Zink OpenGL-Over-Vulkan At ~97% Piglit Testing Conformance

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • nanonyme
    replied
    Originally posted by uid313 View Post
    The Zink code in Mesa is at OpenGL 3.3, so when will the new updated Zink code with support for OpenGL 4+ be merged into mainline Mesa?

    With it being 69% the speed of Intel's OpenGL driver, is there any room for improvement? What can we expect? What can we hope for?
    Also another interesting point is how comparable Zink with DSA will be to regular GL driver with DSA. There might be less overhead emulating OpenGL at that point.

    Leave a comment:


  • oiaohm
    replied
    Something to consider is if you put zink head to head with llvmpipe or any of the other software options its insanely fast. 50 percent of Intel at first does not sound that good.

    Lets say you have a program that for some reason is glitching with the native opengl driver. Current options is drop to llvmpipe and in most cases be lucky to get 5 frames per second.

    If Zink ends up faster that is good. But Zink at it current speed for the cases where drivers don't work right its a hell of a improvement.

    Leave a comment:


  • smitty3268
    replied
    Originally posted by microcode View Post

    Little or no effort has gone into the performance of Zink at this time, so comparing it and especially using it as the basis for unrelated comparisons is silly.
    I don't think that's entirely true. Mike Blumenkrantz spent a while optimizing things over the last month, and made really good progress. He also said that he reached the limit of what is easily achievable for now, which is why he switched back over to fixing bugs. There will certainly be more performance changes to come, but it's at the point where I think performance comparisons are useful and unlikely to massively change day to day anymore.

    Leave a comment:


  • _ONH_
    replied
    If the 4.5 or something in that ballpark is upstream, beforehand it seems to me to take to much effort which changes altogether to fast I think.

    Leave a comment:


  • R41N3R
    replied
    Some benchmarks would be interesting at this point :-) Just to see how well it works already now on ANV & RADV compared to the classic OpenGL drivers.

    Leave a comment:


  • microcode
    replied
    Originally posted by cl333r View Post
    Zink translates from OpenGL and DXVK from DirectX, so they do the same type of job, hence I'm wondering if Zink is 65% of native OpenGL speed how many percent is DXVK of DirectX? Also about 65%?
    Little or no effort has gone into the performance of Zink at this time, so comparing it and especially using it as the basis for unrelated comparisons is silly.

    Leave a comment:


  • log0
    replied
    I may be passing over 97% of the test cases I’m running, but that doesn’t mean that zink is conformant for any versions of GL or ES, which may not actually be possible at present (without huge amounts of awkward hacks) given the persistent issues zink has with provoking vertex handling. I expect this situation to change in the future through the addition of more Vulkan extensions, but for now I’m just accepting that there’s some areas where zink is going to misrender stuff.
    And there is that. So Vulkan has to add specific extensions as was done for DXVK. And Vulkan drivers have to support those extensions for it to work.

    Leave a comment:


  • uid313
    replied
    The Zink code in Mesa is at OpenGL 3.3, so when will the new updated Zink code with support for OpenGL 4+ be merged into mainline Mesa?

    With it being 69% the speed of Intel's OpenGL driver, is there any room for improvement? What can we expect? What can we hope for?

    Leave a comment:


  • R41N3R
    replied
    Mike is doing a fantastic job with Zink and it is very interesting to read about the progress made in has branch! Hope more patches land soon in Mesa mainline :-)

    Leave a comment:


  • AsuMagic
    replied
    Originally posted by _ONH_ View Post
    I wonder how zink runs on radv or amdvkl. Did anyone test thisalready?
    yes, i tested it on radv a while ago. it works.
    if you're talking about performance i didn't really bother to measure that but it was significantly lower than radeonsi. that was weeks ago, though, so it missed the recent optimization stuff. it was a 2d game with pretty lightweight rendering so it was still in the hundreds of fps and bound by draw call performance, essentially.

    Leave a comment:

Working...
X