Announcement

Collapse
No announcement yet.

Radeon ROCm 2.2 Released With Vega 20 Optimization, Caffe2 Multi-GPU Training

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

  • Radeon ROCm 2.2 Released With Vega 20 Optimization, Caffe2 Multi-GPU Training

    Phoronix: Radeon ROCm 2.2 Released With Vega 20 Optimization, Caffe2 Multi-GPU Training

    One month since the release of ROCm 2.1, the Radeon Open Compute stack has now been succeeded by ROCm 2.2...

    http://www.phoronix.com/scan.php?pag...m-2.2-Released

  • #2
    It is virtually impossible to get information on the Windows drivers. I'm hoping @bridgeman would help: is there ANYONE working on ROCm for Windows? I know, no ETA, no promises, but is it a priority whatsoever? Riding a 4-5 year old OpenCL implementation doesn't convey the message that OpenCL is supported in any way on Windows. It's almost as if it were a Linux-only feature (not ROCm, but OpenCL).

    Comment


    • #3
      There might not be a lot of changes, but many of those are a really big deal. That's some good work as far as I'm concerned.

      Comment


      • #4
        Originally posted by Meteorhead View Post
        [...] It's almost as if it were a Linux-only feature (not ROCm, but OpenCL).
        I have the opposite impression. Just installing the official video display driver from AMD on Windows is sufficient for using OpenCL for example with Blender 3D. Enabling OpenCL on Linux (again for the simple purpose of taking advantage of it in end-user applications) on the other hand feels quite confusing.

        Comment


        • #5
          Originally posted by Solid State Brain View Post
          I have the opposite impression. Just installing the official video display driver from AMD on Windows is sufficient for using OpenCL for example with Blender 3D. Enabling OpenCL on Linux (again for the simple purpose of taking advantage of it in end-user applications) on the other hand feels quite confusing.
          I have the impression that the best OpenCL in Linux will be from AMD and Intel once karol land the NIR Clover backend and HMM is implemented on AMDGPU and i also suspect will be a lot less buggy and a lot quicker to fix quirks than RocM/AMDGPU-pro.

          I might be wrong but i honestly believe this will be right based on the history of OpenGL and Vulkan so far

          Comment


          • #6
            Originally posted by Solid State Brain View Post
            I have the opposite impression. Just installing the official video display driver from AMD on Windows is sufficient for using OpenCL for example with Blender 3D. Enabling OpenCL on Linux (again for the simple purpose of taking advantage of it in end-user applications) on the other hand feels quite confusing.
            In fairness, installing AMDGPU-PRO on Linux should get you the same OpenCL code as Windows these days (which in turn is different from OpenCL-over-ROCm).

            Comment


            • #7
              bridgman
              I should have clarified that I meant with the open source drivers on Linux.
              (I'm a Radeon RX480 user under OpenSUSE Tumbleweed, which is a rolling distribution)

              Comment


              • #8
                Originally posted by Solid State Brain View Post
                bridgman
                I should have clarified that I meant with the open source drivers on Linux.
                (I'm a Radeon RX480 user under OpenSUSE Tumbleweed, which is a rolling distribution)
                Ahh, OK. Even there you should be able to install the closed-ish source OpenCL bits from the packaged drivers on top of upstream kernel/userspace and get working OpenCL if you can live with that.

                I say closed-ish because most of the OpenCL/PAL userspace is already open sourced via the ROCM and AMDVLK initiatives - it's just the glue between OpenCL and PAL that AFAIK has not been open sourced yet, and that's a function of time & people rather than anything inherently un-releaseable.

                Comment


                • #9
                  Originally posted by bridgman View Post

                  Ahh, OK. Even there you should be able to install the closed-ish source OpenCL bits from the packaged drivers on top of upstream kernel/userspace and get working OpenCL if you can live with that.

                  I say closed-ish because most of the OpenCL/PAL userspace is already open sourced via the ROCM and AMDVLK initiatives - it's just the glue between OpenCL and PAL that AFAIK has not been open sourced yet, and that's a function of time & people rather than anything inherently un-releaseable.
                  I'm not sure if Tumbleweed would work with our closed opencl stack... it's possible, but a safer bet would be installing the SLED drivers on Leap.
                  E.g. AFAIK, SLE 15 drivers would work on Leap 15.0, as long as you use the --no-dkms option in the installer as the dkms is not designed to work on Leap.

                  Something like this using 18.50:
                  ./amdgpu-install --headless --no-dkms --opencl=pal

                  Although if someone is building ROCm for tumbleweed, that might be a better option.

                  Comment


                  • #10
                    Originally posted by Meteorhead View Post
                    It is virtually impossible to get information on the Windows drivers. I'm hoping @bridgeman would help: is there ANYONE working on ROCm for Windows? I know, no ETA, no promises, but is it a priority whatsoever? Riding a 4-5 year old OpenCL implementation doesn't convey the message that OpenCL is supported in any way on Windows. It's almost as if it were a Linux-only feature (not ROCm, but OpenCL).
                    Boo hoo!

                    I'm roc(m)ing a lovely 2.2 that I just sudo apt updated on Ubuntu 18.04. Try it, you'll like it!

                    Comment

                    Working...
                    X