t*)(p + 0+3*stride)= *(uint32_t*)(p + 0+3*stride)= val;
*(uint32_t*)(p + 4+3*stride)= val; *(uint32_t*)(p + 4+3*stride)= val;
}else if(w==16){ }else if(w==16){
*(uint32_t*)(p + 0+0*stride)= *(uint32_t*)(p + 0+0*stride)= val;
*(uint32_t*)(p + 4+0*stride)= *(uint32_t*)(p + 4+0*stride)= val;
*(uint32_t*)(p + 8+0*stride)= *(uint32_t*)(p + 8+0*stride)= val;
*(uint32_t*)(p +12+0*stride)= *(uint32_t*)(p +12+0*stride)= val;
*(uint32_t*)(p + 0+1*stride)= *(uint32_t*)(p + 0+1*stride)= val;
*(uint32_t*)(p + 4+1*stride)= *(uint32_t*)(p + 4+1*stride)= val;
*(uint32_t*)(p + 8+1*stride)= *(uint32_t*)(p + 8+1*stride)= val;
*(uint32_t*)(p +12+1*stride)= val; *(uint32_t*)(p +12+1*stride)= val;
if(h==2) return; if(h==2) return;
*(uint32_t*)(p + 0+2*stride)= *(uint32_t*)(p + 0+2*stride)= val;
*(uint32_t*)(p + 4+2*stride)= *(uint32_t*)(p + 4+2*stride)= val;
*(uint32_t*)(p + 8+2*stride)= *(uint32_t*)(p + 8+2*stride)= val;
*(uint32_t*)(p +12+2*stride)= *(uint32_t*)(p +12+2*stride)= val;
*(uint32_t*)(p + 0+3*stride)= *(uint32_t*)(p + 0+3*stride)= val;
*(uint32_t*)(p + 4+3*stride)= *(uint32_t*)(p + 4+3*stride)= val;
*(uint32_t*)(p + 8+3*stride)= *(uint32_t*)(p + 8+3*stride)= val;
*(uint32_t*)(p +12+3*stride)= val; *(uint32_t*)(p +12+3*stride)= val;
#endif #endif
}else }else
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment