FTR, I rebased this series on top of current master (minimal changes) and run a full CTS and piglit run with the i965 [gen 8].
The failing CTS now passes and I have not observed any regression nor in CTS neither in piglit. On Tue, 2016-06-07 at 15:25 +1000, Dave Airlie wrote: > So when I implemented shader subroutines, I ignored part of the > spec, because nobody really cares or uses this extension. > > But since CTS has a test for this feature I thought I'd implement it > a bit better now (still not perfect). > > So the spec says that the values in the subroutine uniforms are > stored per context not per program. > > The first patch is enough to pass the CTS test which doesn't do > any rendering checks, it just writes values from multiple contexts > and reads them back. > > The second 4 tests, push the updating of the subroutines into > the constant upload paths of the drivers, so it calls the API > to update the subroutine constants just before it writes the > constants to the hw. This might not be perfect in some threaded > cases, but it's a lot better than nothing, and I don't think > this API will see much use going forward. > > With these and the previous one liner, there is only one > failing test in CTS for subroutines and it is due to program binary > API. > > Dave. > > _______________________________________________ > mesa-dev mailing list > mesa-dev@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/mesa-dev -- Br, Andres _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
