commit | 11d9f3345ba1f6520acd90a5fbc74d5eb8825b16 | [log] [tgz] |
---|---|---|
author | Joseph Huber <huberjn@outlook.com> | Tue Mar 12 10:39:40 2024 -0500 |
committer | Copybara-Service <copybara-worker@google.com> | Tue Mar 12 08:44:27 2024 -0700 |
tree | 8d6fa6126693ad800b8e76299a84c99d6cd711b9 | |
parent | 49bc29aa11bd050a9f5a6915591ab617f398d8e4 [diff] |
[Libomptarget] Use NVPTX lane id intrinsic in DeviceRTL (#84928) Summary: We are currently taking the lower 5 bites of the thread ID as the warp ID. This doesn't work in non-1D grids and is also slower than just using the dedicated hardware register. GitOrigin-RevId: 9f69d3cf88905df5006f93dce536b7e73c0b1735