I just wanted to point out my cool observation i had when watching the friday stream.
if you save to memory InverseAlpha:=(1-alpha) instead of alpha, you can get 2 things:
1. all of your alpha usage in the color section is done by InverseAlpha because of the premultiplied alpha, so it makes things nicer to read i think.
2. by following the mathematical derivetion casey did, the alpha section has the cool feature that (1-Abuffer)=(1-Asource)*(1-Adest),or:
InverseAlphaBuffer=InverseAlphaSource*InverseAlphaDest and this is really nicer to see.
i know it does no good to try to optimize things now, but i think that it's cool enough to write on the forums.
if you save to memory InverseAlpha:=(1-alpha) instead of alpha, you can get 2 things:
1. all of your alpha usage in the color section is done by InverseAlpha because of the premultiplied alpha, so it makes things nicer to read i think.
2. by following the mathematical derivetion casey did, the alpha section has the cool feature that (1-Abuffer)=(1-Asource)*(1-Adest),or:
InverseAlphaBuffer=InverseAlphaSource*InverseAlphaDest and this is really nicer to see.
i know it does no good to try to optimize things now, but i think that it's cool enough to write on the forums.
That doesn't really save memory, but it's a good observation. If alpha is the "coverage" of a pixel, then 1-alpha is the "stuff that is let through".
However, it doesn't realistically save any work, especially given that the world has standardised on alpha, not 1-alpha. I'm also not of the opinion that just because something is a little shorter, that makes it nicer to read. It's possible to make code so concise that the intent is lost.
However, it doesn't realistically save any work, especially given that the world has standardised on alpha, not 1-alpha. I'm also not of the opinion that just because something is a little shorter, that makes it nicer to read. It's possible to make code so concise that the intent is lost.