[Mlir-commits] [mlir] [mlir][gpu] Pattern to promote `gpu.shuffle` to specialized AMDGPU ops (PR #137109)

Krzysztof Drewniak llvmlistbot at llvm.org
Fri May 2 17:01:42 PDT 2025


krzysz00 wrote:

Wanted to ping because, even outside of the subgroup_reduce case, there may be cases where rewriting certain `gpu.shuffle` instances to `ds_swizzle` will be advantageous

https://github.com/llvm/llvm-project/pull/137109


More information about the Mlir-commits mailing list