How to access the GPU in Apple M2 chip when natively compiling to aarch64?

[David Banas]

Hi all,

I'm using a newer GHC version, to natively compile my Haskell code to the Apple M2 chip.

I'd like to access the GPU in the M2. And it seemed like Accelerate was the way to go.

But, it looks like it won't support what I intend.

Does anyone know of a way to access the GPU in the M2 when natively compiling to aarch64?

Thanks,

-db

[Henning Thielemann]

I think Accelerate has only a GPU backend for Nvidia/PTX via LLVM. I guess
other GPUs may be addressed via OpenCL and there was an effort to make an
OpenCL backend for Accelerate:
https://github.com/HIPERFIT/accelerate-opencl

But I am afraid there is nothing that works out of the box.