[Mlir-commits] [mlir] [mlir][linalg] Add FoldConsecutiveScalarMulPattern canonicalization (PR #209697)

Renato Golin llvmlistbot at llvm.org
Fri Aug 21 06:21:37 PDT 2026


================

----------------
rengolin wrote:

Why restrict to multiplication? Any associative op would work.

Also, this is only valid for floating points with `fast-math` flags, otherwise reordering is forbidden.

https://github.com/llvm/llvm-project/pull/209697


More information about the Mlir-commits mailing list