Automatic mipmap generation using glGenerateMipmap (util_gen_mipmap) fills mipmaps with wrong rendering. Piglit tests fbo-generatemipmap* are failing. Software fallback works properly, as does llvmpipe. Hardware: rv670 (HD3870), amd64 Software: kernel 2.6.38rc8, mesa from git, libdrm from git
Created attachment 45598 [details] [review] short patch to test problem Can you try this patch to mesa and say if it fixes the issue?
Indeed it helps. Piglit test fbo-generatemipmap-formats is still failing, but may be unrelated problem. Basic functionality now works properly.
(In reply to comment #1) > Created an attachment (id=45598) [details] > short patch to test problem > > Can you try this patch to mesa and say if it fixes the issue? It also fixes a rv670 lodbias regression - https://bugs.freedesktop.org/show_bug.cgi?id=35367
Patch is strictly a work-around and a bloody hack (dripping even). Indicates there's something should be happening in the driver code that isn't. "Fixes" was a bad word choice. Anyway, I can confirm this bug on RS880 (Radeon HD 4200).
This should be fixed by the following patches: http://cgit.freedesktop.org/mesa/mesa/commit/?id=d116fe51c1aee3453290ac30ffe993bc131c53f7 http://cgit.freedesktop.org/mesa/mesa/commit/?id=5066f7cd0e010db46b42f8f08a33c10cb34f8726 http://cgit.freedesktop.org/mesa/mesa/commit/?id=bf7a3ddca6fcc73eebdd9afdc45a485b38477bd9
Still doesn't work on RS880 until it's added to the conditional on line 813 of r600_hw_context.c I wonder if this should apply to rv620, rv635 (and rs780?) also?
fixed in b9f2750900efb2667bc1dfc428b71d8a08b212be. rv610/630/620/635 all work fine without the additional hack.
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.