Does oneapi::mkl::rng work on Nvidia GPU?

[parody9]

Hello!

I migrated CUDA code to DPC++ using dpct.

But I want to execute migrated code using GPU.

I heard that oneAPI doesn't support NVIDIA GPU yet. Is it true?

If Intel support NVIDIA GPU, How can i install SYCL Platform of GPU and execute migrated code on my GPU? 

And in this case does oneapi::mkl::rng work on GPU? 

Here is my sycl-ls --verbose result. I am using ryzen 5 3700x.

Platforms: 3

Platform [#1]:

    Version  : OpenCL 1.2 Intel(R) FPGA SDK for OpenCL(TM), Version 20.3

    Name     : Intel(R) FPGA Emulation Platform for OpenCL(TM)

    Vendor   : Intel(R) Corporation

    Devices  : 1

        Device [#1]:

        Type       : ACC

        Version    : OpenCL 1.2

        Name       : Intel(R) FPGA Emulation Device

        Vendor     : Intel(R) Corporation

        Driver     : 2020.11.12.0.14_160000

Platform [#2]:

    Version  : OpenCL 2.1 LINUX

    Name     : Intel(R) OpenCL

    Vendor   : Intel(R) Corporation

    Devices  : 1

        Device [#1]:

        Type       : CPU

        Version    : OpenCL 2.1 (Build 0)

        Name       : AMD Ryzen 7 3700X 8-Core Processor             

        Vendor     : Intel(R) Corporation

        Driver     : 2020.11.12.0.14_160000

Platform [#3]:

    Version  : 1.2

    Name     : SYCL host platform

    Vendor   :

    Devices  : 1

        Device [#1]:

        Type       : HOST

        Version    : 1.2

        Name       : SYCL host device

        Vendor     :

        Driver     : 1.2

default_selector()      : CPU : OpenCL 2.1 (Build 0)[ 2020.11.12.0.14_160000 ]

host_selector()         : CPU : OpenCL 2.1 (Build 0)[ 2020.11.12.0.14_160000 ]

accelerator_selector()  : ACC : OpenCL 1.2 [ 2020.11.12.0.14_160000 ]

cpu_selector()          : CPU : OpenCL 2.1 (Build 0)[ 2020.11.12.0.14_160000 ]

gpu_selector()          : No device of requested type available. Please chec...

custom_selector(gpu)    : No device of requested type available. Please chec...

custom_selector(cpu)    : CPU : OpenCL 2.1 (Build 0)[ 2020.11.12.0.14_160000 ]

custom_selector(acc)    : ACC : OpenCL 1.2 [ 2020.11.12.0.14_160000 ]

[parody9]

My clinfo

Platform #0: Intel(R) FPGA Emulation Platform for OpenCL(TM)

`-- Device #0: Intel(R) FPGA Emulation Device

Platform #1: Intel(R) OpenCL

`-- Device #0: AMD Ryzen 7 3700X 8-Core Processor             

Platform #2: NVIDIA CUDA

`-- Device #0: GeForce RTX 2080 SUPER

[AbhishekD_Intel · Accepted solution]

Hi,

Thanks for reaching out to us.

Currently, Intel oneAPI toolkits target Intel specific devices. The individual tools from the toolkits support some additional operating systems and architecture configurations. Please check the system requirements and hardware support articles related to specific tools to get more details regarding the supported hardware.

Please refer to the below links for system requirements and hardware support for Intel oneAPI Basekit.

https://software.intel.com/content/www/us/en/develop/articles/intel-oneapi-base-toolkit-system-requirements.html

https://software.intel.com/content/www/us/en/develop/articles/how-to-get-the-hardware-that-supports-the-oneapi-program.html

For the Intel oneAPI MKL related issue please refer to the below article for more details.

https://software.intel.com/content/www/us/en/develop/articles/oneapi-math-kernel-library-system-requirements.html

If you want to execute migrated DPC++ code on NVIDIA GPU then please refer to the open-source version of DPC++. There is experimental support of DPC++ for CUDA devices. Please refer to the below link for more details.

https://intel.github.io/llvm-docs/GetStartedGuide.html

Hope the provided details will help you to resolve your issues.

Warm Regards,

Abhishek

[parody9]

Hi Abhishek

Thanks for your support!

I hope that oneAPI supports various devices someday.

[AbhishekD_Intel]

Hi,

Thanks for the confirmation. Let's hope for the best. We will no longer monitor this issue, please post a new thread if you have any other issues.

Warm Regards,

Abhishek

[TonyMongkolsmai]

Late late update (2 years later), but for others who may be searching.  The oneAPI spec does support AMD and NVIDIA GPUs, and if you are using the Intel oneAPI Base Toolkit, you can get a supported version of NVIDIA support as well as a version that runs on AMD GPUs using Codeplay plugins.

https://developer.codeplay.com/products/oneapi/nvidia/home/

https://developer.codeplay.com/products/oneapi/amd/home/