Created attachment 131115 [details]
launched UH4 on RS880 / Kernel 4.10.10 with mesa git
demo runs well until you press "F2" in order to activate wireframe mode, which causes GPU lockups
tested mesa: 12, 13, 17, 17.1 and git - all same.
can you please verify, that it is a _combination_ of wireframe _and_ tessellation?
I've got that for ages with r600g, RV730 (HD 4650, 1GB, AGP) and NI/Turks XT (HD 6670, 2 GB PCIe). As 'Heaven' (UH4) start 'normally' with tessellation enabled and 'you' press 'F2' (enable wireframe) it could result in GPU lockup. So please start 'Heaven', press 'F3' (disable tessellation) and last press 'F2' (enable wireframe). This 'version' works fine for me. It seems to be a combination of both. But my systems mostly recover from the GPU lockup.
W'll try to get a new backtrace, soon.
Tessellation is so awfully _slow_ (hardware or 'only' software implemantation) that I think we have to look/point, there.
Bas (Radv) worked on radeonsi tessellation improvements.
See related links:
I ran all the time with MESA_GL_VERSION_OVERRIDE=4.1COMPAT and 'Heaven' show display corruption (missing parts), but wireframe works. If I toggle tessellation (F3) missing parts arise but without tessellation, as expected.
So for 'Heaven' I have to disable MESA_GL_VERSION_OVERRIDE and get the above boom (GPU lockup).
=> tessellation / wireframe correlation
'All' old and Mesa git, Kernel 4.11-rc8 (and former)
My card does not support tess and I also disabled it in the launcher. Also my card is only GL3.3.
I ran apitrace on heaven_x64 and now the thing doesn't crash anymore :-( So maybe commands come to fast in wireframe mode and need some fence/sync point?
-- GitLab Migration Automatic Message --
This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.
You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/600.