[llvm] [AArch64] Fold NEON splats into users by using SVE immediates (PR #165559)
Paul Walker via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 11 05:23:31 PDT 2026
================
@@ -15535,6 +15601,9 @@ SDValue AArch64TargetLowering::LowerBUILD_VECTOR(SDValue Op,
return Op;
}
+ if (SDValue V = tryFoldSplatIntoUsersWithSVE(Op, DAG))
----------------
paulwalker-arm wrote:
Thanks, this is looking much better now.
FYI: I'm just finishing of a patch to clean up the NEON side so that we can remove the many indirections when lowering vector constants, which should make this patch simpler. One doesn't have to gate the other because if this lands first then I'll just refactoring it at the same time as the NEON changes.
https://github.com/llvm/llvm-project/pull/165559
More information about the llvm-commits
mailing list