> On Apr 3, 2015, at 15:18, Sergey Bylokhov <sergey.bylok...@oracle.com> wrote: > > 02.04.15 12:19, Hendrik Schreiber wrote: >> Sergey, do I understand this correctly, when I say, it seems that using the >> "wrong" image type there is a two time penalty, but once the image is cached >> (which happens, once we drew it), it will draw as quickly as any other >> VRAM-cached image? In other words: the gain of adjusting PNGDecoder to >> produce pre-multiplied images is probably minimal? > Yes, you are right.
Alright. Sounds like I overestimated the advantages pre-multiplied images would provide and it's smarter to leave everything the way it is. -hendrik
