[PATCH] D13121: Improve ISel across lane float min/max reduction

James Molloy via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 25 20:04:05 PDT 2015


jmolloy added inline comments.

================
Comment at: lib/Target/AArch64/AArch64ISelLowering.cpp:8601
@@ -8600,3 +8600,3 @@
   int NumVecElts = VTy.getVectorNumElements();
-  if (NumVecElts != 4 && NumVecElts != 8 && NumVecElts != 16)
-    return SDValue();
+  if (Op == ISD::FMAXNUM || Op == ISD::FMINNUM) {
+    if (NumVecElts != 4)
----------------
What about FMAXNAN and FMINNAN (-> FMAXV, FMINV)?

================
Comment at: lib/Target/AArch64/AArch64ISelLowering.cpp:8662
@@ -8656,3 +8661,3 @@
   case ISD::ADD:
-    Opcode = AArch64ISD::UADDV;
+    Opcode = Intrinsic::aarch64_neon_uaddv;
     break;
----------------
Why are you making this change? What's the rationale?


http://reviews.llvm.org/D13121





More information about the llvm-commits mailing list