GPU not accessible on ADL-N platform while running DPC++ program (CL_DEVICE_NOT_FOUND)

[AvdheshChouhan]

Getting CL_DEVICE_NOT_FOUND while running DPCPP program on ADL-N platform. 
Also tried after installing intel compute-runtime libraries, still facing the same issue. 

terminate called after throwing an instance of 'cl::sycl::runtime_error'
what(): No device of requested type available. Please check https://software.intel.com/content/www/us/en/develop/articles/intel-oneapi-dpcpp-system-requirements.html -1 (CL_DEVICE_NOT_FOUND)
Aborted (core dumped)

[HemanthCH_Intel]

Hi,

Thank you for posting in Intel Communities.

Could you please provide the output of the below command to investigate more on your issue?

sycl-ls

Could you please provide the sample reproducer code?

Thanks & Regards,

Hemanth

[AvdheshChouhan]

Hi @HemanthCH_Intel ,

Thank you for the response. Please find the output of sycl-ls command below. 

#sycl-ls
[opencl:acc:0] Intel(R) FPGA Emulation Platform for OpenCL(TM), Intel(R) FPGA Emulation Device 1.2 [2022.13.3.0.16_160000]
[opencl:cpu:1] Intel(R) OpenCL, Genuine Intel(R) 0000 3.0 [2022.13.3.0.16_160000]
[host:host:0] SYCL host platform, SYCL host device 1.2 [1.2]

Its not a code specific issue, its not able to find the gpu device, getting the device using following code. gpu_selector gives "CL_DEVICE_NOT_FOUND" error while default_selector returns CPU.  

auto prop_list = property_list{property::queue::enable_profiling()};

queue q(gpu_selector{}, dpc_common::exception_handler, prop_list);

cout << "Running on " << q.get_device().get_info<info::device::name>();

PS: For my ADL-S setup, the gpu_selector only works with sudo, for for ADL-N its, not working with that either. 

Thanks,
Avdhesh

[HemanthCH_Intel]

Hi,

We can see from the output of "sycl-ls" command that GPU is not present in ADL-N machine. When GPU is not present and when we try to run a program using GPU selector then we get the error as mentioned in the above post. You can use host, CPU, and FPGA selectors instead of GPU selector.

Thanks & Regards,

Hemanth

[Lejin_Joy]

Hi Hemanth,

ADL-N contains integrated GPU and has EU cores available. Hence it should be possible to detect GPUs on ADL-N platform. Please let us know how we can get unblocked. 

Thanks,

Lejin 

[HemanthCH_Intel]

Hi,

Could you please try updating the GPU drivers using the below link:

https://www.intel.in/content/www/in/en/support/intel-driver-support-assistant.html

Thanks & Regards,

Hemanth

[Lejin_Joy]

The link that has been shared is generic driver link and is not helpful. After some exploration it  was found that ADL-N GPU Drivers are supported only in Linux Kernel version : https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/include/drm/i915_pciids.h?h=v5.19.8

/* ADL-N */#define INTEL_ADLN_IDS(info) \    INTEL_VGA_DEVICE(0x46D0, info), \    INTEL_VGA_DEVICE(0x46D1, info), \    INTEL_VGA_DEVICE(0x46D2, info)

However, when installing Ubuntu 20.04 with this version of kernel tADL N board does not even boot up. Could you please give specific links to Linux Ubuntu GPU driver compatible for ADL-N board. There is an urgent need and request your  accuarte support to solve this issue. 

Thanks,

Lejin  

[Jean_Intel]

Hello Lejin_Joy,

There is no official Linux Graphics driver version since the driver is made up of different subsystems like the kernel driver, Mesa 3D driver, and other drivers. Each has its current release version.

Most Linux-based* distributions already include Intel® Graphics Drivers. These drivers are provided and maintained by the Linux* distribution vendors and not by Intel. We recommend contacting the Linux* Operating System Vendors.

https://www.intel.com/content/www/us/en/support/topics/linux-osv.html

To have a better understanding of your issue, please provide me with the following:

• What is the GPU you are using?
• Are you having a problem with a custom app?
• Do you receive any error messages? Please provide me with a screenshot of the error message.
• What happens when you see the error?

Also, I noticed that you are not the original thread owner. However, I would like to confirm if you are working alongside with AvdheshChouhan (thread creator).

Best regards, 

Jean O.  

Intel Customer Support Technician

[AvdheshChouhan]

Hi Jean,

Thank you for the response. So, it is better to contact ubuntu support for the GPU driver issue right ?  I will check with ubuntu support to investigate any possible kernel issue, and in case the kernel driver is not the issue please check the the following logs and see if can find a solution. 

• What is the GPU you are using?
  • We are using the integrated GPU on the ADL-N platform.
• Are you having a problem with a custom app?
  • Trying to compile the DPCPP code using the oneapi base toolkit. 
• Do you receive any error messages? Please provide me with a screenshot of the error message.
  • When using the gpu_selector, getting following error: 
    An exception is caught while thresholding.
    terminate called after throwing an instance of 'cl::sycl::runtime_error'
      what():  No device of requested type available. Please check https://software.intel.com/content/www/us/en/develop/articles/intel-oneapi-dpcpp-system-requirements.html -1 (CL_DEVICE_NOT_FOUND)
    Aborted (core dumped)
• What happens when you see the error?
  • The error is not allowing to run code on GPU. 

I am working with Lejin on the same project and this is a blocker for us, and hence expecting a quick resolution here. 
Thank you in advance.

Thanks 

[Jean_Intel]

Hello AvdheshChouhan,

I appreciate the information provided.

It is correct since Intel does not provide an official Linux-based driver. You should contact your Operating System distro for support regarding their kernel and Driver-related issues.

Also, I understand that you are using the Integrated GPU; however, provide me with the exact model of the IGPU you are using.

Best regards, 

Jean O.  

Intel Customer Support Technician

[AvdheshChouhan · Accepted solution]

The issue was with kernel and the driver version as the support for ADL-N device id is in the latest release of compute runtime library and latest ubuntu kernel.  guc firmware was also not present, but now after all the dependencies I am able to detect the GPU and compile DPCPP code on ADL-N. 

Thank you.


https://github.com/intel/compute-runtime/commit/4b2f301e52ffc522ea09c0123ebf215ef0e6799d#diff-c38bae245ee95c52eb3b23ef1ab34f9060e0d5f430df0f9d105453453a76f115

[Jean_Intel]

Hello AvdheshChouhan,

We're glad to know the information helped, and that you could find the solution to your issue. Hopefully, it will help other community members. Since the thread is now solved, we will close it. If you need any additional information, please submit a new question as this thread will no longer be monitored.  

Best regards, 

Jean O.  

Intel Customer Support Technician