Add some cases to x86 OptimizeCompare to handle DEC and INC, too.

While we are setting the earlier def to true, also make it live.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@164056 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp
index 4f3d824b4a7638c6cbeeb5fd279e09eae6ff91a0..4a9be39fa51cea1e82376120ac6ef784088ce21b 100644 (file)
--- a/lib/Target/X86/X86InstrInfo.cpp
+++ b/lib/Target/X86/X86InstrInfo.cpp
@@ -3127,11 +3127,19 @@ inline static bool isDefConvertible(MachineInstr *MI) {
   case X86::SUB8ri:    case X86::SUB64rr:  case X86::SUB32rr:
   case X86::SUB16rr:   case X86::SUB8rr:   case X86::SUB64rm:
   case X86::SUB32rm:   case X86::SUB16rm:  case X86::SUB8rm:
+  case X86::DEC64r:  case X86::DEC32r:  case X86::DEC16r: case X86::DEC8r:
+  case X86::DEC64m:  case X86::DEC32m:  case X86::DEC16m: case X86::DEC8m:
+  case X86::DEC64_32r: case X86::DEC64_16r:
+  case X86::DEC64_32m: case X86::DEC64_16m:
   case X86::ADD64ri32: case X86::ADD64ri8: case X86::ADD32ri:
   case X86::ADD32ri8:  case X86::ADD16ri:  case X86::ADD16ri8:
   case X86::ADD8ri:    case X86::ADD64rr:  case X86::ADD32rr:
   case X86::ADD16rr:   case X86::ADD8rr:   case X86::ADD64rm:
   case X86::ADD32rm:   case X86::ADD16rm:  case X86::ADD8rm:
+  case X86::INC64r:  case X86::INC32r:  case X86::INC16r: case X86::INC8r:
+  case X86::INC64m:  case X86::INC32m:  case X86::INC16m: case X86::INC8m:
+  case X86::INC64_32r: case X86::INC64_16r:
+  case X86::INC64_32m: case X86::INC64_16m:
   case X86::AND64ri32: case X86::AND64ri8: case X86::AND32ri:
   case X86::AND32ri8:  case X86::AND16ri:  case X86::AND16ri8:
   case X86::AND8ri:    case X86::AND64rr:  case X86::AND32rr:
@@ -3371,12 +3379,14 @@ optimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg, unsigned SrcReg2,
     Sub->getParent()->insert(MachineBasicBlock::iterator(Sub), Movr0Inst);
   }
 
-  // Make sure Sub instruction defines EFLAGS.
+  // Make sure Sub instruction defines EFLAGS and mark the def live.
+  unsigned LastOperand = Sub->getNumOperands() - 1;
   assert(Sub->getNumOperands() >= 2 &&
-         Sub->getOperand(Sub->getNumOperands()-1).isReg() &&
-         Sub->getOperand(Sub->getNumOperands()-1).getReg() == X86::EFLAGS &&
+         Sub->getOperand(LastOperand).isReg() &&
+         Sub->getOperand(LastOperand).getReg() == X86::EFLAGS &&
          "EFLAGS should be the last operand of SUB, ADD, OR, XOR, AND");
-  Sub->getOperand(Sub->getNumOperands()-1).setIsDef(true);
+  Sub->getOperand(LastOperand).setIsDef(true);
+  Sub->getOperand(LastOperand).setIsDead(false);
   CmpInstr->eraseFromParent();
 
   // Modify the condition code of instructions in OpsToUpdate.

diff --git a/test/CodeGen/X86/jump_sign.ll b/test/CodeGen/X86/jump_sign.ll
index 48e21061d20997bd7033e6a3f34bb36011083bce..0cf397d4af050248026f82a19348ed6d73fd9426 100644 (file)
--- a/test/CodeGen/X86/jump_sign.ll
+++ b/test/CodeGen/X86/jump_sign.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 -mcpu=pentiumpro | FileCheck %s
+; RUN: llc < %s -march=x86 -mcpu=pentiumpro -verify-machineinstrs | FileCheck %s
 
 define i32 @f(i32 %X) {
 entry:
@@ -253,3 +253,28 @@ return:
   %retval.0 = phi i8* [ %add.ptr, %if.end ], [ null, %entry ]
   ret i8* %retval.0
 }
+
+; Test optimizations of dec/inc.
+define i32 @dec(i32 %a) nounwind {
+entry:
+; CHECK: dec:
+; CHECK: decl
+; CHECK-NOT: test
+; CHECK: cmovsl
+  %sub = sub nsw i32 %a, 1
+  %cmp = icmp sgt i32 %sub, 0
+  %cond = select i1 %cmp, i32 %sub, i32 0
+  ret i32 %cond
+}
+
+define i32 @inc(i32 %a) nounwind {
+entry:
+; CHECK: inc:
+; CHECK: incl
+; CHECK-NOT: test
+; CHECK: cmovsl
+  %add = add nsw i32 %a, 1
+  %cmp = icmp sgt i32 %add, 0
+  %cond = select i1 %cmp, i32 %add, i32 0
+  ret i32 %cond
+}