[llvm] [X86] Remove LowerFCanonicalize and use generic expansion (PR #147730)

[via llvm-commits]

woruyu wrote:

for shouldExpandVectorFCANONICALIZEInVectorLegalizer
X86 does not require UnrollVectorOp for vector ISD::FCANONICALIZE — it uses the generic expansion
However, targets like AMDGPU and SystemZ currently rely on UnrollVectorOp to handle vector FCANONICALIZE

https://github.com/llvm/llvm-project/pull/147730