[DAGCombiner][X86] Correctly clean up high bits in combinei64TruncSrlAdd

Author: dtcxzyw

llvm/lib/Target/X86/X86ISelLowering.cpp

@@ -53712,23 +53712,20 @@ static SDValue combinei64TruncSrlAdd(SDValue N, EVT VT, SelectionDAG &DAG,
                                  m_ConstInt(SrlConst)))))
     return SDValue();
 
-  if (SrlConst.ule(32) || AddConst.lshr(SrlConst).shl(SrlConst) != AddConst)
+  if (SrlConst.ule(32) || AddConst.countr_zero() < SrlConst.getZExtValue())
     return SDValue();
 
   SDValue AddLHSSrl =
       DAG.getNode(ISD::SRL, DL, MVT::i64, AddLhs, N.getOperand(1));
   SDValue Trunc = DAG.getNode(ISD::TRUNCATE, DL, VT, AddLHSSrl);
 
-  APInt NewAddConstVal =
-      (~((~AddConst).lshr(SrlConst))).trunc(VT.getSizeInBits());
+  APInt NewAddConstVal = AddConst.lshr(SrlConst).trunc(VT.getSizeInBits());
   SDValue NewAddConst = DAG.getConstant(NewAddConstVal, DL, VT);
   SDValue NewAddNode = DAG.getNode(ISD::ADD, DL, VT, Trunc, NewAddConst);
 
-  APInt CleanupSizeConstVal = (SrlConst - 32).zextOrTrunc(VT.getSizeInBits());
   EVT CleanUpVT =
-      EVT::getIntegerVT(*DAG.getContext(), CleanupSizeConstVal.getZExtValue());
-  SDValue CleanUp = DAG.getAnyExtOrTrunc(NewAddNode, DL, CleanUpVT);
-  return DAG.getAnyExtOrTrunc(CleanUp, DL, VT);
+      EVT::getIntegerVT(*DAG.getContext(), 64 - SrlConst.getZExtValue());
+  return DAG.getZeroExtendInReg(NewAddNode, DL, CleanUpVT);
 }
 
 /// Attempt to pre-truncate inputs to arithmetic ops if it will simplify

llvm/test/CodeGen/X86/combine-i64-trunc-srl-add.ll

@@ -7,8 +7,9 @@ define i1 @test_ult_trunc_add(i64 %x) {
 ; X64-LABEL: test_ult_trunc_add:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shrq $48, %rdi
-; X64-NEXT:    addl $-65522, %edi # imm = 0xFFFF000E
-; X64-NEXT:    cmpl $3, %edi
+; X64-NEXT:    addl $14, %edi
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    cmpl $3, %eax
 ; X64-NEXT:    setb %al
 ; X64-NEXT:    retq
   %add = add i64 %x, 3940649673949184
@@ -22,8 +23,9 @@ define i1 @test_ult_add(i64 %x) {
 ; X64-LABEL: test_ult_add:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shrq $48, %rdi
-; X64-NEXT:    addl $-65522, %edi # imm = 0xFFFF000E
-; X64-NEXT:    cmpl $3, %edi
+; X64-NEXT:    addl $14, %edi
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    cmpl $3, %eax
 ; X64-NEXT:    setb %al
 ; X64-NEXT:    retq
   %add = add i64 3940649673949184, %x
@@ -35,8 +37,9 @@ define i1 @test_ugt_trunc_add(i64 %x) {
 ; X64-LABEL: test_ugt_trunc_add:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shrq $48, %rdi
-; X64-NEXT:    addl $-65522, %edi # imm = 0xFFFF000E
-; X64-NEXT:    cmpl $4, %edi
+; X64-NEXT:    addl $14, %edi
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    cmpl $4, %eax
 ; X64-NEXT:    setae %al
 ; X64-NEXT:    retq
   %add = add i64 %x, 3940649673949184
@@ -116,7 +119,8 @@ define i32 @test_trunc_add(i64 %x) {
 ; X64-LABEL: test_trunc_add:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shrq $48, %rdi
-; X64-NEXT:    leal -65522(%rdi), %eax
+; X64-NEXT:    addl $14, %edi
+; X64-NEXT:    movzwl %di, %eax
 ; X64-NEXT:    retq
   %add = add i64 %x, 3940649673949184
   %shr = lshr i64 %add, 48
@@ -157,7 +161,7 @@ define i64 @pr128309(i64 %x) {
 ; X64:       # %bb.0: # %entry
 ; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    andl $18114, %eax # imm = 0x46C2
-; X64-NEXT:    addl $-65530, %eax # imm = 0xFFFF0006
+; X64-NEXT:    addl $6, %eax
 ; X64-NEXT:    andl %edi, %eax
 ; X64-NEXT:    retq
 entry: