[all-commits] [llvm/llvm-project] ccc900: [offload] Use pinned memory for KLE

[Robert Imschweiler via All-commits]

  Branch: refs/heads/users/ro-i/kle-pinned
  Home:   https://github.com/llvm/llvm-project
  Commit: ccc9005ed329f1363ce0903df84e4bb650382bda
      https://github.com/llvm/llvm-project/commit/ccc9005ed329f1363ce0903df84e4bb650382bda
  Author: Robert Imschweiler <robert.imschweiler at amd.com>
  Date:   2026-08-03 (Mon, 03 Aug 2026)

  Changed paths:
    M offload/include/Shared/APITypes.h
    M offload/plugins-nextgen/amdgpu/src/rtl.cpp
    M offload/plugins-nextgen/common/include/PluginInterface.h
    M offload/plugins-nextgen/common/src/PluginInterface.cpp
    M offload/unittests/OffloadAPI/kernel/olLaunchKernel.cpp

  Log Message:
  -----------
  [offload] Use pinned memory for KLE

Reduce kernel launch latency by using the fast path "pinned host memory
-> device memory" for submitting the kernel launch environment.

Claude assisted with this patch.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications