Intel® oneAPI DL Framework Developer Toolkit
Get answers for developing new or customizing existing frameworks using common APIs.
Announcements
Welcome to the Intel Community. If you get an answer you like, please mark it as an Accepted Solution to help others. Thank you!

I want to run oneDNN on GPU

k_higashi
Beginner
521 Views

Please teach me the steps to run the oneDNN sample code on a GPU (DPC ++).

I have confirmed that it works on the CPU by following the steps below.

I have created a project by selecting oneDNN CNN FP32 Inference from Browse Intel oneAPI Samples in the Visual Studio IDE.

I run it referring to the contents of README.md.

step1 setup.

"C:\Program Files (x86)\intel\oneapi\setvars.bat" --dnnl-configuration=cpu_vcomp

step2 cmake

mkdir build
cd build
cmake -G "Visual Studio 16 2019" ..
cmake --build .

step3 run

bin\Debug\cnn-inference-f32-cpp.exe

The code will be run, but on CPU.

Use time: 15.12 ms per iteration.
Example passed on CPU.

Also, if you specify the gpu option in step 4, the following error will be displayed.
" Application couldn't find GPU, please run with CPU instead."

I think the error is wrong because other samples, such as Vector Add, can be run on the GPU.

I think this procedure was compiled with MSVC. (cpu only)
I want to build and run oneDNN sample code on GPU (DPC ++).


<Information>
OS: Windows 10 Home
Toolkit:
Microsoft Visual Studio 2019 Professional (ver.16.7.7)
Intel oneapi beta 10
Compiler: Intel oneAPI DPC++ Compiler (2021.1.2288)

CPU: Intel Core i7-1065G7 1.3GHz
Accelerator:Iris Plus Graphics
driver ver.: 27.20.100.8681


best regard.

0 Kudos
2 Replies
RahulV_intel
Moderator
508 Views

Hi,


We are redirecting your query to an appropriate forum for a faster response.


Thanks,

Rahul


AthiraM_Intel
Moderator
474 Views

Hi,


We are able to reproduce the issue. We are forwarding your case to Subject matter Experts.


Thanks


Reply