Why my same GPU was in different context?

Question

I have two rtx3090 gpus, and i want to bind them in one context. but below code:

int main() {
    std::vector<sycl::device> devices = sycl::device::get_devices();
    std::vector<sycl::device> gpu_devices;
    for (const auto& device : devices) {
        std::cout<<device.get_info<sycl::info::device::name>()<<std::endl;
        if (device.get_info<sycl::info::device::name>().find("3090") != std::string::npos) {
            gpu_devices.push_back(device);
        }
    }
    sycl::context ctx(gpu_devices);
    return 0;
}

it throw error : what(): Can't add devices across platforms to a single context. -33 (PI_ERROR_INVALID_DEVICE)

And the output of this code is:

Platform: Intel(R) OpenCL 
Device: Intel(R) Xeon(R) Gold 5218R CPU @ 2.10GHz 
Platform: Intel(R) OpenCL 
Device: Intel(R) Xeon(R) Gold 5218R CPU @ 2.10GHz 
Platform: Intel(R) FPGA Emulation Platform for OpenCL(TM) 
Device: Intel(R) FPGA Emulation Device 
Platform: NVIDIA CUDA BACKEND 
Device: NVIDIA GeForce RTX 3090 
Platform: NVIDIA CUDA BACKEND 
Device: NVIDIA GeForce RTX 3090

My expected output is

Platform: NVIDIA CUDA BACKEND
Device: NVIDIA GeForce RTX 3090
Device: NVIDIA GeForce RTX 3090

I want to know why the same GPU at different platform

Van Occupanther · Accepted Answer · 2024-11-04 10:59:54Z

0

It's not clear which version of oneAPI you are using here, but it does look quite old as far as I can tell. Earlier this year this patch was merged which should allow for your use-case to work. It was made available in the 2025.0 release of oneAPI. I don't have a multi-GPU setup here to be able to test this but downloading the new release of oneAPI will hopefully fix your problem!

edited Nov 4, 2024 at 10:59

answered Nov 1, 2024 at 20:08

Van Occupanther

12 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Alice Pedom Over a year ago

I use the latest release from the dpcpp offical and version is 2024.2.1 ,data is Aug. 5, 2024. And not only the gpus, but also the cpu was recognized in different platform.

Van Occupanther Over a year ago

I should have clarified in the post: oneAPI 2025.0 was released last week, and will have some updates that affect the code you're specifically trying. If that continues not to fix it please update!

illuhad Over a year ago

... Alice, otherwise if you happen to have an AdaptiveCpp installation around - AdaptiveCpp has supported this use case since ~2020 and should work out of the box :-)

Collectives™ on Stack Overflow

Why my same GPU was in different context?

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related