[llvm-branch-commits] [llvm] [AArch64][DAG] Improve codegen for clearing upper vector lanes (PR #237259)

via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Thu Nov 5 00:29:06 PST 2026


================
@@ -8759,6 +8759,55 @@ defm : Neon_INS_elt_ext_pattern<v16i8, v8i8, v4i16, scalar_to_vector, INSvi8lane
 defm : Neon_INS_elt_ext_pattern<v16i8, v8i8, v2i32, scalar_to_vector_v2f32, INSvi8lane, DUPi8, bsub, VecIndex_x4>;
 defm : Neon_INS_elt_ext_pattern<v8i16, v4i16, v2i32, scalar_to_vector_v2f32, INSvi16lane, DUPi16, hsub, VecIndex_x2>;
 
+multiclass ExtractInsertLane0IntoZeroVector<ValueType VecVT, ValueType ElemTy> {
----------------
CarolineConcatto wrote:

I think we should multiclass when we have more than 1 pattern and we can re-use for  a couple of  type sizes. It is not a rule, but I think it is a good practice.

https://github.com/llvm/llvm-project/pull/237259


More information about the llvm-branch-commits mailing list