This is a bug report to avoid losing Henri Verbeet's patch series which fixes gallium's use of the first_level field of sampler views. Basically, the texture lod should be computed relative to level==first_level, not level=0. We have patches for softpipe and r300 but llvmpipe (at least) needs to be updated too.
Created attachment 42910 [details] [review] patch for softpipe
Created attachment 42911 [details] [review] patch for r300
Created attachment 42912 [details] [review] patch for state tracker
Created attachment 42913 [details] [review] second patch for state tracker
Created attachment 44101 [details] [review] patch for r300 rebased Is there any reason not to commit the patches? They work fine on r300g and softpipe.
I'd like to get this resolved for llvmpipe too. I just haven't had the time.
Created attachment 45093 [details] [review] patch for llvmpipe
Brian, Fabian attached the patch for llvmpipe. Can I now push all of them?
I've taken a look at the llvmpipe changes and they look good. But I haven't had time to apply all the patches and do any testing yet. If you've tested with piglit, etc. and everything looks good then go ahead and commit the patches.
All the variants of the tex-miplevel-selection test pass with softpipe, llvmpipe, and r300g (for the first time!). All patches have been pushed. Closing.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.