[X86] combineCarryThroughADD - recognise X86ISD::ADD(AND(X,1),-1) pat…

…tern can be folded to X86ISD::BT As mentioned on D122482, if we've generated a masked overflow test see if we can fold it to X86ISD::BT to feed a X86ISD::ADC/SBB Differential Revision: https://reviews.llvm.org/D122572
llvm · Mar 31, 2022 · 481b185 · 481b185
1 parent 2f1261a
commit 481b185
Show file tree

Hide file tree

Showing 4 changed files with 23 additions and 13 deletions.
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -44714,12 +44714,15 @@ static bool checkBoolTestAndOrSetCCCombine(SDValue Cond, X86::CondCode &CC0,
 static SDValue combineCarryThroughADD(SDValue EFLAGS, SelectionDAG &DAG) {
   if (EFLAGS.getOpcode() == X86ISD::ADD) {
     if (isAllOnesConstant(EFLAGS.getOperand(1))) {
+      bool FoundAndLSB = false;
       SDValue Carry = EFLAGS.getOperand(0);
       while (Carry.getOpcode() == ISD::TRUNCATE ||
              Carry.getOpcode() == ISD::ZERO_EXTEND ||
              (Carry.getOpcode() == ISD::AND &&
-              isOneConstant(Carry.getOperand(1))))
+              isOneConstant(Carry.getOperand(1)))) {
+        FoundAndLSB |= Carry.getOpcode() == ISD::AND;
         Carry = Carry.getOperand(0);
+      }
       if (Carry.getOpcode() == X86ISD::SETCC ||
           Carry.getOpcode() == X86ISD::SETCC_CARRY) {
         // TODO: Merge this code with equivalent in combineAddOrSubToADCOrSBB?
@@ -44750,6 +44753,21 @@ static SDValue combineCarryThroughADD(SDValue EFLAGS, SelectionDAG &DAG) {
             CarryOp1.getOpcode() == X86ISD::ADD &&
             isOneConstant(CarryOp1.getOperand(1)))
           return CarryOp1;
+      } else if (FoundAndLSB) {
+        SDLoc DL(Carry);
+        SDValue BitNo = DAG.getConstant(0, DL, Carry.getValueType());
+        if (Carry.getOpcode() == ISD::SRL) {
+          BitNo = Carry.getOperand(1);
+          Carry = Carry.getOperand(0);
+        }
+        if (Carry.getScalarValueSizeInBits() < 32)
+          Carry = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i32, Carry);
+        EVT CarryVT = Carry.getValueType();
+        if (DAG.getTargetLoweringInfo().isTypeLegal(CarryVT)) {
+          if (CarryVT != BitNo.getValueType())
+            BitNo = DAG.getNode(ISD::ANY_EXTEND, DL, CarryVT, BitNo);
+          return DAG.getNode(X86ISD::BT, DL, MVT::i32, Carry, BitNo);
+        }
       }
     }
   }

diff --git a/llvm/test/CodeGen/X86/addcarry.ll b/llvm/test/CodeGen/X86/addcarry.ll
@@ -640,8 +640,7 @@ define { i64, i1 } @addcarry_fake_carry(i64 %a, i64 %b, i1 %carryin) {
 ; CHECK-LABEL: addcarry_fake_carry:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    movq %rdi, %rax
-; CHECK-NEXT:    andb $1, %dl
-; CHECK-NEXT:    addb $-1, %dl
+; CHECK-NEXT:    btl $0, %edx
 ; CHECK-NEXT:    adcq %rsi, %rax
 ; CHECK-NEXT:    setb %dl
 ; CHECK-NEXT:    retq

diff --git a/llvm/test/CodeGen/X86/combine-adc.ll b/llvm/test/CodeGen/X86/combine-adc.ll
@@ -67,24 +67,18 @@ define i32 @PR40483_add2(i32*, i32) nounwind {
   ret i32 %10
 }
 
-; FIXME: Fail to convert add+lshr+and to BT
 define i32 @adc_merge_constants(i32 %a0) nounwind {
 ; X86-LABEL: adc_merge_constants:
 ; X86:       # %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    shrl $11, %ecx
-; X86-NEXT:    andb $1, %cl
 ; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    addb $-1, %cl
+; X86-NEXT:    btl $11, {{[0-9]+}}(%esp)
 ; X86-NEXT:    adcl $54, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: adc_merge_constants:
 ; X64:       # %bb.0:
-; X64-NEXT:    shrl $11, %edi
-; X64-NEXT:    andb $1, %dil
 ; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    addb $-1, %dil
+; X64-NEXT:    btl $11, %edi
 ; X64-NEXT:    adcl $54, %eax
 ; X64-NEXT:    retq
   %bit = lshr i32 %a0, 11

diff --git a/llvm/test/CodeGen/X86/subcarry.ll b/llvm/test/CodeGen/X86/subcarry.ll
@@ -349,8 +349,7 @@ define { i64, i1 } @subcarry_fake_carry(i64 %a, i64 %b, i1 %carryin) {
 ; CHECK-LABEL: subcarry_fake_carry:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    movq %rdi, %rax
-; CHECK-NEXT:    andb $1, %dl
-; CHECK-NEXT:    addb $-1, %dl
+; CHECK-NEXT:    btl $0, %edx
 ; CHECK-NEXT:    sbbq %rsi, %rax
 ; CHECK-NEXT:    setb %dl
 ; CHECK-NEXT:    retq