diff --git a/lib/Target/Sparc/InstSelectSimple.cpp b/lib/Target/Sparc/InstSelectSimple.cpp
index a73d97b..3bae6a5 100644
--- a/lib/Target/Sparc/InstSelectSimple.cpp
+++ b/lib/Target/Sparc/InstSelectSimple.cpp
@@ -288,33 +288,53 @@
   unsigned Op1Reg = getReg (I.getOperand (1));
 
   unsigned ResultReg = makeAnotherReg (I.getType ());
-  
+  unsigned OpCase = ~0;
+
   // FIXME: support long, ulong, fp.
   switch (I.getOpcode ()) {
-    case Instruction::Add: 
-      BuildMI (BB, V8::ADDrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Sub: 
-      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Mul: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SMULrr : V8::UMULrr;
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+  case Instruction::Add: OpCase = 0; break;
+  case Instruction::Sub: OpCase = 1; break;
+  case Instruction::Mul: OpCase = 2; break;
+  case Instruction::And: OpCase = 3; break;
+  case Instruction::Or:  OpCase = 4; break;
+  case Instruction::Xor: OpCase = 5; break;
+
+  case Instruction::Div:
+  case Instruction::Rem: {
+    unsigned Dest = ResultReg;
+    if (I.getOpcode() == Instruction::Rem)
+      Dest = makeAnotherReg(I.getType());
+
+    // FIXME: this is probably only right for 32 bit operands.
+    if (I.getType ()->isSigned()) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      // Sign extend into the Y register
+      BuildMI (BB, V8::SRAri, 2, Tmp).addReg (Op0Reg).addZImm (31);
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (Tmp).addReg (V8::G0);
+      BuildMI (BB, V8::SDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
+    } else {
+      // Zero extend into the Y register, ie, just set it to zero
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
+      BuildMI (BB, V8::UDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
     }
-    case Instruction::Div: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SDIVrr : V8::UDIVrr;
-      // Clear out the Y register (top half of LHS of divide) 
-      BuildMI (BB, V8::WRYrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
-      BuildMI (BB, V8::NOP, 0); // WR may take up to 4 cycles to finish
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+
+    if (I.getOpcode() == Instruction::Rem) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      BuildMI (BB, V8::SMULrr, 2, Tmp).addReg(Dest).addReg(Op1Reg);
+      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg(Op0Reg).addReg(Tmp);
     }
-    default:
-      visitInstruction (I);
-      return;
+    break;
+  }
+  default:
+    visitInstruction (I);
+    return;
+  }
+
+  if (OpCase != ~0U) {
+    static const unsigned Opcodes[] = {
+      V8::ADDrr, V8::SUBrr, V8::SMULrr, V8::ANDrr, V8::ORrr, V8::XORrr
+    };
+    BuildMI (BB, Opcodes[OpCase], 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
   }
 
   switch (getClass (I.getType ())) {
diff --git a/lib/Target/Sparc/SparcInstrInfo.td b/lib/Target/Sparc/SparcInstrInfo.td
index db79597..ae4e9e6 100644
--- a/lib/Target/Sparc/SparcInstrInfo.td
+++ b/lib/Target/Sparc/SparcInstrInfo.td
@@ -75,9 +75,12 @@
   def NOP : F2_1<0b100, "nop">;
 
 // Section B.11 - Logical Instructions, p. 106
+def ANDrr : F3_1<2, 0b000001, "and">;
 def ANDri : F3_2<2, 0b000001, "and">;
 def ORrr  : F3_1<2, 0b000010, "or">;
 def ORri  : F3_2<2, 0b000010, "or">;
+def XORrr : F3_1<2, 0b000011, "xor">;
+def XORri : F3_2<2, 0b000011, "xor">;
 
 // Section B.12 - Shift Instructions, p. 107
 def SLLri : F3_1<2, 0b100101, "sll">;
@@ -95,8 +98,14 @@
 def SMULrr : F3_1<2, 0b001011, "smul">;
 
 // Section B.19 - Divide Instructions, p. 115
-def UDIVrr: F3_1<2, 0b001110, "udiv">;
-def SDIVrr: F3_1<2, 0b001111, "sdiv">;
+def UDIVrr   : F3_1<2, 0b001110, "udiv">;
+def UDIVri   : F3_2<2, 0b001110, "udiv">;
+def SDIVrr   : F3_1<2, 0b001111, "sdiv">;
+def SDIVri   : F3_2<2, 0b001111, "sdiv">;
+def UDIVCCrr : F3_1<2, 0b011110, "udivcc">;
+def UDIVCCri : F3_2<2, 0b011110, "udivcc">;
+def SDIVCCrr : F3_1<2, 0b011111, "sdivcc">;
+def SDIVCCri : F3_2<2, 0b011111, "sdivcc">;
 
 // Section B.20 - SAVE and RESTORE, p. 117
 def SAVErr    : F3_1<2, 0b111100, "save">;           // save    r, r, r
@@ -118,7 +127,6 @@
 def JMPLrr : F3_1<2, 0b111000, "jmpl">;              // jmpl [rs1+rs2], rd
 def JMPLri : F3_2<2, 0b111000, "jmpl">;              // jmpl [rs1+imm], rd
 
-// Section B.29 - Write State Register Instructions, p. 133
-let rd = 0 in
-  def WRYrr : F3_1<2, 0b110000, "wr">; // Special case of WRASR
-def WRASRrr : F3_1<2, 0b110000, "wr">; // Special reg = reg ^ reg
+// Section B.29 - Write State Register Instructions
+def WRrr : F3_1<2, 0b110000, "wr">;                    // wr rs1, rs2, rd
+def WRri : F3_2<2, 0b110000, "wr">;                    // wr rs1, imm, rd
diff --git a/lib/Target/Sparc/SparcV8ISelSimple.cpp b/lib/Target/Sparc/SparcV8ISelSimple.cpp
index a73d97b..3bae6a5 100644
--- a/lib/Target/Sparc/SparcV8ISelSimple.cpp
+++ b/lib/Target/Sparc/SparcV8ISelSimple.cpp
@@ -288,33 +288,53 @@
   unsigned Op1Reg = getReg (I.getOperand (1));
 
   unsigned ResultReg = makeAnotherReg (I.getType ());
-  
+  unsigned OpCase = ~0;
+
   // FIXME: support long, ulong, fp.
   switch (I.getOpcode ()) {
-    case Instruction::Add: 
-      BuildMI (BB, V8::ADDrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Sub: 
-      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Mul: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SMULrr : V8::UMULrr;
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+  case Instruction::Add: OpCase = 0; break;
+  case Instruction::Sub: OpCase = 1; break;
+  case Instruction::Mul: OpCase = 2; break;
+  case Instruction::And: OpCase = 3; break;
+  case Instruction::Or:  OpCase = 4; break;
+  case Instruction::Xor: OpCase = 5; break;
+
+  case Instruction::Div:
+  case Instruction::Rem: {
+    unsigned Dest = ResultReg;
+    if (I.getOpcode() == Instruction::Rem)
+      Dest = makeAnotherReg(I.getType());
+
+    // FIXME: this is probably only right for 32 bit operands.
+    if (I.getType ()->isSigned()) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      // Sign extend into the Y register
+      BuildMI (BB, V8::SRAri, 2, Tmp).addReg (Op0Reg).addZImm (31);
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (Tmp).addReg (V8::G0);
+      BuildMI (BB, V8::SDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
+    } else {
+      // Zero extend into the Y register, ie, just set it to zero
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
+      BuildMI (BB, V8::UDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
     }
-    case Instruction::Div: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SDIVrr : V8::UDIVrr;
-      // Clear out the Y register (top half of LHS of divide) 
-      BuildMI (BB, V8::WRYrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
-      BuildMI (BB, V8::NOP, 0); // WR may take up to 4 cycles to finish
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+
+    if (I.getOpcode() == Instruction::Rem) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      BuildMI (BB, V8::SMULrr, 2, Tmp).addReg(Dest).addReg(Op1Reg);
+      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg(Op0Reg).addReg(Tmp);
     }
-    default:
-      visitInstruction (I);
-      return;
+    break;
+  }
+  default:
+    visitInstruction (I);
+    return;
+  }
+
+  if (OpCase != ~0U) {
+    static const unsigned Opcodes[] = {
+      V8::ADDrr, V8::SUBrr, V8::SMULrr, V8::ANDrr, V8::ORrr, V8::XORrr
+    };
+    BuildMI (BB, Opcodes[OpCase], 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
   }
 
   switch (getClass (I.getType ())) {
diff --git a/lib/Target/SparcV8/InstSelectSimple.cpp b/lib/Target/SparcV8/InstSelectSimple.cpp
index a73d97b..3bae6a5 100644
--- a/lib/Target/SparcV8/InstSelectSimple.cpp
+++ b/lib/Target/SparcV8/InstSelectSimple.cpp
@@ -288,33 +288,53 @@
   unsigned Op1Reg = getReg (I.getOperand (1));
 
   unsigned ResultReg = makeAnotherReg (I.getType ());
-  
+  unsigned OpCase = ~0;
+
   // FIXME: support long, ulong, fp.
   switch (I.getOpcode ()) {
-    case Instruction::Add: 
-      BuildMI (BB, V8::ADDrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Sub: 
-      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Mul: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SMULrr : V8::UMULrr;
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+  case Instruction::Add: OpCase = 0; break;
+  case Instruction::Sub: OpCase = 1; break;
+  case Instruction::Mul: OpCase = 2; break;
+  case Instruction::And: OpCase = 3; break;
+  case Instruction::Or:  OpCase = 4; break;
+  case Instruction::Xor: OpCase = 5; break;
+
+  case Instruction::Div:
+  case Instruction::Rem: {
+    unsigned Dest = ResultReg;
+    if (I.getOpcode() == Instruction::Rem)
+      Dest = makeAnotherReg(I.getType());
+
+    // FIXME: this is probably only right for 32 bit operands.
+    if (I.getType ()->isSigned()) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      // Sign extend into the Y register
+      BuildMI (BB, V8::SRAri, 2, Tmp).addReg (Op0Reg).addZImm (31);
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (Tmp).addReg (V8::G0);
+      BuildMI (BB, V8::SDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
+    } else {
+      // Zero extend into the Y register, ie, just set it to zero
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
+      BuildMI (BB, V8::UDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
     }
-    case Instruction::Div: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SDIVrr : V8::UDIVrr;
-      // Clear out the Y register (top half of LHS of divide) 
-      BuildMI (BB, V8::WRYrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
-      BuildMI (BB, V8::NOP, 0); // WR may take up to 4 cycles to finish
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+
+    if (I.getOpcode() == Instruction::Rem) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      BuildMI (BB, V8::SMULrr, 2, Tmp).addReg(Dest).addReg(Op1Reg);
+      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg(Op0Reg).addReg(Tmp);
     }
-    default:
-      visitInstruction (I);
-      return;
+    break;
+  }
+  default:
+    visitInstruction (I);
+    return;
+  }
+
+  if (OpCase != ~0U) {
+    static const unsigned Opcodes[] = {
+      V8::ADDrr, V8::SUBrr, V8::SMULrr, V8::ANDrr, V8::ORrr, V8::XORrr
+    };
+    BuildMI (BB, Opcodes[OpCase], 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
   }
 
   switch (getClass (I.getType ())) {
diff --git a/lib/Target/SparcV8/SparcV8ISelSimple.cpp b/lib/Target/SparcV8/SparcV8ISelSimple.cpp
index a73d97b..3bae6a5 100644
--- a/lib/Target/SparcV8/SparcV8ISelSimple.cpp
+++ b/lib/Target/SparcV8/SparcV8ISelSimple.cpp
@@ -288,33 +288,53 @@
   unsigned Op1Reg = getReg (I.getOperand (1));
 
   unsigned ResultReg = makeAnotherReg (I.getType ());
-  
+  unsigned OpCase = ~0;
+
   // FIXME: support long, ulong, fp.
   switch (I.getOpcode ()) {
-    case Instruction::Add: 
-      BuildMI (BB, V8::ADDrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Sub: 
-      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
-    case Instruction::Mul: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SMULrr : V8::UMULrr;
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+  case Instruction::Add: OpCase = 0; break;
+  case Instruction::Sub: OpCase = 1; break;
+  case Instruction::Mul: OpCase = 2; break;
+  case Instruction::And: OpCase = 3; break;
+  case Instruction::Or:  OpCase = 4; break;
+  case Instruction::Xor: OpCase = 5; break;
+
+  case Instruction::Div:
+  case Instruction::Rem: {
+    unsigned Dest = ResultReg;
+    if (I.getOpcode() == Instruction::Rem)
+      Dest = makeAnotherReg(I.getType());
+
+    // FIXME: this is probably only right for 32 bit operands.
+    if (I.getType ()->isSigned()) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      // Sign extend into the Y register
+      BuildMI (BB, V8::SRAri, 2, Tmp).addReg (Op0Reg).addZImm (31);
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (Tmp).addReg (V8::G0);
+      BuildMI (BB, V8::SDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
+    } else {
+      // Zero extend into the Y register, ie, just set it to zero
+      BuildMI (BB, V8::WRrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
+      BuildMI (BB, V8::UDIVrr, 2, Dest).addReg (Op0Reg).addReg (Op1Reg);
     }
-    case Instruction::Div: {
-      unsigned Opcode = I.getType ()->isSigned () ? V8::SDIVrr : V8::UDIVrr;
-      // Clear out the Y register (top half of LHS of divide) 
-      BuildMI (BB, V8::WRYrr, 2, V8::Y).addReg (V8::G0).addReg (V8::G0);
-      BuildMI (BB, V8::NOP, 0); // WR may take up to 4 cycles to finish
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, V8::NOP, 0);
-      BuildMI (BB, Opcode, 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
-      break;
+
+    if (I.getOpcode() == Instruction::Rem) {
+      unsigned Tmp = makeAnotherReg (I.getType ());
+      BuildMI (BB, V8::SMULrr, 2, Tmp).addReg(Dest).addReg(Op1Reg);
+      BuildMI (BB, V8::SUBrr, 2, ResultReg).addReg(Op0Reg).addReg(Tmp);
     }
-    default:
-      visitInstruction (I);
-      return;
+    break;
+  }
+  default:
+    visitInstruction (I);
+    return;
+  }
+
+  if (OpCase != ~0U) {
+    static const unsigned Opcodes[] = {
+      V8::ADDrr, V8::SUBrr, V8::SMULrr, V8::ANDrr, V8::ORrr, V8::XORrr
+    };
+    BuildMI (BB, Opcodes[OpCase], 2, ResultReg).addReg (Op0Reg).addReg (Op1Reg);
   }
 
   switch (getClass (I.getType ())) {
diff --git a/lib/Target/SparcV8/SparcV8InstrInfo.td b/lib/Target/SparcV8/SparcV8InstrInfo.td
index db79597..ae4e9e6 100644
--- a/lib/Target/SparcV8/SparcV8InstrInfo.td
+++ b/lib/Target/SparcV8/SparcV8InstrInfo.td
@@ -75,9 +75,12 @@
   def NOP : F2_1<0b100, "nop">;
 
 // Section B.11 - Logical Instructions, p. 106
+def ANDrr : F3_1<2, 0b000001, "and">;
 def ANDri : F3_2<2, 0b000001, "and">;
 def ORrr  : F3_1<2, 0b000010, "or">;
 def ORri  : F3_2<2, 0b000010, "or">;
+def XORrr : F3_1<2, 0b000011, "xor">;
+def XORri : F3_2<2, 0b000011, "xor">;
 
 // Section B.12 - Shift Instructions, p. 107
 def SLLri : F3_1<2, 0b100101, "sll">;
@@ -95,8 +98,14 @@
 def SMULrr : F3_1<2, 0b001011, "smul">;
 
 // Section B.19 - Divide Instructions, p. 115
-def UDIVrr: F3_1<2, 0b001110, "udiv">;
-def SDIVrr: F3_1<2, 0b001111, "sdiv">;
+def UDIVrr   : F3_1<2, 0b001110, "udiv">;
+def UDIVri   : F3_2<2, 0b001110, "udiv">;
+def SDIVrr   : F3_1<2, 0b001111, "sdiv">;
+def SDIVri   : F3_2<2, 0b001111, "sdiv">;
+def UDIVCCrr : F3_1<2, 0b011110, "udivcc">;
+def UDIVCCri : F3_2<2, 0b011110, "udivcc">;
+def SDIVCCrr : F3_1<2, 0b011111, "sdivcc">;
+def SDIVCCri : F3_2<2, 0b011111, "sdivcc">;
 
 // Section B.20 - SAVE and RESTORE, p. 117
 def SAVErr    : F3_1<2, 0b111100, "save">;           // save    r, r, r
@@ -118,7 +127,6 @@
 def JMPLrr : F3_1<2, 0b111000, "jmpl">;              // jmpl [rs1+rs2], rd
 def JMPLri : F3_2<2, 0b111000, "jmpl">;              // jmpl [rs1+imm], rd
 
-// Section B.29 - Write State Register Instructions, p. 133
-let rd = 0 in
-  def WRYrr : F3_1<2, 0b110000, "wr">; // Special case of WRASR
-def WRASRrr : F3_1<2, 0b110000, "wr">; // Special reg = reg ^ reg
+// Section B.29 - Write State Register Instructions
+def WRrr : F3_1<2, 0b110000, "wr">;                    // wr rs1, rs2, rd
+def WRri : F3_2<2, 0b110000, "wr">;                    // wr rs1, imm, rd