Am Friday 14 August 2009 12:13:01 schrieb Henri Verbeet:
2009/8/14 Stefan Dösinger stefandoesinger@gmx.at:
Am Friday 14 August 2009 09:23:25 schrieb Henri Verbeet:
For most cards this should make more sense than RTL_READDRAW, even if e.g. surface_upload_data() has some room for improvement.
I don't think this is a good idea until the READDRAW code has the ability to only upload a part of the surface if a limited rectangle was locked
You probably mean RTL_READTEX, surface_upload_data() specifically, but I did mention that in the commit message.
I should not write emails directly after getting up... READTEX was what I meant.
What I am also wondering about this patch: Its general wisdom that texture load + draw a quad is faster than glDrawPixels, although glDrawPixels is the more natural match to framebuffer locking. Why aren't drivers implementing glDrawPixels with texture+draw in that case? Or does glDrawPixels do anything we don't need that slows it down compared to the texture codepath?