projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 69373a5) | patch
[mlir][gpu] NFC change to pass threadID ops to rewriteOneForeachThreadToGpuThreads
authorThomas Raoux <thomasraoux@google.com>
Tue, 14 Feb 2023 01:14:37 +0000 (01:14 +0000)
committerThomas Raoux <thomasraoux@google.com>
Tue, 14 Feb 2023 01:28:11 +0000 (01:28 +0000)
commit288ae0b92f57cc6fcd77a6e5220e67fba7768ceb
tree7cd22139a0075af4cce35fdd1c0c2d87b27e85bdtree | snapshot
parent69373a5d3f45771cfb1896f9938e2444bedb7b33commit | diff
[mlir][gpu] NFC change to pass threadID ops to rewriteOneForeachThreadToGpuThreads

This allows user to give both the thread ids and dimension of the threads we want to distribute on.
This means we can use it to distribute on warps as well.

Reviewed By: harsh

Differential Revision: https://reviews.llvm.org/D143950
mlir/lib/Dialect/GPU/TransformOps/GPUTransformOps.cpp diff | blob | history
Domain: System / Toolchain;