Ten days after my article r6xx/r7xx open source driver evolves fast!, there have been quite a lot of work and commits on mesa but it seems like my corruption problems are not going away.
It is time for another video !
Once again, let’s start with the video and we will discuss what changed (in the right or wrong way) right after.
You can download the video if you want to see it in its native resolution.
What are the changes ?
So first, I want to tell that I was not in a compositing environment when I did the video. But I can assure you it is roughly the same except for gtkperf.
It performs a bit faster than ten days ago (remember, we are not in a composited environment here). Maybe because of this commit.
Ha Ha! I love this! What a boost with the not-composited environment :) We went from 500FPS to 3000FPS. It is also slightly faster with composited environments as we have seen a 100fps improvement.
In a commit today, it was stated that the driver should be able to draw openarena. So, I gave it a try and ….. it works :)
It works but it is unplayable because of artefacts. Seems like mesa still needs some work for the HD4770.
Unfortunately, artefacts are still here and it didn’t really improve over time. Also, there are some problems using the raster engine instead of the dri-driver. But, at least, the computer is usable.
Hope this video will help developers in finding where the problem is.
Good job! It’s getting more and more stable but this is still not doing the trick on my HD4770. I won’t be there for two weeks, I hope it will work when I come back :)
Arch Linux x86_64 packages ?
As you may have seen, I have uploaded packages on the mupuf64 repo. Hope this will help some people in testing the driver. I’ll update it as often as I can until I leave my computer, on Friday. Then, updates will come back the September, 10th.
Thanks for reading :)
After asking for help on phoronix, Alex Deucher told me to set the Option EXANoDownloadFromScreen in xorg.conf. It solved my 2D corruption problem.
Concerning open arena, John Bridgman told me to try running it with the environment variable LIBGL_ALWAYS_INDIRECT=yes and it worked!
So, it seems like all my main problems are gone, thanks to everyone involved :)