Open
Description
As of Chapel 1.25.0 we can natively generate kernels that can run on NVIDIA GPUs from order-independent loops and launch them. See the tech note for more details: https://chapel-lang.org/docs/technotes/gpu.html
In future releases we will continue to enhance GPU support so its possible to express additional types of computations, better manage memory, and target other types.
This issue here is to solicit feedback regarding GPU support and to serve as a general landing page linking to to other issues.
Some more pointed discussions we're having about GPU support include:
Activity