Commit bdebf4ce authored by Vadim Pisarevsky's avatar Vadim Pisarevsky

Merge pull request #3637 from wangyan42164:LBP_mad24_opt

parents f3311f9e df697f65
...@@ -397,8 +397,8 @@ __kernel void runLBPClassifierStumpSimple( ...@@ -397,8 +397,8 @@ __kernel void runLBPClassifierStumpSimple(
for( tileIdx = groupIdx; tileIdx < totalTiles; tileIdx += ngroups ) for( tileIdx = groupIdx; tileIdx < totalTiles; tileIdx += ngroups )
{ {
int iy = ((tileIdx / ntiles.x)*local_size_y + ly)*ystep; int iy = mad24((tileIdx / ntiles.x), local_size_y, ly) * ystep;
int ix = ((tileIdx % ntiles.x)*local_size_x + lx)*ystep; int ix = mad24((tileIdx % ntiles.x), local_size_x, lx) * ystep;
if( ix < worksize.x && iy < worksize.y ) if( ix < worksize.x && iy < worksize.y )
{ {
......
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