lib/Target/Sparc/SparcInstrInfo.td

//===- SparcV8Instrs.td - Target Description for SparcV8 Target -----------===//
// 
//                     The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
// 
//===----------------------------------------------------------------------===//
//
// This file describes the SparcV8 instructions in TableGen format.
//
//===----------------------------------------------------------------------===//

//===----------------------------------------------------------------------===//
// Instruction format superclass
//===----------------------------------------------------------------------===//

include "SparcV8InstrFormats.td"

//===----------------------------------------------------------------------===//
// Feature predicates.
//===----------------------------------------------------------------------===//

// HasV9 - This predicate is true when the target processor supports V9
// instructions.  Note that the machine may be running in 32-bit mode.
def HasV9   : Predicate<"Subtarget.isV9()">;

// HasVIS - This is true when the target processor has VIS extensions.
def HasVIS : Predicate<"Subtarget.isVIS()">;

// UseDeprecatedInsts - This predicate is true when the target processor is a
// V8, or when it is V9 but the V8 deprecated instructions are efficient enough
// to use when appropriate.  In either of these cases, the instruction selector
// will pick deprecated instructions.
def UseDeprecatedInsts : Predicate<"Subtarget.useDeprecatedV8Instructions()">;

//===----------------------------------------------------------------------===//
// Instruction Pattern Stuff
//===----------------------------------------------------------------------===//

def simm13  : PatLeaf<(imm), [{
  // simm13 predicate - True if the imm fits in a 13-bit sign extended field.
  return (((int)N->getValue() << (32-13)) >> (32-13)) == (int)N->getValue();
}]>;

def LO10 : SDNodeXForm<imm, [{
  return CurDAG->getTargetConstant((unsigned)N->getValue() & 1023, MVT::i32);
}]>;

def HI22 : SDNodeXForm<imm, [{
  // Transformation function: shift the immediate value down into the low bits.
  return CurDAG->getTargetConstant((unsigned)N->getValue() >> 10, MVT::i32);
}]>;

def SETHIimm : PatLeaf<(imm), [{
  return (((unsigned)N->getValue() >> 10) << 10) == (unsigned)N->getValue();
}], HI22>;

// Addressing modes.
def ADDRrr : ComplexPattern<i32, 2, "SelectADDRrr", []>;
def ADDRri : ComplexPattern<i32, 2, "SelectADDRri", []>;

// Address operands
def MEMrr : Operand<i32> {
  let PrintMethod = "printMemOperand";
  let NumMIOperands = 2;
  let MIOperandInfo = (ops IntRegs, IntRegs);
}
def MEMri : Operand<i32> {
  let PrintMethod = "printMemOperand";
  let NumMIOperands = 2;
  let MIOperandInfo = (ops IntRegs, i32imm);
}

// Branch targets have OtherVT type.
def brtarget : Operand<OtherVT>;
def calltarget : Operand<i32>;

def SDTV8cmpfcc : 
SDTypeProfile<1, 2, [SDTCisVT<0, FlagVT>, SDTCisFP<1>, SDTCisSameAs<1, 2>]>;
def SDTV8brcc : 
SDTypeProfile<0, 3, [SDTCisVT<0, OtherVT>, SDTCisVT<1, OtherVT>,
                     SDTCisVT<2, FlagVT>]>;
def SDTV8selectcc :
SDTypeProfile<1, 4, [SDTCisSameAs<0, 1>, SDTCisSameAs<1, 2>, 
                     SDTCisVT<3, i32>, SDTCisVT<4, FlagVT>]>;
def SDTV8FTOI :
SDTypeProfile<1, 1, [SDTCisVT<0, f32>, SDTCisFP<1>]>;
def SDTV8ITOF :
SDTypeProfile<1, 1, [SDTCisFP<0>, SDTCisVT<1, f32>]>;

def V8cmpicc : SDNode<"V8ISD::CMPICC", SDTIntBinOp, [SDNPOutFlag]>;
def V8cmpfcc : SDNode<"V8ISD::CMPFCC", SDTV8cmpfcc, [SDNPOutFlag]>;
def V8bricc : SDNode<"V8ISD::BRICC", SDTV8brcc, [SDNPHasChain]>;
def V8brfcc : SDNode<"V8ISD::BRFCC", SDTV8brcc, [SDNPHasChain]>;

def V8hi    : SDNode<"V8ISD::Hi", SDTIntUnaryOp>;
def V8lo    : SDNode<"V8ISD::Lo", SDTIntUnaryOp>;

def V8ftoi  : SDNode<"V8ISD::FTOI", SDTV8FTOI>;
def V8itof  : SDNode<"V8ISD::ITOF", SDTV8ITOF>;

def V8selecticc : SDNode<"V8ISD::SELECT_ICC", SDTV8selectcc>;
def V8selectfcc : SDNode<"V8ISD::SELECT_FCC", SDTV8selectcc>;

// These are target-independent nodes, but have target-specific formats.
def SDT_V8CallSeq : SDTypeProfile<0, 1, [ SDTCisVT<0, i32> ]>;
def callseq_start : SDNode<"ISD::CALLSEQ_START", SDT_V8CallSeq, [SDNPHasChain]>;
def callseq_end   : SDNode<"ISD::CALLSEQ_END",   SDT_V8CallSeq, [SDNPHasChain]>;

def SDT_V8Call    : SDTypeProfile<0, 1, [SDTCisVT<0, i32>]>;
def call          : SDNode<"V8ISD::CALL", SDT_V8Call,
	                   [SDNPHasChain, SDNPOptInFlag, SDNPOutFlag]>;

def SDT_V8RetFlag : SDTypeProfile<0, 0, []>;
def retflag       : SDNode<"V8ISD::RET_FLAG", SDT_V8RetFlag,
	                   [SDNPHasChain, SDNPOptInFlag]>;

//===----------------------------------------------------------------------===//
// Instructions
//===----------------------------------------------------------------------===//

// Pseudo instructions.
class Pseudo<dag ops, string asmstr, list<dag> pattern>
   : InstV8<ops, asmstr, pattern>;

def ADJCALLSTACKDOWN : Pseudo<(ops i32imm:$amt),
                               "!ADJCALLSTACKDOWN $amt",
                               [(callseq_start imm:$amt)]>;
def ADJCALLSTACKUP : Pseudo<(ops i32imm:$amt),
                            "!ADJCALLSTACKUP $amt",
                            [(callseq_end imm:$amt)]>;
def IMPLICIT_DEF_Int : Pseudo<(ops IntRegs:$dst),
                              "!IMPLICIT_DEF $dst",
                              [(set IntRegs:$dst, (undef))]>;
def IMPLICIT_DEF_FP  : Pseudo<(ops FPRegs:$dst), "!IMPLICIT_DEF $dst",
                              [(set FPRegs:$dst, (undef))]>;
def IMPLICIT_DEF_DFP : Pseudo<(ops DFPRegs:$dst), "!IMPLICIT_DEF $dst",
                              [(set DFPRegs:$dst, (undef))]>;
                              
// FpMOVD/FpNEGD/FpABSD - These are lowered to single-precision ops by the 
// fpmover pass.
def FpMOVD : Pseudo<(ops DFPRegs:$dst, DFPRegs:$src),
                    "!FpMOVD $src, $dst", []>;   // pseudo 64-bit double move
def FpNEGD : Pseudo<(ops DFPRegs:$dst, DFPRegs:$src),
                    "!FpNEGD $src, $dst",
                    [(set DFPRegs:$dst, (fneg DFPRegs:$src))]>;
def FpABSD : Pseudo<(ops DFPRegs:$dst, DFPRegs:$src),
                    "!FpABSD $src, $dst",
                    [(set DFPRegs:$dst, (fabs DFPRegs:$src))]>;

// SELECT_CC_* - Used to implement the SELECT_CC DAG operation.  Expanded by the
// scheduler into a branch sequence.  This has to handle all permutations of
// selection between i32/f32/f64 on ICC and FCC.
let usesCustomDAGSchedInserter = 1 in {  // Expanded by the scheduler.
  def SELECT_CC_Int_ICC
   : Pseudo<(ops IntRegs:$dst, IntRegs:$T, IntRegs:$F, i32imm:$Cond),
            "; SELECT_CC_Int_ICC PSEUDO!",
            [(set IntRegs:$dst, (V8selecticc IntRegs:$T, IntRegs:$F,
                                             imm:$Cond, ICC))]>;
  def SELECT_CC_Int_FCC
   : Pseudo<(ops IntRegs:$dst, IntRegs:$T, IntRegs:$F, i32imm:$Cond),
            "; SELECT_CC_Int_FCC PSEUDO!",
            [(set IntRegs:$dst, (V8selectfcc IntRegs:$T, IntRegs:$F,
                                             imm:$Cond, FCC))]>;
  def SELECT_CC_FP_ICC
   : Pseudo<(ops FPRegs:$dst, FPRegs:$T, FPRegs:$F, i32imm:$Cond),
            "; SELECT_CC_FP_ICC PSEUDO!",
            [(set FPRegs:$dst, (V8selecticc FPRegs:$T, FPRegs:$F,
                                            imm:$Cond, ICC))]>;
  def SELECT_CC_FP_FCC
   : Pseudo<(ops FPRegs:$dst, FPRegs:$T, FPRegs:$F, i32imm:$Cond),
            "; SELECT_CC_FP_FCC PSEUDO!",
            [(set FPRegs:$dst, (V8selectfcc FPRegs:$T, FPRegs:$F,
                                            imm:$Cond, FCC))]>;
  def SELECT_CC_DFP_ICC
   : Pseudo<(ops DFPRegs:$dst, DFPRegs:$T, DFPRegs:$F, i32imm:$Cond),
            "; SELECT_CC_DFP_ICC PSEUDO!",
            [(set DFPRegs:$dst, (V8selecticc DFPRegs:$T, DFPRegs:$F,
                                             imm:$Cond, ICC))]>;
  def SELECT_CC_DFP_FCC
   : Pseudo<(ops DFPRegs:$dst, DFPRegs:$T, DFPRegs:$F, i32imm:$Cond),
            "; SELECT_CC_DFP_FCC PSEUDO!",
            [(set DFPRegs:$dst, (V8selectfcc DFPRegs:$T, DFPRegs:$F,
                                             imm:$Cond, FCC))]>;
}


// Section A.3 - Synthetic Instructions, p. 85
// special cases of JMPL:
let isReturn = 1, isTerminator = 1, hasDelaySlot = 1, noResults = 1 in {
  let rd = O7.Num, rs1 = G0.Num, simm13 = 8 in
    def RETL: F3_2<2, 0b111000, (ops), "retl", [(retflag)]>;
}

// Section B.1 - Load Integer Instructions, p. 90
def LDSBrr : F3_1<3, 0b001001,
                  (ops IntRegs:$dst, MEMrr:$addr),
                  "ldsb [$addr], $dst",
                  [(set IntRegs:$dst, (sextload ADDRrr:$addr, i8))]>;
def LDSBri : F3_2<3, 0b001001,
                  (ops IntRegs:$dst, MEMri:$addr),
                  "ldsb [$addr], $dst",
                  [(set IntRegs:$dst, (sextload ADDRri:$addr, i8))]>;
def LDSHrr : F3_1<3, 0b001010,
                  (ops IntRegs:$dst, MEMrr:$addr),
                  "ldsh [$addr], $dst",
                  [(set IntRegs:$dst, (sextload ADDRrr:$addr, i16))]>;
def LDSHri : F3_2<3, 0b001010,
                  (ops IntRegs:$dst, MEMri:$addr),
                  "ldsh [$addr], $dst",
                  [(set IntRegs:$dst, (sextload ADDRri:$addr, i16))]>;
def LDUBrr : F3_1<3, 0b000001,
                  (ops IntRegs:$dst, MEMrr:$addr),
                  "ldub [$addr], $dst",
                  [(set IntRegs:$dst, (zextload ADDRrr:$addr, i8))]>;
def LDUBri : F3_2<3, 0b000001,
                  (ops IntRegs:$dst, MEMri:$addr),
                  "ldub [$addr], $dst",
                  [(set IntRegs:$dst, (zextload ADDRri:$addr, i8))]>;
def LDUHrr : F3_1<3, 0b000010,
                  (ops IntRegs:$dst, MEMrr:$addr),
                  "lduh [$addr], $dst",
                  [(set IntRegs:$dst, (zextload ADDRrr:$addr, i16))]>;
def LDUHri : F3_2<3, 0b000010,
                  (ops IntRegs:$dst, MEMri:$addr),
                  "lduh [$addr], $dst",
                  [(set IntRegs:$dst, (zextload ADDRri:$addr, i16))]>;
def LDrr   : F3_1<3, 0b000000,
                  (ops IntRegs:$dst, MEMrr:$addr),
                  "ld [$addr], $dst",
                  [(set IntRegs:$dst, (load ADDRrr:$addr))]>;
def LDri   : F3_2<3, 0b000000,
                  (ops IntRegs:$dst, MEMri:$addr),
                  "ld [$addr], $dst",
                  [(set IntRegs:$dst, (load ADDRri:$addr))]>;

// Section B.2 - Load Floating-point Instructions, p. 92
def LDFrr  : F3_1<3, 0b100000,
                  (ops FPRegs:$dst, MEMrr:$addr),
                  "ld [$addr], $dst",
                  [(set FPRegs:$dst, (load ADDRrr:$addr))]>;
def LDFri  : F3_2<3, 0b100000,
                  (ops FPRegs:$dst, MEMri:$addr),
                  "ld [$addr], $dst",
                  [(set FPRegs:$dst, (load ADDRri:$addr))]>;
def LDDFrr : F3_1<3, 0b100011,
                  (ops DFPRegs:$dst, MEMrr:$addr),
                  "ldd [$addr], $dst",
                  [(set DFPRegs:$dst, (load ADDRrr:$addr))]>;
def LDDFri : F3_2<3, 0b100011,
                  (ops DFPRegs:$dst, MEMri:$addr),
                  "ldd [$addr], $dst",
                  [(set DFPRegs:$dst, (load ADDRri:$addr))]>;

// Section B.4 - Store Integer Instructions, p. 95
def STBrr : F3_1<3, 0b000101,
                 (ops MEMrr:$addr, IntRegs:$src),
                 "stb $src, [$addr]",
                 [(truncstore IntRegs:$src, ADDRrr:$addr, i8)]>;
def STBri : F3_2<3, 0b000101,
                 (ops MEMri:$addr, IntRegs:$src),
                 "stb $src, [$addr]",
                 [(truncstore IntRegs:$src, ADDRri:$addr, i8)]>;
def STHrr : F3_1<3, 0b000110,
                 (ops MEMrr:$addr, IntRegs:$src),
                 "sth $src, [$addr]",
                 [(truncstore IntRegs:$src, ADDRrr:$addr, i16)]>;
def STHri : F3_2<3, 0b000110,
                 (ops MEMri:$addr, IntRegs:$src),
                 "sth $src, [$addr]",
                 [(truncstore IntRegs:$src, ADDRri:$addr, i16)]>;
def STrr  : F3_1<3, 0b000100,
                 (ops MEMrr:$addr, IntRegs:$src),
                 "st $src, [$addr]",
                 [(store IntRegs:$src, ADDRrr:$addr)]>;
def STri  : F3_2<3, 0b000100,
                 (ops MEMri:$addr, IntRegs:$src),
                 "st $src, [$addr]",
                 [(store IntRegs:$src, ADDRri:$addr)]>;

// Section B.5 - Store Floating-point Instructions, p. 97
def STFrr   : F3_1<3, 0b100100,
                   (ops MEMrr:$addr, FPRegs:$src),
                   "st $src, [$addr]",
                   [(store FPRegs:$src, ADDRrr:$addr)]>;
def STFri   : F3_2<3, 0b100100,
                   (ops MEMri:$addr, FPRegs:$src),
                   "st $src, [$addr]",
                   [(store FPRegs:$src, ADDRri:$addr)]>;
def STDFrr  : F3_1<3, 0b100111,
                   (ops MEMrr:$addr, DFPRegs:$src),
                   "std  $src, [$addr]",
                   [(store DFPRegs:$src, ADDRrr:$addr)]>;
def STDFri  : F3_2<3, 0b100111,
                   (ops MEMri:$addr, DFPRegs:$src),
                   "std $src, [$addr]",
                   [(store DFPRegs:$src, ADDRri:$addr)]>;

// Section B.9 - SETHI Instruction, p. 104
def SETHIi: F2_1<0b100,
                 (ops IntRegs:$dst, i32imm:$src),
                 "sethi $src, $dst",
                 [(set IntRegs:$dst, SETHIimm:$src)]>;

// Section B.10 - NOP Instruction, p. 105
// (It's a special case of SETHI)
let rd = 0, imm22 = 0 in
  def NOP : F2_1<0b100, (ops), "nop", []>;

// Section B.11 - Logical Instructions, p. 106
def ANDrr   : F3_1<2, 0b000001,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "and $b, $c, $dst",
                   [(set IntRegs:$dst, (and IntRegs:$b, IntRegs:$c))]>;
def ANDri   : F3_2<2, 0b000001,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "and $b, $c, $dst",
                   [(set IntRegs:$dst, (and IntRegs:$b, simm13:$c))]>;
def ANDNrr  : F3_1<2, 0b000101,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "andn $b, $c, $dst",
                   [(set IntRegs:$dst, (and IntRegs:$b, (not IntRegs:$c)))]>;
def ANDNri  : F3_2<2, 0b000101,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "andn $b, $c, $dst", []>;
def ORrr    : F3_1<2, 0b000010,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "or $b, $c, $dst",
                   [(set IntRegs:$dst, (or IntRegs:$b, IntRegs:$c))]>;
def ORri    : F3_2<2, 0b000010,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "or $b, $c, $dst",
                   [(set IntRegs:$dst, (or IntRegs:$b, simm13:$c))]>;
def ORNrr   : F3_1<2, 0b000110,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "orn $b, $c, $dst",
                   [(set IntRegs:$dst, (or IntRegs:$b, (not IntRegs:$c)))]>;
def ORNri   : F3_2<2, 0b000110,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "orn $b, $c, $dst", []>;
def XORrr   : F3_1<2, 0b000011,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "xor $b, $c, $dst",
                   [(set IntRegs:$dst, (xor IntRegs:$b, IntRegs:$c))]>;
def XORri   : F3_2<2, 0b000011,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "xor $b, $c, $dst",
                   [(set IntRegs:$dst, (xor IntRegs:$b, simm13:$c))]>;
def XNORrr  : F3_1<2, 0b000111,
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "xnor $b, $c, $dst",
                   [(set IntRegs:$dst, (not (xor IntRegs:$b, IntRegs:$c)))]>;
def XNORri  : F3_2<2, 0b000111,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "xnor $b, $c, $dst", []>;

// Section B.12 - Shift Instructions, p. 107
def SLLrr : F3_1<2, 0b100101,
                 (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                 "sll $b, $c, $dst",
                 [(set IntRegs:$dst, (shl IntRegs:$b, IntRegs:$c))]>;
def SLLri : F3_2<2, 0b100101,
                 (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                 "sll $b, $c, $dst",
                 [(set IntRegs:$dst, (shl IntRegs:$b, simm13:$c))]>;
def SRLrr : F3_1<2, 0b100110, 
                 (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                  "srl $b, $c, $dst",
                  [(set IntRegs:$dst, (srl IntRegs:$b, IntRegs:$c))]>;
def SRLri : F3_2<2, 0b100110,
                 (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                 "srl $b, $c, $dst", 
                 [(set IntRegs:$dst, (srl IntRegs:$b, simm13:$c))]>;
def SRArr : F3_1<2, 0b100111, 
                 (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                  "sra $b, $c, $dst",
                  [(set IntRegs:$dst, (sra IntRegs:$b, IntRegs:$c))]>;
def SRAri : F3_2<2, 0b100111,
                 (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                 "sra $b, $c, $dst",
                 [(set IntRegs:$dst, (sra IntRegs:$b, simm13:$c))]>;

// Section B.13 - Add Instructions, p. 108
def ADDrr   : F3_1<2, 0b000000, 
                  (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                  "add $b, $c, $dst",
                   [(set IntRegs:$dst, (add IntRegs:$b, IntRegs:$c))]>;
def ADDri   : F3_2<2, 0b000000,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "add $b, $c, $dst",
                   [(set IntRegs:$dst, (add IntRegs:$b, simm13:$c))]>;
def ADDCCrr : F3_1<2, 0b010000, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "addcc $b, $c, $dst", []>;
def ADDCCri : F3_2<2, 0b010000,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "addcc $b, $c, $dst", []>;
def ADDXrr  : F3_1<2, 0b001000, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "addx $b, $c, $dst", []>;
def ADDXri  : F3_2<2, 0b001000,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "addx $b, $c, $dst", []>;

// Section B.15 - Subtract Instructions, p. 110
def SUBrr   : F3_1<2, 0b000100, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "sub $b, $c, $dst",
                   [(set IntRegs:$dst, (sub IntRegs:$b, IntRegs:$c))]>;
def SUBri   : F3_2<2, 0b000100,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "sub $b, $c, $dst",
                   [(set IntRegs:$dst, (sub IntRegs:$b, simm13:$c))]>;
def SUBXrr  : F3_1<2, 0b001100, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "subx $b, $c, $dst", []>;
def SUBXri  : F3_2<2, 0b001100,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "subx $b, $c, $dst", []>;
def SUBCCrr : F3_1<2, 0b010100, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "subcc $b, $c, $dst",
                   [(set IntRegs:$dst, (V8cmpicc IntRegs:$b, IntRegs:$c))]>;
def SUBCCri : F3_2<2, 0b010100,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "subcc $b, $c, $dst",
                   [(set IntRegs:$dst, (V8cmpicc IntRegs:$b, simm13:$c))]>;
def SUBXCCrr: F3_1<2, 0b011100, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "subxcc $b, $c, $dst", []>;

// Section B.18 - Multiply Instructions, p. 113
def UMULrr  : F3_1<2, 0b001010, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "umul $b, $c, $dst", []>;
def UMULri  : F3_2<2, 0b001010,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "umul $b, $c, $dst", []>;
def SMULrr  : F3_1<2, 0b001011, 
                   (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                   "smul $b, $c, $dst",
                   [(set IntRegs:$dst, (mul IntRegs:$b, IntRegs:$c))]>;
def SMULri  : F3_2<2, 0b001011,
                   (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                   "smul $b, $c, $dst",
                   [(set IntRegs:$dst, (mul IntRegs:$b, simm13:$c))]>;

// Section B.19 - Divide Instructions, p. 115
def UDIVrr   : F3_1<2, 0b001110, 
                    (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                    "udiv $b, $c, $dst", []>;
def UDIVri   : F3_2<2, 0b001110,
                    (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                    "udiv $b, $c, $dst", []>;
def SDIVrr   : F3_1<2, 0b001111,
                    (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                    "sdiv $b, $c, $dst", []>;
def SDIVri   : F3_2<2, 0b001111,
                    (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                    "sdiv $b, $c, $dst", []>;

// Section B.20 - SAVE and RESTORE, p. 117
def SAVErr    : F3_1<2, 0b111100,
                     (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                     "save $b, $c, $dst", []>;
def SAVEri    : F3_2<2, 0b111100,
                     (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                     "save $b, $c, $dst", []>;
def RESTORErr : F3_1<2, 0b111101,
                     (ops IntRegs:$dst, IntRegs:$b, IntRegs:$c),
                     "restore $b, $c, $dst", []>;
def RESTOREri : F3_2<2, 0b111101,
                     (ops IntRegs:$dst, IntRegs:$b, i32imm:$c),
                     "restore $b, $c, $dst", []>;

// Section B.21 - Branch on Integer Condition Codes Instructions, p. 119

// conditional branch class:
class BranchV8<bits<4> cc, dag ops, string asmstr, list<dag> pattern>
 : F2_2<cc, 0b010, ops, asmstr, pattern> {
  let isBranch = 1;
  let isTerminator = 1;
  let hasDelaySlot = 1;
  let noResults = 1;
}

let isBarrier = 1 in
  def BA   : BranchV8<0b1000, (ops brtarget:$dst),
                      "ba $dst",
                      [(br bb:$dst)]>;
def BNE  : BranchV8<0b1001, (ops brtarget:$dst),
                    "bne $dst",
                    [(V8bricc bb:$dst, SETNE, ICC)]>;
def BE   : BranchV8<0b0001, (ops brtarget:$dst),
                    "be $dst",
                    [(V8bricc bb:$dst, SETEQ, ICC)]>;
def BG   : BranchV8<0b1010, (ops brtarget:$dst),
                    "bg $dst",
                    [(V8bricc bb:$dst, SETGT, ICC)]>;
def BLE  : BranchV8<0b0010, (ops brtarget:$dst),
                    "ble $dst",
                    [(V8bricc bb:$dst, SETLE, ICC)]>;
def BGE  : BranchV8<0b1011, (ops brtarget:$dst),
                    "bge $dst",
                    [(V8bricc bb:$dst, SETGE, ICC)]>;
def BL   : BranchV8<0b0011, (ops brtarget:$dst),
                    "bl $dst",
                    [(V8bricc bb:$dst, SETLT, ICC)]>;
def BGU  : BranchV8<0b1100, (ops brtarget:$dst),
                    "bgu $dst",
                    [(V8bricc bb:$dst, SETUGT, ICC)]>;
def BLEU : BranchV8<0b0100, (ops brtarget:$dst),
                    "bleu $dst",
                    [(V8bricc bb:$dst, SETULE, ICC)]>;
def BCC  : BranchV8<0b1101, (ops brtarget:$dst),
                    "bcc $dst",
                    [(V8bricc bb:$dst, SETUGE, ICC)]>;
def BCS  : BranchV8<0b0101, (ops brtarget:$dst),
                    "bcs $dst",
                    [(V8bricc bb:$dst, SETULT, ICC)]>;

// Section B.22 - Branch on Floating-point Condition Codes Instructions, p. 121

// floating-point conditional branch class:
class FPBranchV8<bits<4> cc, dag ops, string asmstr, list<dag> pattern>
 : F2_2<cc, 0b110, ops, asmstr, pattern> {
  let isBranch = 1;
  let isTerminator = 1;
  let hasDelaySlot = 1;
  let noResults = 1;
}

def FBU  : FPBranchV8<0b0111, (ops brtarget:$dst),
                      "fbu $dst",
                      [(V8brfcc bb:$dst, SETUO, FCC)]>;
def FBG  : FPBranchV8<0b0110, (ops brtarget:$dst),
                      "fbg $dst",
                      [(V8brfcc bb:$dst, SETGT, FCC)]>;
def FBUG : FPBranchV8<0b0101, (ops brtarget:$dst),
                      "fbug $dst",
                      [(V8brfcc bb:$dst, SETUGT, FCC)]>;
def FBL  : FPBranchV8<0b0100, (ops brtarget:$dst),
                      "fbl $dst",
                      [(V8brfcc bb:$dst, SETLT, FCC)]>;
def FBUL : FPBranchV8<0b0011, (ops brtarget:$dst),
                      "fbul $dst",
                      [(V8brfcc bb:$dst, SETULT, FCC)]>;
def FBLG : FPBranchV8<0b0010, (ops brtarget:$dst),
                      "fblg $dst",
                      [(V8brfcc bb:$dst, SETONE, FCC)]>;
def FBNE : FPBranchV8<0b0001, (ops brtarget:$dst),
                      "fbne $dst",
                      [(V8brfcc bb:$dst, SETNE, FCC)]>;
def FBE  : FPBranchV8<0b1001, (ops brtarget:$dst),
                      "fbe $dst",
                      [(V8brfcc bb:$dst, SETEQ, FCC)]>;
def FBUE : FPBranchV8<0b1010, (ops brtarget:$dst),
                      "fbue $dst",
                      [(V8brfcc bb:$dst, SETUEQ, FCC)]>;
def FBGE : FPBranchV8<0b1011, (ops brtarget:$dst),
                      "fbge $dst",
                      [(V8brfcc bb:$dst, SETGE, FCC)]>;
def FBUGE: FPBranchV8<0b1100, (ops brtarget:$dst),
                      "fbuge $dst",
                      [(V8brfcc bb:$dst, SETUGE, FCC)]>;
def FBLE : FPBranchV8<0b1101, (ops brtarget:$dst),
                      "fble $dst",
                      [(V8brfcc bb:$dst, SETLE, FCC)]>;
def FBULE: FPBranchV8<0b1110, (ops brtarget:$dst),
                      "fbule $dst",
                      [(V8brfcc bb:$dst, SETULE, FCC)]>;
def FBO  : FPBranchV8<0b1111, (ops brtarget:$dst),
                      "fbo $dst",
                      [(V8brfcc bb:$dst, SETO, FCC)]>;



// Section B.24 - Call and Link Instruction, p. 125
// This is the only Format 1 instruction
let Uses = [O0, O1, O2, O3, O4, O5],
    hasDelaySlot = 1, isCall = 1, noResults = 1,
    Defs = [O0, O1, O2, O3, O4, O5, O7, G1, G2, G3, G4, G5, G6, G7,
    D0, D1, D2, D3, D4, D5, D6, D7, D8, D9, D10, D11, D12, D13, D14, D15] in { 
  def CALL : InstV8<(ops calltarget:$dst),
                    "call $dst", []> {
    bits<30> disp;
    let op = 1;
    let Inst{29-0} = disp;
  }
  
  // indirect calls
  def JMPLrr : F3_1<2, 0b111000,
                    (ops MEMrr:$ptr),
                    "call $ptr",
                    [(call  ADDRrr:$ptr)]>;
  def JMPLri : F3_2<2, 0b111000,
                    (ops MEMri:$ptr),
                    "call $ptr",
                    [(call  ADDRri:$ptr)]>;
}

// Section B.28 - Read State Register Instructions
def RDY : F3_1<2, 0b101000,
               (ops IntRegs:$dst),
               "rd %y, $dst", []>;

// Section B.29 - Write State Register Instructions
def WRYrr : F3_1<2, 0b110000,
                 (ops IntRegs:$b, IntRegs:$c),
                 "wr $b, $c, %y", []>;
def WRYri : F3_2<2, 0b110000,
                 (ops IntRegs:$b, i32imm:$c),
                 "wr $b, $c, %y", []>;

// Convert Integer to Floating-point Instructions, p. 141
def FITOS : F3_3<2, 0b110100, 0b011000100,
                 (ops FPRegs:$dst, FPRegs:$src),
                 "fitos $src, $dst",
                 [(set FPRegs:$dst, (V8itof FPRegs:$src))]>;
def FITOD : F3_3<2, 0b110100, 0b011001000, 
                 (ops DFPRegs:$dst, FPRegs:$src),
                 "fitod $src, $dst",
                 [(set DFPRegs:$dst, (V8itof FPRegs:$src))]>;

// Convert Floating-point to Integer Instructions, p. 142
def FSTOI : F3_3<2, 0b110100, 0b011010001,
                 (ops FPRegs:$dst, FPRegs:$src),
                 "fstoi $src, $dst",
                 [(set FPRegs:$dst, (V8ftoi FPRegs:$src))]>;
def FDTOI : F3_3<2, 0b110100, 0b011010010,
                 (ops FPRegs:$dst, DFPRegs:$src),
                 "fdtoi $src, $dst",
                 [(set FPRegs:$dst, (V8ftoi DFPRegs:$src))]>;

// Convert between Floating-point Formats Instructions, p. 143
def FSTOD : F3_3<2, 0b110100, 0b011001001, 
                 (ops DFPRegs:$dst, FPRegs:$src),
                 "fstod $src, $dst",
                 [(set DFPRegs:$dst, (fextend FPRegs:$src))]>;
def FDTOS : F3_3<2, 0b110100, 0b011000110,
                 (ops FPRegs:$dst, DFPRegs:$src),
                 "fdtos $src, $dst",
                 [(set FPRegs:$dst, (fround DFPRegs:$src))]>;

// Floating-point Move Instructions, p. 144
def FMOVS : F3_3<2, 0b110100, 0b000000001,
                 (ops FPRegs:$dst, FPRegs:$src),
                 "fmovs $src, $dst", []>;
def FNEGS : F3_3<2, 0b110100, 0b000000101, 
                 (ops FPRegs:$dst, FPRegs:$src),
                 "fnegs $src, $dst",
                 [(set FPRegs:$dst, (fneg FPRegs:$src))]>;
def FABSS : F3_3<2, 0b110100, 0b000001001, 
                 (ops FPRegs:$dst, FPRegs:$src),
                 "fabss $src, $dst",
                 [(set FPRegs:$dst, (fabs FPRegs:$src))]>;


// Floating-point Square Root Instructions, p.145
def FSQRTS : F3_3<2, 0b110100, 0b000101001, 
                  (ops FPRegs:$dst, FPRegs:$src),
                  "fsqrts $src, $dst",
                  [(set FPRegs:$dst, (fsqrt FPRegs:$src))]>;
def FSQRTD : F3_3<2, 0b110100, 0b000101010, 
                  (ops DFPRegs:$dst, DFPRegs:$src),
                  "fsqrtd $src, $dst",
                  [(set DFPRegs:$dst, (fsqrt DFPRegs:$src))]>;



// Floating-point Add and Subtract Instructions, p. 146
def FADDS  : F3_3<2, 0b110100, 0b001000001,
                  (ops FPRegs:$dst, FPRegs:$src1, FPRegs:$src2),
                  "fadds $src1, $src2, $dst",
                  [(set FPRegs:$dst, (fadd FPRegs:$src1, FPRegs:$src2))]>;
def FADDD  : F3_3<2, 0b110100, 0b001000010,
                  (ops DFPRegs:$dst, DFPRegs:$src1, DFPRegs:$src2),
                  "faddd $src1, $src2, $dst",
                  [(set DFPRegs:$dst, (fadd DFPRegs:$src1, DFPRegs:$src2))]>;
def FSUBS  : F3_3<2, 0b110100, 0b001000101,
                  (ops FPRegs:$dst, FPRegs:$src1, FPRegs:$src2),
                  "fsubs $src1, $src2, $dst",
                  [(set FPRegs:$dst, (fsub FPRegs:$src1, FPRegs:$src2))]>;
def FSUBD  : F3_3<2, 0b110100, 0b001000110,
                  (ops DFPRegs:$dst, DFPRegs:$src1, DFPRegs:$src2),
                  "fsubd $src1, $src2, $dst",
                  [(set DFPRegs:$dst, (fsub DFPRegs:$src1, DFPRegs:$src2))]>;

// Floating-point Multiply and Divide Instructions, p. 147
def FMULS  : F3_3<2, 0b110100, 0b001001001,
                  (ops FPRegs:$dst, FPRegs:$src1, FPRegs:$src2),
                  "fmuls $src1, $src2, $dst",
                  [(set FPRegs:$dst, (fmul FPRegs:$src1, FPRegs:$src2))]>;
def FMULD  : F3_3<2, 0b110100, 0b001001010,
                  (ops DFPRegs:$dst, DFPRegs:$src1, DFPRegs:$src2),
                  "fmuld $src1, $src2, $dst",
                  [(set DFPRegs:$dst, (fmul DFPRegs:$src1, DFPRegs:$src2))]>;
def FSMULD : F3_3<2, 0b110100, 0b001101001,
                  (ops DFPRegs:$dst, FPRegs:$src1, FPRegs:$src2),
                  "fsmuld $src1, $src2, $dst",
                  [(set DFPRegs:$dst, (fmul (fextend FPRegs:$src1),
                                            (fextend FPRegs:$src2)))]>;
def FDIVS  : F3_3<2, 0b110100, 0b001001101,
                 (ops FPRegs:$dst, FPRegs:$src1, FPRegs:$src2),
                 "fdivs $src1, $src2, $dst",
                 [(set FPRegs:$dst, (fdiv FPRegs:$src1, FPRegs:$src2))]>;
def FDIVD  : F3_3<2, 0b110100, 0b001001110,
                 (ops DFPRegs:$dst, DFPRegs:$src1, DFPRegs:$src2),
                 "fdivd $src1, $src2, $dst",
                 [(set DFPRegs:$dst, (fdiv DFPRegs:$src1, DFPRegs:$src2))]>;

// Floating-point Compare Instructions, p. 148
// Note: the 2nd template arg is different for these guys.
// Note 2: the result of a FCMP is not available until the 2nd cycle
// after the instr is retired, but there is no interlock. This behavior
// is modelled with a forced noop after the instruction.
def FCMPS  : F3_3<2, 0b110101, 0b001010001,
                  (ops FPRegs:$src1, FPRegs:$src2),
                  "fcmps $src1, $src2\n\tnop",
                  [(set FCC, (V8cmpfcc FPRegs:$src1, FPRegs:$src2))]>;
def FCMPD  : F3_3<2, 0b110101, 0b001010010,
                  (ops DFPRegs:$src1, DFPRegs:$src2),
                  "fcmpd $src1, $src2\n\tnop",
                  [(set FCC, (V8cmpfcc DFPRegs:$src1, DFPRegs:$src2))]>;


//===----------------------------------------------------------------------===//
// V9 Instructions
//===----------------------------------------------------------------------===//

// V9 Conditional Moves.
let Predicates = [HasV9], isTwoAddress = 1 in {
  // FIXME: Add instruction encodings for the JIT some day.
  def MOVNE : Pseudo<(ops IntRegs:$dst, IntRegs:$T, IntRegs:$F),
                     "movne %icc, $F, $dst",
                     [(set IntRegs:$dst,
                           (V8selecticc IntRegs:$F, IntRegs:$T, 22, ICC))]>;
  def MOVEQ : Pseudo<(ops IntRegs:$dst, IntRegs:$T, IntRegs:$F),
                     "move %icc, $F, $dst",
                     [(set IntRegs:$dst,
                           (V8selecticc IntRegs:$F, IntRegs:$T, 17, ICC))]>;
}

//===----------------------------------------------------------------------===//
// Non-Instruction Patterns
//===----------------------------------------------------------------------===//

// Small immediates.
def : Pat<(i32 simm13:$val),
          (ORri G0, imm:$val)>;
// Arbitrary immediates.
def : Pat<(i32 imm:$val),
          (ORri (SETHIi (HI22 imm:$val)), (LO10 imm:$val))>;

// Global addresses, constant pool entries
def : Pat<(V8hi tglobaladdr:$in), (SETHIi tglobaladdr:$in)>;
def : Pat<(V8lo tglobaladdr:$in), (ORri G0, tglobaladdr:$in)>;
def : Pat<(V8hi tconstpool:$in), (SETHIi tconstpool:$in)>;
def : Pat<(V8lo tconstpool:$in), (ORri G0, tconstpool:$in)>;

// Add reg, lo.  This is used when taking the addr of a global/constpool entry.
def : Pat<(add IntRegs:$r, (V8lo tglobaladdr:$in)),
          (ADDri IntRegs:$r, tglobaladdr:$in)>;
def : Pat<(add IntRegs:$r, (V8lo tconstpool:$in)),
          (ADDri IntRegs:$r, tconstpool:$in)>;


// Calls: 
def : Pat<(call tglobaladdr:$dst),
          (CALL tglobaladdr:$dst)>;
def : Pat<(call externalsym:$dst),
          (CALL externalsym:$dst)>;

def : Pat<(ret), (RETL)>;

// Map integer extload's to zextloads.
def : Pat<(i32 (extload ADDRrr:$src, i1)), (LDUBrr ADDRrr:$src)>;
def : Pat<(i32 (extload ADDRri:$src, i1)), (LDUBri ADDRri:$src)>;
def : Pat<(i32 (extload ADDRrr:$src, i8)), (LDUBrr ADDRrr:$src)>;
def : Pat<(i32 (extload ADDRri:$src, i8)), (LDUBri ADDRri:$src)>;
def : Pat<(i32 (extload ADDRrr:$src, i16)), (LDUHrr ADDRrr:$src)>;
def : Pat<(i32 (extload ADDRri:$src, i16)), (LDUHri ADDRri:$src)>;

// zextload bool -> zextload byte
def : Pat<(i32 (zextload ADDRrr:$src, i1)), (LDUBrr ADDRrr:$src)>;
def : Pat<(i32 (zextload ADDRri:$src, i1)), (LDUBri ADDRri:$src)>;

// truncstore bool -> truncstore byte.
def : Pat<(truncstore IntRegs:$src, ADDRrr:$addr, i1), 
          (STBrr ADDRrr:$addr, IntRegs:$src)>;
def : Pat<(truncstore IntRegs:$src, ADDRri:$addr, i1), 
          (STBri ADDRri:$addr, IntRegs:$src)>;