5 releases

0.2.3 Jul 9, 2023
0.2.2 Jul 9, 2023
0.2.1 Jul 9, 2023
0.2.0 Jul 9, 2023
0.0.1 May 7, 2023

#379 in Concurrency

MIT license

1.5MB
23K SLoC

this is early but under full time development.

an extremely simple interface to SIMT-style (CUDA, HIP, compute shaders) computation.

current focus is CUDA + ROCm/HIP support.

kernels are written in C++. this makes it easy to port existing code and also saves a lot of headaches dealing with emitting NVVM/AMDGPU code.

Dependencies

~0.6–6MB
~25K SLoC