1 // It's actually (but surprisingly) not correct to do a mix() here;
2 // it would be if we had postmultiplied alpha and didn't have to worry
3 // about alpha in the bottom layer, but given that we use premultiplied
4 // alpha all over, top shouldn't actually be multiplied by anything.
6 // These formulas come from Wikipedia:
8 // http://en.wikipedia.org/wiki/Alpha_compositing
10 // We use the associative version given. However, note that since we want
11 // _output_ to be premultiplied, C_o from Wikipedia is not what we want,
12 // but rather c_o (which is not explicitly given, but obviously is just
13 // C_o without the division by alpha_o).
15 vec4 FUNCNAME(vec2 tc) {
16 // SWAP_INPUTS will be #defined to 1 if we want to swap the two inputs,
18 vec4 bottom = INPUT2(tc);
19 vec4 top = INPUT1(tc);
21 vec4 bottom = INPUT1(tc);
22 vec4 top = INPUT2(tc);
24 return top + (1.0 - top.a) * bottom;