Re: Flipping the switch to Mesa 21?
[email protected] Tue, 23 Apr 2024 14:05:10 +0000
| Newsgroups | gmane.os.netbsd.devel.x11 |
|---|---|
| Message-ID | <[email protected]> |
On Tue, Apr 23, 2024 at 08:53:54AM +0100, Robert Swindells wrote: > > [sorry for the delay in replying, I have been away and wanted to update > my test systems to latest -current] > > matthew green <[email protected]> wrote: > >[email protected] writes: > >> Hi all, > >> > >> I tested out mesa 21, made some changes for intel only, it seems to work > >> well enough on my combination of test hardware (I have one GPU for each > >> DRM driver, and tested modesetting too with all of them). > > What have you tried nouveau on? > > I only have a really old NV44 laptop. I tested with an nvidia GT 730 card > > The missing intel files had also been reported by prlw1@, I had been > trying to work out where to put the intel_perf files. > > I got most of the way through adding files for Intel Iris and Crocus > but I don't think they are supported by the kernel code so stopped. > > >> Any reasons not to switch to HAVE_MESA_VER=21? > >> I am not aware of anything wrong with it. > > I see you have made the switch now, not suggesting you revert. > > > i tested on amdgpu ("R7", seems to be GCN 3, from bristol ridge APU), > > radeon 5450 (trusty!), and llvmpipe on pinebookpro. seems at least > > as functional as old mesa with a few tests. > > I have been using HAVE_MESA_VER=21 on a Radeon 5450 as well with no > problems. > > Tried amdgpu on a ThinkPad T295, doesn't really work but looks like > kernel problems, maybe fixed by a DRI update. > > For aarch64, I have been building with LLVMRT disabled so that I could > be sure that the lima or panfrost code was being used, I haven't tried > llvmpipe on it for a while. > > Also have the Makefile changes to build libGLESv2 if we wanted to add > that. Neat, wiz just said that there's a package trying to use it - x11/zutty, and wondered whether we can add it. If you can share the diff, that would be nice (or commit it, if you prefer) > > > rjs had reported a prior issue i don't recall, but this seems like a > > good time to at least test, and we can revert if there's something > > immediately bad to deal with.. > > I still get a core dump from glxgears on my one nouveau system, this > didn't happen with Mesa 19. > Can you share a backtrace?