[llvm-branch-commits] [llvm] [offload] Copy the kernel launch environment with a direct-write memcpy (PR #231923)

Robert Imschweiler via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Thu Oct 29 01:45:56 PDT 2026


ro-i wrote:

yes, I already performed them for the earlier versions of this patch (so, the previous PRs). I just verified that the changed design doesn't change anything performance-wise.
So, for these two kernels: https://github.com/ro-i/omp-microtest/blob/main/src/omp_launch_latency.cpp
The kernel without KLE is roughly performance-neutral. The kernel with KLE takes ~40% less time.
Since the current implementation is confined to the AMDGPU plugin (as you preferred), I don't see how other plugins could be negatively affected.

https://github.com/llvm/llvm-project/pull/231923


More information about the llvm-branch-commits mailing list