[TargetLowering] SimplifyDemandedBits - add ANY_EXTEND_VECTOR_INREG support

Add 'lowest' demanded elt -> bitcast fold to all *_EXTEND_VECTOR_INREG cases.

Reapplies rL363856.

llvm-svn: 364311

diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 933d4f3ce9ad20139b089d8462bf94829c3abc54..d83bd108b673ba2bf704273c725e297cbf934fda 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -1415,6 +1415,13 @@ bool TargetLowering::SimplifyDemandedBits(
 
     // If none of the top bits are demanded, convert this into an any_extend.
     if (DemandedBits.getActiveBits() <= InBits) {
+      // If we only need the non-extended bits of the bottom element
+      // then we can just bitcast to the result.
+      if (IsVecInReg && DemandedElts == 1 &&
+          VT.getSizeInBits() == SrcVT.getSizeInBits() &&
+          TLO.DAG.getDataLayout().isLittleEndian())
+        return TLO.CombineTo(Op, TLO.DAG.getBitcast(VT, Src));
+
       unsigned Opc =
           IsVecInReg ? ISD::ANY_EXTEND_VECTOR_INREG : ISD::ANY_EXTEND;
       if (!TLO.LegalOperations() || isOperationLegal(Opc, VT))
@@ -1446,12 +1453,21 @@ bool TargetLowering::SimplifyDemandedBits(
     }
     break;
   }
-  case ISD::ANY_EXTEND: {
-    // TODO: Add ISD::ANY_EXTEND_VECTOR_INREG support.
+  case ISD::ANY_EXTEND:
+  case ISD::ANY_EXTEND_VECTOR_INREG: {
     SDValue Src = Op.getOperand(0);
     EVT SrcVT = Src.getValueType();
     unsigned InBits = SrcVT.getScalarSizeInBits();
     unsigned InElts = SrcVT.isVector() ? SrcVT.getVectorNumElements() : 1;
+    bool IsVecInReg = Op.getOpcode() == ISD::ANY_EXTEND_VECTOR_INREG;
+
+    // If we only need the bottom element then we can just bitcast.
+    // TODO: Handle ANY_EXTEND?
+    if (IsVecInReg && DemandedElts == 1 &&
+        VT.getSizeInBits() == SrcVT.getSizeInBits() &&
+        TLO.DAG.getDataLayout().isLittleEndian())
+      return TLO.CombineTo(Op, TLO.DAG.getBitcast(VT, Src));
+
     APInt InDemandedBits = DemandedBits.trunc(InBits);
     APInt InDemandedElts = DemandedElts.zextOrSelf(InElts);
     if (SimplifyDemandedBits(Src, InDemandedBits, InDemandedElts, Known, TLO,

diff --git a/llvm/test/CodeGen/X86/vselect.ll b/llvm/test/CodeGen/X86/vselect.ll
index 0d43a50330fd8c4351e7d0f73c8eded239e6651b..a8afad6f9bb783d0a71a31f1c3135a20e257ef1f 100644 (file)
--- a/llvm/test/CodeGen/X86/vselect.ll
+++ b/llvm/test/CodeGen/X86/vselect.ll
@@ -663,7 +663,6 @@ define i64 @vselect_any_extend_vector_inreg_crash(<8 x i8>* %x) {
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; SSE41-NEXT:    pcmpeqw {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
 ; SSE41-NEXT:    psllq $56, %xmm0
 ; SSE41-NEXT:    movl $32768, %eax # imm = 0x8000
 ; SSE41-NEXT:    movq %rax, %xmm1