On Tue, Aug 1, 2017 at 3:50 AM, Dave Airlie <airl...@gmail.com> wrote: > On 1 August 2017 at 09:40, Marek Olšák <mar...@gmail.com> wrote: >> Hi, >> >> This might slightly increase MRT performance. You need the amdgpu >> kernel driver if you want it for SI and CI. >> >> I've fixed a few of issues with the original code, enabled it for >> mipmapped textures, and added GFX9 support. I wasn't able to make it >> work with mipmapping on SI. There is also a hack to make mipmapping >> work with DCC on VI - my solution is to overallocate DCC to make >> corruption go away with tile swizzling. >> >> I've tested SI, CI, VI. >> >> I haven't tested GFX9. The GFX9 patch is just the first version for >> testing. With current GB_ADDR_CONFIG, only Vega10 can do tile >> swizzling. The code doesn't do anything for Raven due to its (possibly >> incorrect) GB_ADDR_CONFIG. >> > I think this will regress radv, as I don't see it setting the > SHAREABLE flag anywhere > in the radv code, if I missed that.
I think we should actually be ok, since we don't set the surf_index ptr in radv if the image is shareable. > > If you have a branch I'm happy to check it, or just run the Vulkan > deferred demo from > https://github.com/SaschaWillems/Vulkan/ > > Otherwise it all looks reasonable, I should probably add the fmask > counter to radv once it lands. > > So for the rest of them > Reviewed-by: Dave Airlie <airl...@redhat.com> > > Dave. > _______________________________________________ > mesa-dev mailing list > mesa-dev@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/mesa-dev _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
