Notes for compiling and running OpenCL code on Intel CPUs

1. Install Intel OneApi basekit

2. Set up environment by running

source /opt/intel/oneapi/setvars.sh

This sets up paths and runtime environment for OpenCL execution. For example, the following paths are added. The compiler

/opt/intel/oneapi/compiler/latest/linux/bin/dpcpp

the includes in

/opt/intel/oneapi/compiler/latest/linux/include/sycl/

and libs in

/opt/intel/oneapi/compiler/latest/linux/lib/

and

/opt/intel/oneapi/compiler/latest/linux/compiler/lib/intel64_lin/

3. Clone repo and OpenCL-CLHPP headers for OpenCL C++ API, then update CPATH

git clone --recurse-submodules https://github.com/hokkanen/opencl.git
export CPATH=$CPATH:$(pwd)/opencl/OpenCL-CLHPP/include

4. Compile examples with

dpcpp cl_devices.c -lOpenCL 

Alternatively, if you have CUDA installed, OpenCL code can easily be compiled with nvcc by

nvcc -arch=sm_80 cl_devices.c -lOpenCL 

where -arch=sm_80 is the target architecture (NVIDIA A100 GPU).

5. Run examples with

./a.out