bridgman
I am on Gentoo so I can actually move in any direction that can lead me to working opencl again: custom kernel, dksm or hybrid stack, whatever. Last time I tried opencl from 17.30.465504 packaged separately for Gentoo to be used with mainline kernel driver. From the fact Gentoo packaged it, I may conclude such a hybrid works for somebody (otherwise they wouldn't package it), but not for me this far.
However FYI: it is reported that amdgpu-pro 17.30 doesn't work on Cape Verde on Ubuntu 16.04, too. From darktable bug tracker:
The same error is reported that I get on my hybrid (mainline 4.12 amdgpu + opencl from amdgpu-pro 17.30):
Here's some explanation I got from darktable devs:
(Quoted from here https://redmine.darktable.org/issues/11702 ).
That makes me think that the bug is effective on both closed and hybrid stacks on Cape Verde (say hi to AMD's QA). However all my previous attempts resulted in userspace segfaults and last time both clinfo and darktable see opencl but the apps still can't create a commmand queue so it's maybe a progress.
I opened bugs at Gentoo bugzilla on dev-libs/amdgpu-pro-opencl package and at kernel bugzilla on amdgpu with no useful feedback so far... It would be so kind of you if you suggest how to make this goddamned thing eventually work again in full...
My current configuration is:
kernel 4.12.0-pf7 (tried few 4.11 and 4.12 kernels)
xorg-server-1.19.3
libdrm-2.4.82
x11-drivers/xf86-video-amdgpu-1.3.0
dev-libs/amdgpu-pro-opencl-17.30.465504
llvm/clang 4.0.1
I am on Gentoo so I can actually move in any direction that can lead me to working opencl again: custom kernel, dksm or hybrid stack, whatever. Last time I tried opencl from 17.30.465504 packaged separately for Gentoo to be used with mainline kernel driver. From the fact Gentoo packaged it, I may conclude such a hybrid works for somebody (otherwise they wouldn't package it), but not for me this far.
However FYI: it is reported that amdgpu-pro 17.30 doesn't work on Cape Verde on Ubuntu 16.04, too. From darktable bug tracker:
I get the same error with my Radeon R7 250 (GCN 1.0) on a clean install of Ubuntu GNOME 16.04.3, and the newest AMDGPU-PRO driver (17.30). Darktable's OpenCL works fine with this card on Ubuntu GNOME 14.04 with the discontinued fglrx driver.
[opencl_init] could not create command queue for device 0: -6
Error code -6 relates to CL_OUT_OF_HOST_MEMORY. And that's what the docs tell us about this error during command queue creation:
CL_OUT_OF_HOST_MEMORY if there is a failure to allocate resources required by the OpenCL implementation on the host.
CL_OUT_OF_HOST_MEMORY if there is a failure to allocate resources required by the OpenCL implementation on the host.
That makes me think that the bug is effective on both closed and hybrid stacks on Cape Verde (say hi to AMD's QA). However all my previous attempts resulted in userspace segfaults and last time both clinfo and darktable see opencl but the apps still can't create a commmand queue so it's maybe a progress.
I opened bugs at Gentoo bugzilla on dev-libs/amdgpu-pro-opencl package and at kernel bugzilla on amdgpu with no useful feedback so far... It would be so kind of you if you suggest how to make this goddamned thing eventually work again in full...
My current configuration is:
kernel 4.12.0-pf7 (tried few 4.11 and 4.12 kernels)
xorg-server-1.19.3
libdrm-2.4.82
x11-drivers/xf86-video-amdgpu-1.3.0
dev-libs/amdgpu-pro-opencl-17.30.465504
llvm/clang 4.0.1
Comment