Major Performance Breakthrough Discovered For Intel's Mesa Driver

[johnc]

So the Intel Linux driver team couldn't just pick up the phone and ask their Windows driver team counterpart about the performance differences? Valve had to spend its own money to hire a third-party developer to investigate the issue?

Intel strikes me as a completely dysfunctional company.

[souenzzo]

meanwhile, in nVidia....
GeForce 7025 still not work with Nouveau (full system crash) and work unstable with nVidia 304xx...

[FutureSuture]

So the Intel Linux driver team couldn't just pick up the phone and ask their Windows driver team counterpart about the performance differences? Valve had to spend its own money to hire a third-party developer to investigate the issue?

Intel strikes me as a completely dysfunctional company.

If you think that is dysfunctional, then head on over to AMD. How many drivers for Linux will AMD be maintaining? 1? 2? 3? I am not even sure now. I wish Valve would spend its own money to hire a third party developer to aid AMD. Competition is good but AMD needs some help in that department.

[bastiaan]

So the Intel Linux driver team couldn't just pick up the phone and ask their Windows driver team counterpart about the performance differences? Valve had to spend its own money to hire a third-party developer to investigate the issue?

Intel strikes me as a completely dysfunctional company.

To be fair, LunarG did more than just pick up the phone:

We started to suspect there was a bigger bottleneck masking the improvements, and sure enough we were able to generate a test program that showed a huge performance issue with how the hardware samplers were working as compared to the OpenGL driver running under windows.

So LunarG were evidently successful in pinpointing the issue to such an extent that Intel engineers were able to identify the source of the problem.

Now, it is quite possible that some people at Intel had been looking into the issue, but had not succeeded in narrowing it down. Or perhaps they had, but had not taken seriously the possibility that the Windows driver writers had information they didn't.

[dungeon]

Major Performance Breakthrough Discovered For Intel's Mesa Driver

They said "little chicken" is in kernel drm actually

We shared our test program with Intel, where they have the advantage of working directly with the hardware engineers. Sure enough, they found that little chicken bit this week.

The Intel DRM driver in Linux is the place to set this bit correctly, so a kernel patch is needed in order to see the base bottle neck addressed, then you can get yourself a GlassyMesa driver and see the improvements, too.

[dungeon]

While they went public today with this performance breakthrough, no patch is publicly available. Jens told me in an email that there are concerns that just setting this special bit and leaving it on isn't good enough as it could break some specific video acceleration functionality when enabled.

If i can guess they just playing with gart/vram managment, i can break a little vdpau like that on radeon driver and performance goes 40-60% up in some games too on my Kabini APU, while slowing in another games...

So it is not steady improvment for sure, showcase just for some games

[smitty3268]

Very broadly experienced bug for MONTHS: Random radeonsi crashes

After all this time, there's still not one person who's bisected the regression.

[brad0]

'we've discovered a way to make the radeonsi driver STABLE' would be even better.

'NVIDIA has discovered how to contribute documentation for their hardware and provide resources to develop the open source driver' that would be the best.

As if pigs will fly and hell freezes over.

[gigaplex]

After all this time, there's still not one person who's bisected the regression.

I had a look through that bug report. It quickly devolved into a dumping ground of "me too" reports of unrelated issues. They couldn't get a clear, consistent answer on which kernels were actually affected, so where would the bisection start?

[MrCooper]

Very broadly experienced bug for MONTHS: Random radeonsi crashes
Nobody knows what the heck is causing this. Radeon HD 7770 here on Arch Linux but with Mesa 10.1.4 to at least avoiding this blocker bug (but apparently Mesa is not to blame).

So far, everybody who claimed "doesn't happen with Mesa x.y" has later retracted that claim, because it still happened eventually. If it really never happens with 10.1.4 for you, can you try bisecting it?