Hi, so i just sent out a patch series to mesa-devel to add XRGB2101010/ARGB2101010 fbconfig/visuals/rendering support to the Intel i965 driver. The patches here try to fix observed piglit failures with the piglit "gpu" subset of tests, as far as i think they are false positives.
Patch 1 and 2 fix false positives when running under a depth 30 x-screen or with RGBA 10 10 10 2/0 visuals exposed under glx/egl. About patch 3 and 4 i'm not sure if adding them makes sense? Without patch 3, running with color depth 30 / 10 bpc, causes piglit to skip a couple of tests. Adding the patch will make it more accepting, but then many tests fail because they assume the alpha channel has at least 8 bits, an assumption that is violated in 10 bpc rendering with only 2 bits of alpha. Patch 4 fixes some of those false failures, but not all of them. I'm also not sure if overriding the alpha bits inside piglit_set_tolerance_for_bits() is really the right approach for all tests. Maybe just skipping sets of tests for the depth 30 rendering is more sensible? I don't have commit access, so if some of these get accepted, somebody else has to apply them. Thanks, -mario _______________________________________________ Piglit mailing list Piglit@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/piglit
