Ángel Guzmán Maeso wrote:
A small change:
+int x,y; +for (x = 0; x < widthSrc; x++, tmpBits += 4)
- for(y=0;y<4;y++)
tmpBits[y] = tmpBits[y] * blendfn.SourceConstantAlpha / 255;
I'm not fussed one way or the other, but I can see at least two advantages in keeping the original code:
1) It's slightly more obvious that the data we are dealing with is four bytes of R, G, B and A - although a comment to that effect would be a better solution.
2) Compiler optimisers may do a slightly better job on the already unrolled loop.
I'll leave it to Alexandre and co. to pick and choose which version makes it in.