Archives Discussions

[zazola]

Hi all,

I have been trying to have the AMDGPU-PRO OpenCL SDK in a Radeon Vega FE GPU I recently got, but despite I have tried many versions (17.20-445420, 17.30-465504 mainly) I always get this error when running clinfo:

terminate called after throwing an instance of 'cl::Error'

  what():  clGetPlatformIDs

Aborted

When I try to compile the sample from: https://raw.githubusercontent.com/bgaster/opencl-book-samples/master/src/Chapter_2/HelloWorld

I see that is not able to see any platform in the system:

Failed to find any OpenCL platforms.

Failed to create OpenCL context.

The OS is CentOS 7.3, kernel version 3.10.0-693.2.2

The CPU is a Intel Xeon Silver 4110

Motherboard is a Supermicro X11DPG-QT

Installed packages:

amdgpu-pro.x86_64                    17.20-445420.el7          @amdgpu-pro-local

rocm-amdgpu-pro.x86_64               17.20-445420.el7          @amdgpu-pro-local

(and many more amdgpu-pro related)

ls /etc/OpenCL/vendors reports:

amdocl32.icd  amdocl64.icd  amdocl-rocr64.icd

lspci reporst this:

b1:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Device 6863

b1:00.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Device aaf8

But the most thing was what dmesg reports about amd:

[    0.893006] amd_nb: Cannot enumerate AMD northbridges

I attach also many other lspci outputs may be interesting. If anyone knows if I'm missing something or there is something wrong in my hardware configuration, I really would be grateful you could share it with me.

Thanks in advance for your help,

Aritz.

[dipak]

Please make sure that ROCm component (optional) required for running OpenCL applications is installed correctly [i.e. installing “rocm-amdgpu-pro” and setting “LLVM_BIN” path].

Btw, searching for frontier edition points to this link: Radeon Pro Software . Please try this one.

[though the page shows version 17.10 but the downloaded package indicates AMDGPU PRO 17.30]

Regards,

[zazola]

Yes, I have amdgpu-pro and rocm-amdgpu-pro installed, with LLVM_BIN env. var set up.

I tried 17.30-465504 version of the driver from the link you provided, and now I can see v17.30-479365 is available but still I get the same error.

Regards,

[dipak]

Thanks. I'll check with the concerned team and get back to you.

Regards,

[dipak]

Hi,

It looks like a kernel compatibility issue. The kernel version ( “3.10.0-693.2.2” ) used is same kernel as RHEL 7.4 ( https://access.redhat.com/articles/3078 ). If so, currently we do not support RHEL 7.4/ CentOS 7.4. Someone else also reported a similar issue here: amdgpu-pro-17.30-465504 wont install on centos 7.3 . Please downgrade the kernel and try.