From f3ff6ae06652d2755906e576a102a740f11d5425 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@gmail.com>
Date: Mon, 17 Dec 2012 05:12:30 +0000
Subject: [PATCH] Simplify BMI ANDN matching to use patterns instead of a DAG
 combine. Also add ANDN to isDefConvertible.

llvm-svn: 170305
---
 llvm/lib/Target/X86/X86ISelLowering.cpp   | 10 +---------
 llvm/lib/Target/X86/X86ISelLowering.h     |  2 --
 llvm/lib/Target/X86/X86InstrArithmetic.td | 15 +++++++++++++--
 llvm/lib/Target/X86/X86InstrInfo.cpp      |  2 ++
 4 files changed, 16 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 70089fcc6588..34ca24f6d103 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -12026,7 +12026,6 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::OR:                 return "X86ISD::OR";
   case X86ISD::XOR:                return "X86ISD::XOR";
   case X86ISD::AND:                return "X86ISD::AND";
-  case X86ISD::ANDN:               return "X86ISD::ANDN";
   case X86ISD::BLSI:               return "X86ISD::BLSI";
   case X86ISD::BLSMSK:             return "X86ISD::BLSMSK";
   case X86ISD::BLSR:               return "X86ISD::BLSR";
@@ -15632,7 +15631,7 @@ static SDValue PerformAndCombine(SDNode *N, SelectionDAG &DAG,
 
   EVT VT = N->getValueType(0);
 
-  // Create ANDN, BLSI, and BLSR instructions
+  // Create BLSI, and BLSR instructions
   // BLSI is X & (-X)
   // BLSR is X & (X-1)
   if (Subtarget->hasBMI() && (VT == MVT::i32 || VT == MVT::i64)) {
@@ -15640,13 +15639,6 @@ static SDValue PerformAndCombine(SDNode *N, SelectionDAG &DAG,
     SDValue N1 = N->getOperand(1);
     DebugLoc DL = N->getDebugLoc();
 
-    // Check LHS for not
-    if (N0.getOpcode() == ISD::XOR && isAllOnes(N0.getOperand(1)))
-      return DAG.getNode(X86ISD::ANDN, DL, VT, N0.getOperand(0), N1);
-    // Check RHS for not
-    if (N1.getOpcode() == ISD::XOR && isAllOnes(N1.getOperand(1)))
-      return DAG.getNode(X86ISD::ANDN, DL, VT, N1.getOperand(0), N0);
-
     // Check LHS for neg
     if (N0.getOpcode() == ISD::SUB && N0.getOperand(1) == N1 &&
         isZero(N0.getOperand(0)))
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 60f34099dde8..c51460bc42ee 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -273,8 +273,6 @@ namespace llvm {
       ADD, SUB, ADC, SBB, SMUL,
       INC, DEC, OR, XOR, AND,
 
-      ANDN, // ANDN - Bitwise AND NOT with FLAGS results.
-
       BLSI,   // BLSI - Extract lowest set isolated bit
       BLSMSK, // BLSMSK - Get mask up to lowest set bit
       BLSR,   // BLSR - Reset lowest set bit
diff --git a/llvm/lib/Target/X86/X86InstrArithmetic.td b/llvm/lib/Target/X86/X86InstrArithmetic.td
index f790611b8f8c..478c42d6575c 100644
--- a/llvm/lib/Target/X86/X86InstrArithmetic.td
+++ b/llvm/lib/Target/X86/X86InstrArithmetic.td
@@ -1204,12 +1204,12 @@ multiclass bmi_andn<string mnemonic, RegisterClass RC, X86MemOperand x86memop,
                     PatFrag ld_frag> {
   def rr : I<0xF2, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
             !strconcat(mnemonic, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-            [(set RC:$dst, EFLAGS, (X86andn_flag RC:$src1, RC:$src2))],
+            [(set RC:$dst, EFLAGS, (X86and_flag (not RC:$src1), RC:$src2))],
             IIC_BIN_NONMEM>;
   def rm : I<0xF2, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
             !strconcat(mnemonic, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
             [(set RC:$dst, EFLAGS,
-             (X86andn_flag RC:$src1, (ld_frag addr:$src2)))], IIC_BIN_MEM>;
+             (X86and_flag (not RC:$src1), (ld_frag addr:$src2)))], IIC_BIN_MEM>;
 }
 
 let Predicates = [HasBMI], Defs = [EFLAGS] in {
@@ -1217,6 +1217,17 @@ let Predicates = [HasBMI], Defs = [EFLAGS] in {
   defm ANDN64 : bmi_andn<"andn{q}", GR64, i64mem, loadi64>, T8, VEX_4V, VEX_W;
 }
 
+let Predicates = [HasBMI] in {
+  def : Pat<(and (not GR32:$src1), GR32:$src2),
+            (ANDN32rr GR32:$src1, GR32:$src2)>;
+  def : Pat<(and (not GR64:$src1), GR64:$src2),
+            (ANDN64rr GR64:$src1, GR64:$src2)>;
+  def : Pat<(and (not GR32:$src1), (loadi32 addr:$src2)),
+            (ANDN32rm GR32:$src1, addr:$src2)>;
+  def : Pat<(and (not GR64:$src1), (loadi64 addr:$src2)),
+            (ANDN64rm GR64:$src1, addr:$src2)>;
+}
+
 //===----------------------------------------------------------------------===//
 // MULX Instruction
 //
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index 962977f6520b..d02e12fc4f8a 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -3201,6 +3201,8 @@ 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::ANDN32rr:  case X86::ANDN32rm:
+  case X86::ANDN64rr:  case X86::ANDN64rm:
     return true;
   }
 }