Announcement

Collapse
No announcement yet.

New Heterogeneous Memory Management For Linux, Will Be Supported By NVIDIA/Nouveau

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

  • New Heterogeneous Memory Management For Linux, Will Be Supported By NVIDIA/Nouveau

    Phoronix: New Heterogeneous Memory Management For Linux, Will Be Supported By NVIDIA/Nouveau

    Jerome Glisse has sent out the latest version of his patches now for Heterogeneous Memory Management (HMM), which he's been working on the Linux kernel since 2014...

    http://www.phoronix.com/scan.php?pag...=Linux-HMM-V13

  • #2
    Page fault?! Why isn't this done in compile time? If the compiler knows those instructions are going to the GPU, it should infer the pointer needs GPU mapping. What the hell...

    This madness is why people still get paid for writing assembly.

    Comment


    • #3
      opencl 3 he says...lol. as if. we'd be lucky to see these dirtbags support 2.x within our lifetimes. maybe if they are lucky future generations will be able to utilize opencl 3.x ... on their quantum computers in their lab on mars.

      Comment


      • #4
        Can someone who knows post what hardware the code supports? I see Mellanox & NVidia Pascal mentioned but no explicit mention of AMD APUs or GPUs. The basic idea described sound like what AMD has been trying to achieve with HSA. I'm curious as to how they are related.

        Comment


        • #5
          Originally posted by c117152 View Post
          Page fault?! Why isn't this done in compile time? If the compiler knows those instructions are going to the GPU, it should infer the pointer needs GPU mapping. What the hell...

          This madness is why people still get paid for writing assembly.
          because it isn't known at compile time..

          btw, do you know how mm works? Basically every page of memory you touch is faulted the first time it is accessed. And lots of stuff is CoW which works by page faulting too.

          Comment


          • #6
            Originally posted by robclark View Post

            because it isn't known at compile time..

            btw, do you know how mm works? Basically every page of memory you touch is faulted the first time it is accessed. And lots of stuff is CoW which works by page faulting too.
            Yeah but don't let the facts get in the way of a good rant!

            Comment


            • #7
              Who casts the result of malloc uuuurgh. it returns a void* so no need to cast!

              http://www.dirtcellar.net

              Comment


              • #8
                Originally posted by c117152 View Post
                Page fault?! Why isn't this done in compile time?
                compile-time malloc? are you on drugs?

                Comment


                • #9
                  Originally posted by waxhead View Post
                  Who casts the result of malloc uuuurgh. it returns a void* so no need to cast!
                  only in inferior languages

                  Comment


                  • #10
                    Originally posted by bsp2020 View Post
                    The basic idea described sound like what AMD has been trying to achieve with HSA. I'm curious as to how they are related.
                    me too

                    Comment

                    Working...
                    X