[SelectionDAG][X86] CombineBT - more aggressively determine demanded bits

This patch is in 2 parts:

1 - replace combineBT's use of SimplifyDemandedBits (hasOneUse only) with SelectionDAG::GetDemandedBits to more aggressively determine the lower bits used by BT.

2 - update SelectionDAG::GetDemandedBits to support ANY_EXTEND - if the demanded bits are only in the non-extended portion, then peek through and demand from the source value and then ANY_EXTEND that if we found a match.

Differential Revision: https://reviews.llvm.org/D35896

llvm-svn: 309486

3 files changed, 21 insertions, 17 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index da69ef3b231..a396f3368cf 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -1999,6 +1999,18 @@ SDValue SelectionDAG::GetDemandedBits(SDValue V, const APInt &Mask) {
       return V.getOperand(0);
     break;
   }
+  case ISD::ANY_EXTEND: {
+    SDValue Src = V.getOperand(0);
+    unsigned SrcBitWidth = Src.getScalarValueSizeInBits();
+    // Being conservative here - only peek through if we only demand bits in the
+    // non-extended source (even though the extended bits are technically undef).
+    if (Mask.getActiveBits() > SrcBitWidth)
+      break;
+    APInt SrcMask = Mask.trunc(SrcBitWidth);
+    if (SDValue DemandedSrc = GetDemandedBits(Src, SrcMask))
+      return getNode(ISD::ANY_EXTEND, SDLoc(V), V.getValueType(), DemandedSrc);
+    break;
+  }
   }
   return SDValue();
 }
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index a0fd72dd6a9..da9adcb50fb 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -34200,19 +34200,15 @@ static SDValue combineAndnp(SDNode *N, SelectionDAG &DAG,
 
 static SDValue combineBT(SDNode *N, SelectionDAG &DAG,
                          TargetLowering::DAGCombinerInfo &DCI) {
+  SDValue N0 = N->getOperand(0);
+  SDValue N1 = N->getOperand(1);
+
   // BT ignores high bits in the bit index operand.
-  SDValue Op1 = N->getOperand(1);
-  if (Op1.hasOneUse()) {
-    unsigned BitWidth = Op1.getValueSizeInBits();
-    APInt DemandedMask = APInt::getLowBitsSet(BitWidth, Log2_32(BitWidth));
-    KnownBits Known;
-    TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
-                                          !DCI.isBeforeLegalizeOps());
-    const TargetLowering &TLI = DAG.getTargetLoweringInfo();
-    if (TLI.ShrinkDemandedConstant(Op1, DemandedMask, TLO) ||
-        TLI.SimplifyDemandedBits(Op1, DemandedMask, Known, TLO))
-      DCI.CommitTargetLoweringOpt(TLO);
-  }
+  unsigned BitWidth = N1.getValueSizeInBits();
+  APInt DemandedMask = APInt::getLowBitsSet(BitWidth, Log2_32(BitWidth));
+  if (SDValue DemandedN1 = DAG.GetDemandedBits(N1, DemandedMask))
+    return DAG.getNode(X86ISD::BT, SDLoc(N), MVT::i32, N0, DemandedN1);
+
   return SDValue();
 }
 
diff --git a/llvm/test/CodeGen/X86/bt.ll b/llvm/test/CodeGen/X86/bt.ll
index 5cfad109541..c8050a33916 100644
--- a/llvm/test/CodeGen/X86/bt.ll
+++ b/llvm/test/CodeGen/X86/bt.ll
@@ -1101,8 +1101,6 @@ define void @demanded_i32(i32* nocapture readonly, i32* nocapture, i32) nounwind
 ; X86-NEXT:    movl (%edx,%eax,4), %esi
 ; X86-NEXT:    movl $1, %edx
 ; X86-NEXT:    shll %cl, %edx
-; X86-NEXT:    andb $31, %cl
-; X86-NEXT:    movzbl %cl, %ecx
 ; X86-NEXT:    btl %ecx, %esi
 ; X86-NEXT:    jae .LBB30_2
 ; X86-NEXT:  # BB#1:
@@ -1120,9 +1118,7 @@ define void @demanded_i32(i32* nocapture readonly, i32* nocapture, i32) nounwind
 ; X64-NEXT:    movl $1, %edi
 ; X64-NEXT:    movl %edx, %ecx
 ; X64-NEXT:    shll %cl, %edi
-; X64-NEXT:    andb $31, %dl
-; X64-NEXT:    movzbl %dl, %ecx
-; X64-NEXT:    btl %ecx, %r8d
+; X64-NEXT:    btl %edx, %r8d
 ; X64-NEXT:    jae .LBB30_2
 ; X64-NEXT:  # BB#1:
 ; X64-NEXT:    orl %edi, (%rsi,%rax,4)