diff options
| author | Jakub Kuderski <jakub@nod-labs.com> | 2023-11-22 10:30:04 -0500 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-11-22 10:30:04 -0500 |
| commit | d33bad66d86a6fdb443c59561f9524f451a82db0 (patch) | |
| tree | b681b2657878d295a600104a18dfa54d1263cb63 /llvm/lib/CodeGen/MachineCopyPropagation.cpp | |
| parent | 8c02b34e3b9b1e2596651959ba76c66a7afaf545 (diff) | |
| download | llvm-d33bad66d86a6fdb443c59561f9524f451a82db0.zip llvm-d33bad66d86a6fdb443c59561f9524f451a82db0.tar.gz llvm-d33bad66d86a6fdb443c59561f9524f451a82db0.tar.bz2 | |
[mlir][vector] Add patterns to simplify chained reductions (#73048)
Chained reductions get created during vector unrolling. These patterns
simplify them into a series of adds followed by a final reductions.
This is preferred on GPU targets like SPIR-V/Vulkan where vector
reduction gets lowered into subgroup operations that are generally more
expensive than simple vector additions.
For now, only the `add` combining kind is handled.
Diffstat (limited to 'llvm/lib/CodeGen/MachineCopyPropagation.cpp')
0 files changed, 0 insertions, 0 deletions