We definitely want to fall through to the unsynchronized map case, instead
of wasting bandwidth on a copy. Prevents a -43.2407% +/- 1.06113% (n=49)
performance regression on aa10perf when teaching glamor to provide the
GL_INVALIDATE_RANGE_BIT information.
This is a performance fix, which I usuall
https://bugs.freedesktop.org/show_bug.cgi?id=72889
--- Comment #5 from Michel Dänzer ---
I suspect the problem is that fast colour clears are not resolved during buffer
swaps on Wayland (linear buffers cannot use fast clears, which is why your
r600_choose_tiling change works around the problem) a