[PATCH] D60150: [DAGCombiner][x86] scalarize splatted vector FP ops

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/include/llvm/CodeGen/TargetLowering.h:2447
+  virtual bool shouldScalarizeBinopSplat(SDValue VecOp,
+                                         unsigned SplatIdx) const {
+    return false;
----------------
Would we gain anything from supporting separate LHS/RHS splat indices? x86 probably not but other targets where larger vector register types often alias smaller/scalar (ARM NEON?)


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.h:1075
+      EVT EltVT = VecOp.getValueType().getScalarType();
+      return (EltVT == MVT::f32 || EltVT == MVT::f64) && !SplatIndex;
+    }
----------------
I'm still amazed we haven't ended up with a isExtractElementCheap helper yet.


================
Comment at: llvm/test/CodeGen/X86/scalarize-fp.ll:481
+  ret <2 x i64> %r
+}
+
----------------
Add these to trunk?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60150/new/

https://reviews.llvm.org/D60150