[X86] Simplify X86ISD::ADD/SUB if we don't use the result flag

[RKSimon]

Summary

Simplify to the generic ISD::ADD/SUB if we don't make use of the result flag.

This mainly helps with ADDCARRY/SUBBORROW intrinsics which get expanded to X86ISD::ADD/SUB but could be simplified further.

Noticed in some of the test cases in PR31754

[spatel]

LGTM

lib/Target/X86/X86ISelLowering.cpp

40451

Assert that the incoming opcode is X86ISD::ADD or X86ISD::SUB?

lib/Target/X86/X86ISelLowering.cpp

static SDValue combineCMP(SDNode *N, SelectionDAG &DAG) {
   if (NewOpc == X86ISD::AND)
     return DAG.getNode(X86ISD::CMP, dl, MVT::i32, Op,
                        DAG.getConstant(0, dl, VT));
 
   // Return the flags.
   return Op.getValue(1);
 }
 
+static SDValue combineX86AddSub(SDNode *N, SelectionDAG &DAG) {
+  // If we don't use the flag result, simplify back to a simple ADD/SUB.
+  if (N->hasAnyUseOfValue(1))
+    return SDValue();
+
+  SDLoc DL(N);
+  SDValue LHS = N->getOperand(0);
+  SDValue RHS = N->getOperand(1);
+  SDValue Res = DAG.getNode(X86ISD::ADD == N->getOpcode() ? ISD::ADD : ISD::SUB,
+                            DL, LHS.getSimpleValueType(), LHS, RHS);
+  return DAG.getMergeValues({Res, DAG.getConstant(0, DL, MVT::i32)}, DL);
+}
+
 static SDValue combineSBB(SDNode *N, SelectionDAG &DAG) {
   if (SDValue Flags = combineCarryThroughADD(N->getOperand(2))) {
     MVT VT = N->getSimpleValueType(0);
     SDVTList VTs = DAG.getVTList(VT, MVT::i32);
     return DAG.getNode(X86ISD::SBB, SDLoc(N), VTs,
                        N->getOperand(0), N->getOperand(1),
                        Flags);
   }
SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
   case ISD::VSELECT:
   case ISD::SELECT:
   case X86ISD::BLENDV:      return combineSelect(N, DAG, DCI, Subtarget);
   case ISD::BITCAST:        return combineBitcast(N, DAG, DCI, Subtarget);
   case X86ISD::CMOV:        return combineCMov(N, DAG, DCI, Subtarget);
   case X86ISD::CMP:         return combineCMP(N, DAG);
   case ISD::ADD:            return combineAdd(N, DAG, Subtarget);
   case ISD::SUB:            return combineSub(N, DAG, Subtarget);
+  case X86ISD::ADD:
+  case X86ISD::SUB:         return combineX86AddSub(N, DAG);
   case X86ISD::SBB:         return combineSBB(N, DAG);
   case X86ISD::ADC:         return combineADC(N, DAG, DCI);
   case ISD::MUL:            return combineMul(N, DAG, DCI, Subtarget);
   case ISD::SHL:
   case ISD::SRA:
   case ISD::SRL:            return combineShift(N, DAG, DCI, Subtarget);
   case ISD::AND:            return combineAnd(N, DAG, DCI, Subtarget);
   case ISD::OR:             return combineOr(N, DAG, DCI, Subtarget);

test/CodeGen/X86/adx-intrinsics-upgrade.ll

; CHECK-NEXT:    retq ## encoding: [0xc3]
   %conv = zext i8 %4 to i32
   ret i32 %conv
 }
 
 ; Try a really simple all zero input case, which also used to crash
 define void @allzeros() {
 ; CHECK-LABEL: allzeros:
 ; CHECK:       ## %bb.0: ## %entry
-; CHECK-NEXT:    xorl %eax, %eax ## encoding: [0x31,0xc0]
-; CHECK-NEXT:    addq $0, %rax ## encoding: [0x48,0x83,0xc0,0x00]
-; CHECK-NEXT:    movq %rax, 0 ## encoding: [0x48,0x89,0x04,0x25,0x00,0x00,0x00,0x00]
+; CHECK-NEXT:    movq $0, 0 ## encoding: [0x48,0xc7,0x04,0x25,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
 entry:
   %0 = tail call i8 @llvm.x86.addcarryx.u64(i8 0, i64 0, i64 0, i8* null)
   ret void
 }

test/CodeGen/X86/adx-intrinsics.ll

; CHECK-NEXT:    retq ## encoding: [0xc3]
   %conv = zext i8 %7 to i32
   ret i32 %conv
 }
 
 ; Try a really simple all zero input case, which also used to crash
 define void @allzeros() {
 ; CHECK-LABEL: allzeros:
 ; CHECK:       ## %bb.0: ## %entry
-; CHECK-NEXT:    xorl %eax, %eax ## encoding: [0x31,0xc0]
-; CHECK-NEXT:    addq $0, %rax ## encoding: [0x48,0x83,0xc0,0x00]
-; CHECK-NEXT:    movq %rax, 0 ## encoding: [0x48,0x89,0x04,0x25,0x00,0x00,0x00,0x00]
+; CHECK-NEXT:    movq $0, 0 ## encoding: [0x48,0xc7,0x04,0x25,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
 entry:
   %0 = call { i8, i64 } @llvm.x86.addcarry.64(i8 0, i64 0, i64 0)
   %1 = extractvalue { i8, i64 } %0, 1
   store i64 %1, i64* null, align 1
   %2 = extractvalue { i8, i64 } %0, 0
   ret void
 }

test/CodeGen/X86/combine-adx.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown | FileCheck %s --check-prefixes=X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=X64
 
 define i32 @test_addcarry_32_x_0_false(i32 %a0)  {
 ; X86-LABEL: test_addcarry_32_x_0_false:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    addl $0, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: test_addcarry_32_x_0_false:
 ; X64:       # %bb.0:
-; X64-NEXT:    # kill: def $edi killed $edi def $rdi
-; X64-NEXT:    leal (%rdi), %eax
+; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    retq
   %1 = tail call { i8, i32 } @llvm.x86.addcarry.32(i8 0, i32 %a0, i32 0)
   %2 = extractvalue { i8, i32 } %1, 1
   ret i32 %2
 }
 
 define i32 @test_addcarry_32_0_x_false(i32 %a0)  {
 ; X86-LABEL: test_addcarry_32_0_x_false:
 ; X86:       # %bb.0:
-; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: test_addcarry_32_0_x_false:
 ; X64:       # %bb.0:
-; X64-NEXT:    # kill: def $edi killed $edi def $rdi
-; X64-NEXT:    leal (%rdi), %eax
+; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    retq
   %1 = tail call { i8, i32 } @llvm.x86.addcarry.32(i8 0, i32 0, i32 %a0)
   %2 = extractvalue { i8, i32 } %1, 1
   ret i32 %2
 }
 
 define i32 @test_subborrow_32_x_0_false(i32 %a0)  {
 ; X86-LABEL: test_subborrow_32_x_0_false:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    subl $0, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: test_subborrow_32_x_0_false:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    subl $0, %eax
 ; X64-NEXT:    retq
   %1 = tail call { i8, i32 } @llvm.x86.subborrow.32(i8 0, i32 %a0, i32 0)
   %2 = extractvalue { i8, i32 } %1, 1
   ret i32 %2
 }
 
 define i32 @test_subborrow_32_0_x_false(i32 %a0)  {
 ; X86-LABEL: test_subborrow_32_0_x_false: