4 ms·

Why do you say the Mesa OpenGL is a total joke?

by madez 9y ago

orbifold 9y ago
This is a bit outdated http://richg42.blogspot.de/2014/05/the-truth-on-opengl-driver-quality.html?m=1 http://richg42.blogspot.de/2014/05/the-truth-on-opengl-drive... overview of driver status by a game developer, vendor A is NVIDIA and as the article points out they are the only one with a performant relatively bug free implementation. Also notice how he mocks Intel for having two driver teams: That Linux expects to get special treatment by integrating the driver into their Graphics abstractions leads demonstrably to worse performance and less features than if you bypass all those abstractions and use essentially the same driver for all kernels.
- madez 9y ago
  Thanks for that link. I must say I deem it maybe not completely outdated, but at least worthy an update. But I'm in awe of the what one can read there. "This vendor[Nvidia] is extremely savvy and strategic about embedding its devs directly into key game teams to make things happen. (...). These embedded devs will purposely do things that they know are performant on their driver, with no idea how these things impact other drivers. (...) Vendor A[Nvidia] is also jokingly known as the "Graphics Mafia". Be very careful if a dev from Vendor A gets embedded into your team. These guys are serious business." So, basically Nvidia is sabotaging OpenGL to fuck up the specs and then implement other working variations and make the game developers use their version? If that is true, fuck Nvidia. "On the bright side, Vendor C[Intel] feeds this driver team[Windows Driver Team] more internal information about their hardware than the other team[Linux Driver team]. So it tends to be a few percent faster than driver #1 on the same title/hardware - when it works at all." What the fuck is going on in this industry? Intel is sabotaging its own Linux driver team? Why? "I don't have any real experience or hard data with these drivers, because I've been fearful that working with these open source/reverse engineered drivers would have pissed off each vendor's closed source teams so much that they wouldn't help. Vendor A[Nvidia] hates these drivers because they are deeply entrenched in the current way things are done." That, now finally, makes sense. Nvidia is strong-arming developers to not support Mesa because they are afraid of it. Nvidia is afraid of Mesa. I think this should be more widely known.
  - orbifold 9y ago
    The way I read this was a bit different: NVidia actually is the only vendor that offers a performant, complete and relatively bug free implementation. For example if you consider this http://gdcvault.com/play/1020791/ http://gdcvault.com/play/1020791/ presentation then it is relatively clear that most major innovations were first available as OpenGL extensions by NVidia. The playing field might have levelled somewhat with the introduction of vulkan, which eliminates a lot of code that had to reside in the driver before. The main reason why Mesa is unlikely to catch up is because the backend compiler code is platform specific, so unless NVidia decides to publish their platform specification, it is unlikely that they will achieve meaningful success. Even if NVidia did publish a specification and left driver development to the community it is unclear to me who would be willing to do the free work for them.
    - Nullabillity 9y ago
      Keep in mind that the blog post is from 2014. Since then AMD has rewritten their Linux driver (fglrx -> AMDGPU) which didn't really pay off before their 4xx series (released 2016).

Why do you say the Mesa OpenGL is a total joke?

by madez 9y ago

Why do you say the Mesa OpenGL is a total joke?

This is a bit outdated http://richg42.blogspot.de/2014/05/the-truth-on-opengl-driver-quality.html?m=1 http://richg42.blogspot.de/2014/05/the-truth-on-opengl-drive... overview of driver status by a game developer, vendor A is NVIDIA and as the article points out they are the only one with a performant relatively bug free implementation. Also notice how he mocks Intel for having two driver teams: That Linux expects to get special treatment by integrating the driver into their Graphics abstractions leads demonstrably to worse performance and less features than if you bypass all those abstractions and use essentially the same driver for all kernels.

madez 9y ago
Thanks for that link. I must say I deem it maybe not completely outdated, but at least worthy an update. But I'm in awe of the what one can read there. "This vendor[Nvidia] is extremely savvy and strategic about embedding its devs directly into key game teams to make things happen. (...). These embedded devs will purposely do things that they know are performant on their driver, with no idea how these things impact other drivers. (...) Vendor A[Nvidia] is also jokingly known as the "Graphics Mafia". Be very careful if a dev from Vendor A gets embedded into your team. These guys are serious business." So, basically Nvidia is sabotaging OpenGL to fuck up the specs and then implement other working variations and make the game developers use their version? If that is true, fuck Nvidia. "On the bright side, Vendor C[Intel] feeds this driver team[Windows Driver Team] more internal information about their hardware than the other team[Linux Driver team]. So it tends to be a few percent faster than driver #1 on the same title/hardware - when it works at all." What the fuck is going on in this industry? Intel is sabotaging its own Linux driver team? Why? "I don't have any real experience or hard data with these drivers, because I've been fearful that working with these open source/reverse engineered drivers would have pissed off each vendor's closed source teams so much that they wouldn't help. Vendor A[Nvidia] hates these drivers because they are deeply entrenched in the current way things are done." That, now finally, makes sense. Nvidia is strong-arming developers to not support Mesa because they are afraid of it. Nvidia is afraid of Mesa. I think this should be more widely known.
- orbifold 9y ago
  The way I read this was a bit different: NVidia actually is the only vendor that offers a performant, complete and relatively bug free implementation. For example if you consider this http://gdcvault.com/play/1020791/ http://gdcvault.com/play/1020791/ presentation then it is relatively clear that most major innovations were first available as OpenGL extensions by NVidia. The playing field might have levelled somewhat with the introduction of vulkan, which eliminates a lot of code that had to reside in the driver before. The main reason why Mesa is unlikely to catch up is because the backend compiler code is platform specific, so unless NVidia decides to publish their platform specification, it is unlikely that they will achieve meaningful success. Even if NVidia did publish a specification and left driver development to the community it is unclear to me who would be willing to do the free work for them.
  - Nullabillity 9y ago
    Keep in mind that the blog post is from 2014. Since then AMD has rewritten their Linux driver (fglrx -> AMDGPU) which didn't really pay off before their 4xx series (released 2016).