[Feature Request] OpenCl Integration like CUDA.

[SohamTilekar]

Feature request

Using OpenCl Gives the Support to Program the non-Nvidia GPU programing.

• Almost Every CPU has Integrated GPU (Intel Integrated Graphs | AMD Integrated GPU)
• You Can Speedup CPU Parallelization Significantly.

You can also do CPU Parallelization Easily.

It Is little bit High-level than CUDA. And it is almost same as CUDA.
And many more Features.

[SohamTilekar]

I Can Help on anything related to OpenCl but I Do not Understand the numba backend that much Otherwise I will make a PR for it.

If Any Help Related to write code in CL then Feel free for the Code.

[diptorupd]

@SohamTilekar Have a look at https://github.com/IntelPython/numba-dpex. Numba-dpex provides an generic kernel programming API that mimics the C++ SYCL embedded domain-specific language. Kernels written in the API can presently be run on OpenCL CPU devices (Intel and AMD) and Intel OpenCL and Level Zero integrated and discrete GPUs.

The architecture of numba-dpex is extensible enough that it should be possible to plug-in support for CUDA devices, but such support is currently missing.