前提

1.安装Intel® oneAPI Toolkits
https://software.intel.com/content/www/us/en/develop/documentation/installation-guide-for-intel-oneapi-toolkits-linux/top.html
下载安装Base版，注意版本，尽量安装新版本

2.安装GPU驱动与CUDA
https://developer.nvidia.com/cuda-downloads
建议为11.8及以上版本
nvidia-smi能出现cuda版本

安装插件

1.依赖库

Ubuntu

sudo apt update
sudo apt -y install cmake pkg-config build-essential

Red Hat and Fedora

sudo yum update
sudo yum -y install cmake pkgconfig
sudo yum groupinstall "Development Tools"

SUSE

sudo zypper update
sudo zypper --non-interactive install cmake pkg-config
sudo zypper --non-interactive install pattern devel_C_C++

验证

which cmake pkg-config make gcc g++显示
/usr/bin/cmake
/usr/bin/pkg-config
/usr/bin/make
/usr/bin/gcc
/usr/bin/g++

2.下载

https://developer.codeplay.com/products/oneapi/nvidia/download/
对应自己的版本，没有选低一点的版本

安装

chmod +x oneapi-for-nvidia-gpus-2023.1.0-cuda-12.0-linux.sh
sh oneapi-for-nvidia-gpus-2023.1.0-cuda-12.0-linux.sh

安装之前oneapi安装的位置运行

. /opt/intel/oneapi/setvars.sh --include-intel-llvm
或者
. ~/intel/oneapi/setvars.sh --include-intel-llvm

配置.bashrc(按自己路径)

export PATH=/PATH_TO_CUDA_ROOT/bin:$PATH
export LD_LIBRARY_PATH=/PATH_TO_CUDA_ROOT/lib:$LD_LIBRARY_PATH

查看GPU

sycl-ls

显示本机的gpu如[ext_oneapi_cuda:gpu:0] NVIDIA CUDA BACKEND, TITAN RTX 0.0 [CUDA 11.0]

验证

#include <sycl/sycl.hpp>int main() {// Creating buffer of 4 ints to be used inside the kernel codesycl::buffer<sycl::cl_int, 1> Buffer(4);// Creating SYCL queuesycl::queue Queue;// Size of index space for kernelsycl::range<1> NumOfWorkItems{Buffer.size()};// Submitting command group(work) to queueQueue.submit([&](sycl::handler &cgh) {// Getting write only access to the buffer on a deviceauto Accessor = Buffer.get_access<sycl::access::mode::write>(cgh);// Executing kernelcgh.parallel_for<class FillBuffer>(NumOfWorkItems, [=](sycl::id<1> WIid) {// Fill buffer with indexesAccessor[WIid] = (sycl::cl_int)WIid.get(0);});});// Getting read only access to the buffer on the host.// Implicit barrier waiting for queue to complete the work.const auto HostAccessor = Buffer.get_access<sycl::access::mode::read>();// Check the resultsbool MismatchFound = false;for (size_t I = 0; I < Buffer.size(); ++I) {if (HostAccessor[I] != I) {std::cout << "The result is incorrect for element: " << I<< " , expected: " << I << " , got: " << HostAccessor[I]<< std::endl;MismatchFound = true;}}if (!MismatchFound) {std::cout << "The results are correct!" << std::endl;}return MismatchFound;
}

编译

icpx -fsycl -fsycl-targets=nvptx64-nvidia-cuda simple-sycl-app.cpp -o simple-sycl-app

可无视的警告

icpx: warning: CUDA version is newer than the latest supported version 11.8 [-Wunknown-cuda-version]

运行

SYCL_DEVICE_FILTER=cuda SYCL_PI_TRACE=1 ./simple-sycl-app

结果

SYCL_PI_TRACE[basic]: Plugin found and successfully loaded: libpi_cuda.so [ PluginVersion: 11.15.1 ]
SYCL_PI_TRACE[all]: Selected device: -> final score = 1500
SYCL_PI_TRACE[all]:   platform: NVIDIA CUDA BACKEND
SYCL_PI_TRACE[all]:   device: NVIDIA GeForce RTX 2060
The results are correct!