Since OpenGL ES 2.0, I always discarded pixels in my pixel shaders on all platforms, but often with strong performance impact. When I asked someone in one of the Metal Labs at WWDC about it, I was told that discard in general is not a good idea and I should use alpha testing instead (not sure why though, because early depth should still not work with it). Now that I finally got around to look into it, it doesn't seem like there is such a thing as alpha testing in metal?
What I did find/am already using for some things is Alpha To Coverage, should I just use that instead? Can I expect it to be faster than discarding fragments? Are there any disadvantages?
