Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
GPU technote updates for 1.32 (chapel-lang#23496)
Important changes: - Reduces defensiveness in the language - But at the same time, adds a `.. warning::` to emphasize unstability of the interface - Creates a "Tested Configurations" at the very end and moves that information there - Moves the "kernels are generated from loops" paragraph into "Overview". - Removes "CUDA as early as 7.x may work" - Adds information about ROCm 5.x - Adjusts the memory strategy default to be `array_on_device` - Points to ROCm definition for correct `CHPL_GPU_ARCH` setting for AMD gpus. The table there is much easier to consume: https://rocm.docs.amd.com/en/latest/release/gpu_os_support.html#linux-supported-gpus - Adds a short section "Chapel Tasks and GPU Execution" [Reviewed by @stonea and @ShreyasKhandekar]
- Loading branch information