projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 9657385) | patch
[MLIR] Constant fold multiplies in deriveStaticUpperBound.
authorTres Popp <tpopp@google.com>
Tue, 7 Apr 2020 12:12:04 +0000 (14:12 +0200)
committerTres Popp <tpopp@google.com>
Tue, 14 Apr 2020 11:04:16 +0000 (13:04 +0200)
commit58516718fc66de01f1221c7d30b06c297296a264
tree3c541ea9e6dc131551f376de840fc0b1bcde2ab1tree | snapshot
parent9657385960350150b77ed652175b4c3801abd7facommit | diff
[MLIR] Constant fold multiplies in deriveStaticUpperBound.

Summary:
This operation occurs during collapseParallelLoops, so we constant fold
them also to allow more situations of determining a loop invariant upper
bound when lowering to the GPU dialect from the Loop dialect.

Differential Revision: https://reviews.llvm.org/D77723
mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp diff | blob | history
mlir/test/Conversion/LoopsToGPU/parallel_loop.mlir diff | blob | history
Domain: System / Toolchain;