Announcement

Collapse
No announcement yet.

Blumenkrantz "Massively Improves" Mesa's glReadPixels Performance With 7 Lines Of Code

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

  • Blumenkrantz "Massively Improves" Mesa's glReadPixels Performance With 7 Lines Of Code

    Phoronix: Blumenkrantz "Massively Improves" Mesa's glReadPixels Performance With 7 Lines Of Code

    Mike Blumenkrantz of Valve's open-source/Linux graphics driver team has submitted a merge request to "massively" improve the OpenGL glReadPixels performance within the common Mesa state tracker...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    What does it translate into in real life tasks? I wonder if it will mean performance for non Zink as well.

    Comment


    • #3
      It's a kind of magic magic magic

      Comment


      • #4
        7 lines of code yields such a clear & obvious improvement? Yikes! Not even the White House spin doctors could call that "a cheap fake".

        I guess the developer (or developers?) that wrote the original code (the code that Mr. Blumenkrantz improved) are now sharing a double facepalm ?

        Comment


        • #5
          Btw this patch applies cleanly on top of 24.1.2

          Comment


          • #6
            As someone who rights a lot of code these days... this sort of thing drives me nuts! I have personally done this myself... looked at code written years ago and realised that a couple of lines gives a massive performance boost. It makes me hate programming sometimes because you just cant see it when your writing it yourself.

            There was that patch to the kernel a couple years back that yielded some massive 100% improvement in drive performance on some SSD drives, and it just changed the order of the controller initialisation bits when turning the drive on... apparently the drives had some internal quirks... but it was crazy, it was a basically zero line patch but unless you looked at it with fresh eyes a long time later you could never see the path to the improvement.

            Just last night I threw away some two weeks of code work that I just could not find a path through, tackled the problem from what I thought was going to be the much harder angle and ended up with everything working after about four hours... it was both awesome and depressing at the same to when thinking about the two weeks of waste.

            Comment


            • #7
              Originally posted by zexelon View Post
              Just last night I threw away some two weeks of code work that I just could not find a path through, tackled the problem from what I thought was going to be the much harder angle and ended up with everything working after about four hours... it was both awesome and depressing at the same to when thinking about the two weeks of waste.
              That was not a waste, that was a learning experience. If we are viable human beings, we keep learning until our final breath.

              Comment


              • #8
                Originally posted by zexelon View Post
                As someone who rights a lot of code these days... this sort of thing drives me nuts! I have personally done this myself... looked at code written years ago and realised that a couple of lines gives a massive performance boost. It makes me hate programming sometimes because you just cant see it when your writing it yourself.

                There was that patch to the kernel a couple years back that yielded some massive 100% improvement in drive performance on some SSD drives, and it just changed the order of the controller initialisation bits when turning the drive on... apparently the drives had some internal quirks... but it was crazy, it was a basically zero line patch but unless you looked at it with fresh eyes a long time later you could never see the path to the improvement.

                Just last night I threw away some two weeks of code work that I just could not find a path through, tackled the problem from what I thought was going to be the much harder angle and ended up with everything working after about four hours... it was both awesome and depressing at the same to when thinking about the two weeks of waste.
                It’s always the last thing you try…

                Comment


                • #9
                  Brilliant. 7 lines of code.

                  Comment


                  • #10
                    Originally posted by zexelon View Post
                    it was both awesome and depressing at the same to when thinking about the two weeks of waste.
                    programming is iterative, you don't know what you don't know and the only way forward is to try stuff.
                    as you try more stuff you'll get a clearer picture of what you want/need to do, sometimes it does not go as planned and you have to start from scratch, maybe it was an obvious deadend for the trained eye but it does not matter, you now have a better understanding of the issue at hand and you'll be able to recognize similar patterns and will be able to save time focusing your energy on the right path in the future.
                    don't waste sleep on that, blumenkrantz wasn't born a graphic api genius, i'm sure he still makes silly mistakes sometimes and still miss "obvious" optimisation here and there and that's fine, that's the nature of programming, that's why we work as teams and review each other code.
                    the perfet programmer that gets everything right for the getgo does not exst, whoever claim otherwise is either a fraud or god himself.


                    Comment

                    Working...
                    X