Announcement

Collapse
No announcement yet.

AMD vs. Intel Contributions To The Linux Kernel Over The Past Decade

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

  • #11
    Originally posted by torsionbar28 View Post
    It takes $$$ to hire coders and AMD is a lot smaller than intel. The AMD code contributions align perfectly with the Zen CPU launch. Glad to see they're investing some of that Zen cash back into Linux.
    It surprises me that people don't see this. I'm not going to say that their allocation of employees is perfect but cash does mean new programmers and frankly those programmers take time to get up to speed. If AMD maintains its advantages into 2021 I suspect we will see a massive increase in driver and tools quality.

    Comment


    • #12
      It would've been more interesting to do the statistics per subsystem. E.g. AMD vs. Intel contributions within DRM. Intel works in more areas than AMD (storage, networking, etc.), which distorts the statistics.

      Also, "more is better" is a bogus premise.

      Also, Nvidia is missing from the graph 😂

      Comment


      • #13
        Originally posted by torsionbar28 View Post
        It takes $$$ to hire coders and AMD is a lot smaller than intel
        Intel software division makes AMD's look like a garage company division/operation.

        Comment


        • #14
          Originally posted by insilications View Post

          Intel software division makes AMD's look like a garage company division/operation.
          Yes, one cannot argue with money Intel has at its disposal.
          Intel produces: Ethernet, Wifi, FPGA, optical probably as well. Not only CPU, GPU and chipsets. Given this, almost 50% AMD's kernel contribution compared to Intel is quite large. Yes commits were big on AMD side due to generated GPU code being committed, but still, contribution nevertheless.
          Last edited by reavertm; 01-24-2020, 08:02 AM.

          Comment


          • #15
            It would be interesting to see what percentage of the development team at each company is working on linux. AMDs development team is tiny compared to intel.

            Both however should be commended on the work they do on linux, some just ignore it outside there own binary drivers... looking at you nvidia!

            Comment


            • #16
              Originally posted by atomsymbol View Post
              AMD has submitted some big C files which are basically machine generated and contain a lot of redundant/duplicate information from data compression viewpoint. This artificially inflates the number of lines changed. A more reliable indicator of contribution activity would be "By bytes changed after compression by gzip, xz or zstd".
              That might miss quite a bit of refactoring work, which is important for keeping the code base sane. But it would be interesting to see stats with anything purely machine generated filtered out.

              Regarding the other issue of limiting the stats to intel.com addresses only. That entirely misses a huge chunk of drm and i915 contributions, not sure about other subsystems.

              Comment


              • #17
                Do those numbers tell something important?
                Do they show quality of contribution?
                Or is this some kind of race?
                IMO useless statistic supporting fanboyism.
                Last edited by BaumKuchen; 01-24-2020, 04:06 PM. Reason: Edit: grammar

                Comment


                • #18
                  Originally posted by BaumKuchen View Post
                  Do those numbers tell something important?
                  Do they show quality of contribution?
                  Or is this some kind of race?
                  IMO useless statistic supporting fanboyism.
                  An interesting measurement would be to show by how much the contributions are increasing the cyclomatic complexity of the Linux kernel source code.

                  There does not exist an objective "final" indisputable measure of source code complexity. All there is are approximations and indirect indicators of source code complexity, upper and lower bounds, the "real" complexity number being somewhere within those bounds.

                  Comment


                  • #19
                    Originally posted by syrjala View Post
                    That might miss quite a bit of refactoring work, which is important for keeping the code base sane.
                    I am not sure what you mean exactly. In some cases, the keyboard input complexity of refactoring a function name of length 10 characters called from 500 callsites involves pressing just a few keyboard buttons in an IDE, so about 100 bits of information, which is much smaller than 10*500*log2(ASCII char) = 10*500*8 = 40000 bits of information.

                    Comment


                    • #20
                      Originally posted by wizard69 View Post

                      I think this charting shows you exactly what you are asking for. As the company has turned around they have more people committing. If they can deliver this year (all the promised new products), I can see a lot more effort flowing into software development for all platforms.

                      Frankly I just went out and bought a new AMD based system because I really believe that they are in recovery mode. It is frankly the time to further support the platform because it can only get better as software gets the attention profits allow.
                      Yes, but compared to Intel they seem to be pretty little.
                      And yes, in the last few years I bought only AMD GPUs because of their open source contributions and last year also a new AMD CPU.
                      Hopefully with the revenue they make from the GPUs and CPUs, they can affor to catch somewhat to Intel's contributions to the kernel and other open source software.

                      Comment

                      Working...
                      X