[X86] Pre-truncate gather/scatter indices that have element sizes larger than 64-bits before Legalize.

The wider element type will normally cause legalize to try to split and scalarize the gather/scatter, but we can't handle that. Instead, truncate the index early so the gather/scatter node is insulated from the legalization.

This really shouldn't happen in practice since InstCombine will normalize index types to the same size as pointers.

llvm-svn: 318452

1 files changed, 19 insertions, 2 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 67420fe0ea3..5c2e13134df 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -35829,8 +35829,25 @@ static SDValue combineSetCC(SDNode *N, SelectionDAG &DAG,
   return SDValue();
 }
 
-static SDValue combineGatherScatter(SDNode *N, SelectionDAG &DAG) {
+static SDValue combineGatherScatter(SDNode *N, SelectionDAG &DAG,
+                                    TargetLowering::DAGCombinerInfo &DCI) {
   SDLoc DL(N);
+
+  // Pre-shrink oversized index elements to avoid triggering scalarization.
+  if (DCI.isBeforeLegalize()) {
+    SDValue Index = N->getOperand(4);
+    if (Index.getValueType().getScalarSizeInBits() > 64) {
+      EVT IndexVT = EVT::getVectorVT(*DAG.getContext(), MVT::i64,
+                                   Index.getValueType().getVectorNumElements());
+      SDValue Trunc = DAG.getNode(ISD::TRUNCATE, DL, IndexVT, Index);
+      SmallVector<SDValue, 5> NewOps(N->op_begin(), N->op_end());
+      NewOps[4] = Trunc;
+      DAG.UpdateNodeOperands(N, NewOps);
+      DCI.AddToWorklist(N);
+      return SDValue(N, 0);
+    }
+  }
+
   // Gather and Scatter instructions use k-registers for masks. The type of
   // the masks is v*i1. So the mask will be truncated anyway.
   // The SIGN_EXTEND_INREG my be dropped.
@@ -36949,7 +36966,7 @@ SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
   case X86ISD::FMADDSUB:
   case X86ISD::FMSUBADD:    return combineFMADDSUB(N, DAG, Subtarget);
   case ISD::MGATHER:
-  case ISD::MSCATTER:       return combineGatherScatter(N, DAG);
+  case ISD::MSCATTER:       return combineGatherScatter(N, DAG, DCI);
   case X86ISD::TESTM:       return combineTestM(N, DAG, Subtarget);
   case X86ISD::PCMPEQ:
   case X86ISD::PCMPGT:      return combineVectorCompare(N, DAG, Subtarget);