[llvm-branch-commits] [llvm] [AMDGPU] Precommit si-fold-bitmask.mir (PR #131310)

[Pierre van Houtryve via llvm-branch-commits]

Pierre-vh wrote:

> Then isUnneededShiftMask should probably recognize more forms of the pattern

Yes, but with a cast operation it gets tricky. If I add `(zext node:$src0)` to the PatFrag, it doesn't work because the PatFrag is always used like this: `(cshl_32 (i32 $src0))`. As `$src0` is inferred to i32, it proceeds to infer the zext as a zext to i64 and then we get a type mismatch error.
I'm not sure how to proceed with this one, I need someone to help me figure out the right syntax

https://github.com/llvm/llvm-project/pull/131310