X86: Remove redundant test instructions

Increase the number of instructions LLVM recognizes as setting the ZF
flag. This allows us to remove test instructions that redundantly
recalculate the flag.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181937 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 193 insertions, 7 deletions

diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp
index 7c0423f818..582add7854 100644
--- a/lib/Target/X86/X86InstrInfo.cpp
+++ b/lib/Target/X86/X86InstrInfo.cpp
@@ -3200,8 +3200,37 @@ inline static bool isDefConvertible(MachineInstr *MI) {
   case X86::OR8ri:     case X86::OR64rr:   case X86::OR32rr:
   case X86::OR16rr:    case X86::OR8rr:    case X86::OR64rm:
   case X86::OR32rm:    case X86::OR16rm:   case X86::OR8rm:
+  case X86::NEG8r:     case X86::NEG16r:   case X86::NEG32r: case X86::NEG64r:
+  case X86::SAR8r1:    case X86::SAR16r1:  case X86::SAR32r1:case X86::SAR64r1:
+  case X86::SHR8r1:    case X86::SHR16r1:  case X86::SHR32r1:case X86::SHR64r1:
+  case X86::SHL8r1:    case X86::SHL16r1:  case X86::SHL32r1:case X86::SHL64r1:
+  case X86::ADC32ri:   case X86::ADC32ri8:
+  case X86::ADC32rr:   case X86::ADC64ri32:
+  case X86::ADC64ri8:  case X86::ADC64rr:
+  case X86::SBB32ri:   case X86::SBB32ri8:
+  case X86::SBB32rr:   case X86::SBB64ri32:
+  case X86::SBB64ri8:  case X86::SBB64rr:
   case X86::ANDN32rr:  case X86::ANDN32rm:
   case X86::ANDN64rr:  case X86::ANDN64rm:
+  case X86::BEXTR32rr: case X86::BEXTR64rr:
+  case X86::BEXTR32rm: case X86::BEXTR64rm:
+  case X86::BLSI32rr:  case X86::BLSI32rm:
+  case X86::BLSI64rr:  case X86::BLSI64rm:
+  case X86::BLSMSK32rr:case X86::BLSMSK32rm:
+  case X86::BLSMSK64rr:case X86::BLSMSK64rm:
+  case X86::BLSR32rr:  case X86::BLSR32rm:
+  case X86::BLSR64rr:  case X86::BLSR64rm:
+  case X86::BZHI32rr:  case X86::BZHI32rm:
+  case X86::BZHI64rr:  case X86::BZHI64rm:
+  case X86::LZCNT16rr: case X86::LZCNT16rm:
+  case X86::LZCNT32rr: case X86::LZCNT32rm:
+  case X86::LZCNT64rr: case X86::LZCNT64rm:
+  case X86::POPCNT16rr:case X86::POPCNT16rm:
+  case X86::POPCNT32rr:case X86::POPCNT32rm:
+  case X86::POPCNT64rr:case X86::POPCNT64rm:
+  case X86::TZCNT16rr: case X86::TZCNT16rm:
+  case X86::TZCNT32rr: case X86::TZCNT32rm:
+  case X86::TZCNT64rr: case X86::TZCNT64rm:
     return true;
   }
 }
@@ -3427,13 +3456,16 @@ optimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg, unsigned SrcReg2,
   }
 
   // Make sure Sub instruction defines EFLAGS and mark the def live.
-  unsigned LastOperand = Sub->getNumOperands() - 1;
-  assert(Sub->getNumOperands() >= 2 &&
-         Sub->getOperand(LastOperand).isReg() &&
-         Sub->getOperand(LastOperand).getReg() == X86::EFLAGS &&
-         "EFLAGS should be the last operand of SUB, ADD, OR, XOR, AND");
-  Sub->getOperand(LastOperand).setIsDef(true);
-  Sub->getOperand(LastOperand).setIsDead(false);
+  unsigned i = 0, e = Sub->getNumOperands();
+  for (; i != e; ++i) {
+    MachineOperand &MO = Sub->getOperand(i);
+    if (MO.isReg() && MO.isDef() && MO.getReg() == X86::EFLAGS) {
+      MO.setIsDead(false);
+      break;
+    }
+  }
+  assert(i != e && "Unable to locate a def EFLAGS operand");
+
   CmpInstr->eraseFromParent();
 
   // Modify the condition code of instructions in OpsToUpdate.
diff --git a/test/CodeGen/X86/peep-test-4.ll b/test/CodeGen/X86/peep-test-4.ll
new file mode 100644
index 0000000000..91c37fbffb
--- /dev/null
+++ b/test/CodeGen/X86/peep-test-4.ll
@@ -0,0 +1,154 @@
+; RUN: llc < %s -march=x86-64 -mattr=+bmi,+bmi2,+popcnt | FileCheck %s
+declare void @foo(i32)
+
+; CHECK: neg:
+; CHECK: negl %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+define void @neg(i32 %x) nounwind {
+  %sub = sub i32 0, %x
+  %cmp = icmp eq i32 %sub, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %sub)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: sar:
+; CHECK: sarl %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+define void @sar(i32 %x) nounwind {
+  %ashr = ashr i32 %x, 1
+  %cmp = icmp eq i32 %ashr, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %ashr)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: shr:
+; CHECK: shrl %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+define void @shr(i32 %x) nounwind {
+  %ashr = lshr i32 %x, 1
+  %cmp = icmp eq i32 %ashr, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %ashr)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: shl:
+; CHECK: addl %edi, %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+define void @shl(i32 %x) nounwind {
+  %shl = shl i32 %x, 1
+  %cmp = icmp eq i32 %shl, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %shl)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: adc:
+; CHECK: movabsq $-9223372036854775808, %rax
+; CHECK-NEXT: addq  %rdi, %rax
+; CHECK-NEXT: adcq  $0, %rsi
+; CHECK-NEXT: sete  %al
+; CHECK-NEXT: ret
+define zeroext i1 @adc(i128 %x) nounwind {
+  %add = add i128 %x, 9223372036854775808
+  %cmp = icmp ult i128 %add, 18446744073709551616
+  ret i1 %cmp
+}
+
+; CHECK: sbb:
+; CHECK: cmpq  %rdx, %rdi
+; CHECK-NEXT: sbbq  %rcx, %rsi
+; CHECK-NEXT: setns %al
+; CHECK-NEXT: ret
+define zeroext i1 @sbb(i128 %x, i128 %y) nounwind {
+  %sub = sub i128 %x, %y
+  %cmp = icmp sge i128 %sub, 0
+  ret i1 %cmp
+}
+
+; CHECK: andn:
+; CHECK: andnl   %esi, %edi, %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+define void @andn(i32 %x, i32 %y) nounwind {
+  %not = xor i32 %x, -1
+  %andn = and i32 %y, %not
+  %cmp = icmp eq i32 %andn, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %andn)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: bextr:
+; CHECK: bextrl   %esi, %edi, %edi
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+declare i32 @llvm.x86.bmi.bextr.32(i32, i32) nounwind readnone
+define void @bextr(i32 %x, i32 %y) nounwind {
+  %bextr = tail call i32 @llvm.x86.bmi.bextr.32(i32 %x, i32 %y)
+  %cmp = icmp eq i32 %bextr, 0
+  br i1 %cmp, label %return, label %bb
+
+bb:
+  tail call void @foo(i32 %bextr)
+  br label %return
+
+return:
+  ret void
+}
+
+; CHECK: popcnt:
+; CHECK: popcntl
+; CHECK-NEXT: je
+; CHECK: jmp foo
+; CHECK: ret
+declare i32 @llvm.ctpop.i32(i32) nounwind readnone
+define void @popcnt(i32 %x) nounwind {
+  %popcnt = tail call i32 @llvm.ctpop.i32(i32 %x)
+  %cmp = icmp eq i32 %popcnt, 0
+  br i1 %cmp, label %return, label %bb
+;
+bb:
+  tail call void @foo(i32 %popcnt)
+  br label %return
+;
+return:
+  ret void
+}