Skip to content

freedreno/turnip: Fix bugs around the maximum varying count

Connor Abbott requested to merge cwabbott0/mesa:review/tu-varying-count into master

The first patch fixes hangs in a few dEQP-VK.glsl.limits.near_max.fragment_input.components_* tests. The second patch unfortunately disables the tests that hanged, due to an off-by-one error in the dEQP I have so that it's testing the wrong number of components, but it's still the right thing to do. With this we could also potentially lift the limit in GL to what both the GL and vulkan blobs expose.

Merge request reports