Radeon Software 18.20 Preview Offers Early Support For Ubuntu 18.04 LTS & RHEL 7.5

[bridgman]

will we get the ROCm option back in the driver in the future? for me the Ethereum Performance was much better 140MHs to 180MHs for 6 vega-64 cards

I don't know what the plan is re: including ROCm in the packaged all-open/workstation drivers, but we are continuing to ship regular ROCm stack releases as well.

(Edit) bridgman is it possible for AMD to release the PAL as open-source to replace Clover OpenCL in mesa?

Most of the PAL code has been open sourced as part of AMDVLK - we have asked about open sourcing the OpenCL-to-PAL portion but it hasn't really been discussed much yet.

[bridgman]

Also what's the difference between ROCm and PAL ?

ROCm is our compute-focused stack, using hardware features initially developed for HSA then extended to mid-high end dGPUs. PAL is a more conventional graphics/compute stack, with command submission via ioctl to the kernel DRM driver.

Which one will be used for next AMD graphics cards ?

PAL will be implemented on all new hardware, integrated or discrete, large or small, while ROCm will be implemented mostly on the larger parts.

[RhumArrange]

I installed the opencl driver (PAL mode) on my debian (testing) system, it seems to work. I got a warning about failing to install some dkms driver, but everything seem to be fine anyway (it's running the 4.17 kernel, so maybe things have been mainlined).
What components should be built in order to get the same stack (opencl/PAL mode), in addition to PAL and llvm (found on the GPUOpen driver repo on github) ?
Also what's the difference between ROCm and PAL ? Which one will be used for next AMD graphics cards ?

[RussianNeuroMancer]

Yeah, that page is a couple of years old. I believe the latest information is here:

https://github.com/RadeonOpenCompute/ROCm

Good to know, thanks!

[bridgman]

I'm not sure if the recent packaged drivers actually include ROCm userspace - will check but I believe that was removed and replaced by PAL paths.

[bridgman]

It's alright. I take it that success is not guaranteed with Phenom II and it's even unlikely on the latest GPUs.

AFAIK PCIE atomics are still required for newer GPUs, so with a Phenom your best chance of success would be with a 290/390 GPU.

[bridgman]

I guess 5xx should be supported too, right bridgman?

Yeah, that page is a couple of years old. I believe the latest information is here:

https://github.com/RadeonOpenCompute/ROCm

[RussianNeuroMancer]

Probably just on your documentation: https://rocm.github.io/hardware.html

I guess 5xx should be supported too, right bridgman?

[Tomin]

Whoops, you're right - you said "platform" not "GPU". Even worse, English IS my first language

It's alright. I take it that success is not guaranteed with Phenom II and it's even unlikely on the latest GPUs.

I noticed that OpenMI has also OpenCL version. Does that work on other OpenCL stacks than just ROCm OpenCL? It seems interesting and I wonder why people porting Tensorflow to OpenCL haven't mentioned it.

[juno]

Do you know if OpenCL/PAL requires any non-upstreamed patches for kernel, libdrm and LLVM or would it "just work" on an up-to-date (rolling release) system?

Sorry for the bump, bridgman