[llvm] [AArch64] Lower partial add reduction to udot or svdot (PR #101010)
Paul Walker via llvm-commits
llvm-commits at lists.llvm.org
Tue Aug 27 10:41:27 PDT 2024
================
@@ -21229,6 +21249,101 @@ static SDValue tryCombineWhileLo(SDNode *N,
return SDValue(N, 0);
}
+SDValue tryLowerPartialReductionToDot(SDNode *N,
+ const AArch64Subtarget *Subtarget,
+ SelectionDAG &DAG) {
+
+ if (!Subtarget->isSVEAvailable() && !Subtarget->isNeonAvailable())
----------------
paulwalker-arm wrote:
Given from the above I'd expect this first implementation to be limited to scalable vectors, why do you care about NEON being available. Also, is there anything here that wouldn't working in streaming mode? If not then can you use isSVEorStreamingSVEAvailable?
https://github.com/llvm/llvm-project/pull/101010
More information about the llvm-commits
mailing list