github.com/miolini/go@v0.0.0-20160405192216-fca68c8cb408/src/cmd/compile/internal/ssa/gen/AMD64Ops.go (about)

     1  // Copyright 2015 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  // +build ignore
     6  
     7  package main
     8  
     9  import "strings"
    10  
    11  // Notes:
    12  //  - Integer types live in the low portion of registers. Upper portions are junk.
    13  //  - Boolean types use the low-order byte of a register. 0=false, 1=true.
    14  //    Upper bytes are junk.
    15  //  - Floating-point types live in the low natural slot of an sse2 register.
    16  //    Unused portions are junk.
    17  //  - We do not use AH,BH,CH,DH registers.
    18  //  - When doing sub-register operations, we try to write the whole
    19  //    destination register to avoid a partial-register write.
    20  //  - Unused portions of AuxInt (or the Val portion of ValAndOff) are
    21  //    filled by sign-extending the used portion.  Users of AuxInt which interpret
    22  //    AuxInt as unsigned (e.g. shifts) must be careful.
    23  
    24  // Suffixes encode the bit width of various instructions.
    25  // Q (quad word) = 64 bit
    26  // L (long word) = 32 bit
    27  // W (word)      = 16 bit
    28  // B (byte)      = 8 bit
    29  
    30  // copied from ../../amd64/reg.go
    31  var regNamesAMD64 = []string{
    32  	"AX",
    33  	"CX",
    34  	"DX",
    35  	"BX",
    36  	"SP",
    37  	"BP",
    38  	"SI",
    39  	"DI",
    40  	"R8",
    41  	"R9",
    42  	"R10",
    43  	"R11",
    44  	"R12",
    45  	"R13",
    46  	"R14",
    47  	"R15",
    48  	"X0",
    49  	"X1",
    50  	"X2",
    51  	"X3",
    52  	"X4",
    53  	"X5",
    54  	"X6",
    55  	"X7",
    56  	"X8",
    57  	"X9",
    58  	"X10",
    59  	"X11",
    60  	"X12",
    61  	"X13",
    62  	"X14",
    63  	"X15",
    64  
    65  	// pseudo-registers
    66  	"SB",
    67  	"FLAGS",
    68  }
    69  
    70  func init() {
    71  	// Make map from reg names to reg integers.
    72  	if len(regNamesAMD64) > 64 {
    73  		panic("too many registers")
    74  	}
    75  	num := map[string]int{}
    76  	for i, name := range regNamesAMD64 {
    77  		num[name] = i
    78  	}
    79  	buildReg := func(s string) regMask {
    80  		m := regMask(0)
    81  		for _, r := range strings.Split(s, " ") {
    82  			if n, ok := num[r]; ok {
    83  				m |= regMask(1) << uint(n)
    84  				continue
    85  			}
    86  			panic("register " + r + " not found")
    87  		}
    88  		return m
    89  	}
    90  
    91  	// Common individual register masks
    92  	var (
    93  		ax         = buildReg("AX")
    94  		cx         = buildReg("CX")
    95  		dx         = buildReg("DX")
    96  		x15        = buildReg("X15")
    97  		gp         = buildReg("AX CX DX BX BP SI DI R8 R9 R10 R11 R12 R13 R14 R15")
    98  		fp         = buildReg("X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15")
    99  		gpsp       = gp | buildReg("SP")
   100  		gpspsb     = gpsp | buildReg("SB")
   101  		flags      = buildReg("FLAGS")
   102  		callerSave = gp | fp | flags
   103  	)
   104  	// Common slices of register masks
   105  	var (
   106  		gponly    = []regMask{gp}
   107  		fponly    = []regMask{fp}
   108  		flagsonly = []regMask{flags}
   109  	)
   110  
   111  	// Common regInfo
   112  	var (
   113  		gp01      = regInfo{inputs: []regMask{}, outputs: gponly}
   114  		gp11      = regInfo{inputs: []regMask{gpsp}, outputs: gponly, clobbers: flags}
   115  		gp11nf    = regInfo{inputs: []regMask{gpsp}, outputs: gponly} // nf: no flags clobbered
   116  		gp11sb    = regInfo{inputs: []regMask{gpspsb}, outputs: gponly}
   117  		gp21      = regInfo{inputs: []regMask{gpsp, gpsp}, outputs: gponly, clobbers: flags}
   118  		gp21sb    = regInfo{inputs: []regMask{gpspsb, gpsp}, outputs: gponly}
   119  		gp21shift = regInfo{inputs: []regMask{gpsp, cx}, outputs: []regMask{gp &^ cx}, clobbers: flags}
   120  		gp11div   = regInfo{inputs: []regMask{ax, gpsp &^ dx}, outputs: []regMask{ax},
   121  			clobbers: dx | flags}
   122  		gp11hmul = regInfo{inputs: []regMask{ax, gpsp}, outputs: []regMask{dx},
   123  			clobbers: ax | flags}
   124  		gp11mod = regInfo{inputs: []regMask{ax, gpsp &^ dx}, outputs: []regMask{dx},
   125  			clobbers: ax | flags}
   126  
   127  		gp2flags = regInfo{inputs: []regMask{gpsp, gpsp}, outputs: flagsonly}
   128  		gp1flags = regInfo{inputs: []regMask{gpsp}, outputs: flagsonly}
   129  		flagsgp  = regInfo{inputs: flagsonly, outputs: gponly}
   130  
   131  		// for CMOVconst -- uses AX to hold constant temporary. AX input is moved before temp.
   132  		gp1flagsgp = regInfo{inputs: []regMask{gp, flags}, clobbers: ax | flags, outputs: []regMask{gp &^ ax}}
   133  
   134  		readflags = regInfo{inputs: flagsonly, outputs: gponly}
   135  		flagsgpax = regInfo{inputs: flagsonly, clobbers: ax | flags, outputs: []regMask{gp &^ ax}}
   136  
   137  		gpload    = regInfo{inputs: []regMask{gpspsb, 0}, outputs: gponly}
   138  		gploadidx = regInfo{inputs: []regMask{gpspsb, gpsp, 0}, outputs: gponly}
   139  
   140  		gpstore         = regInfo{inputs: []regMask{gpspsb, gpsp, 0}}
   141  		gpstoreconst    = regInfo{inputs: []regMask{gpspsb, 0}}
   142  		gpstoreidx      = regInfo{inputs: []regMask{gpspsb, gpsp, gpsp, 0}}
   143  		gpstoreconstidx = regInfo{inputs: []regMask{gpspsb, gpsp, 0}}
   144  
   145  		fp01    = regInfo{inputs: []regMask{}, outputs: fponly}
   146  		fp21    = regInfo{inputs: []regMask{fp, fp}, outputs: fponly}
   147  		fp21x15 = regInfo{inputs: []regMask{fp &^ x15, fp &^ x15},
   148  			clobbers: x15, outputs: []regMask{fp &^ x15}}
   149  		fpgp     = regInfo{inputs: fponly, outputs: gponly}
   150  		gpfp     = regInfo{inputs: gponly, outputs: fponly}
   151  		fp11     = regInfo{inputs: fponly, outputs: fponly}
   152  		fp2flags = regInfo{inputs: []regMask{fp, fp}, outputs: flagsonly}
   153  
   154  		fpload    = regInfo{inputs: []regMask{gpspsb, 0}, outputs: fponly}
   155  		fploadidx = regInfo{inputs: []regMask{gpspsb, gpsp, 0}, outputs: fponly}
   156  
   157  		fpstore    = regInfo{inputs: []regMask{gpspsb, fp, 0}}
   158  		fpstoreidx = regInfo{inputs: []regMask{gpspsb, gpsp, fp, 0}}
   159  	)
   160  
   161  	var AMD64ops = []opData{
   162  		// fp ops
   163  		{name: "ADDSS", argLength: 2, reg: fp21, asm: "ADDSS", commutative: true, resultInArg0: true}, // fp32 add
   164  		{name: "ADDSD", argLength: 2, reg: fp21, asm: "ADDSD", commutative: true, resultInArg0: true}, // fp64 add
   165  		{name: "SUBSS", argLength: 2, reg: fp21x15, asm: "SUBSS", resultInArg0: true},                 // fp32 sub
   166  		{name: "SUBSD", argLength: 2, reg: fp21x15, asm: "SUBSD", resultInArg0: true},                 // fp64 sub
   167  		{name: "MULSS", argLength: 2, reg: fp21, asm: "MULSS", commutative: true, resultInArg0: true}, // fp32 mul
   168  		{name: "MULSD", argLength: 2, reg: fp21, asm: "MULSD", commutative: true, resultInArg0: true}, // fp64 mul
   169  		{name: "DIVSS", argLength: 2, reg: fp21x15, asm: "DIVSS", resultInArg0: true},                 // fp32 div
   170  		{name: "DIVSD", argLength: 2, reg: fp21x15, asm: "DIVSD", resultInArg0: true},                 // fp64 div
   171  
   172  		{name: "MOVSSload", argLength: 2, reg: fpload, asm: "MOVSS", aux: "SymOff"},            // fp32 load
   173  		{name: "MOVSDload", argLength: 2, reg: fpload, asm: "MOVSD", aux: "SymOff"},            // fp64 load
   174  		{name: "MOVSSconst", reg: fp01, asm: "MOVSS", aux: "Float32", rematerializeable: true}, // fp32 constant
   175  		{name: "MOVSDconst", reg: fp01, asm: "MOVSD", aux: "Float64", rematerializeable: true}, // fp64 constant
   176  		{name: "MOVSSloadidx1", argLength: 3, reg: fploadidx, asm: "MOVSS", aux: "SymOff"},     // fp32 load indexed by i
   177  		{name: "MOVSSloadidx4", argLength: 3, reg: fploadidx, asm: "MOVSS", aux: "SymOff"},     // fp32 load indexed by 4*i
   178  		{name: "MOVSDloadidx1", argLength: 3, reg: fploadidx, asm: "MOVSD", aux: "SymOff"},     // fp64 load indexed by i
   179  		{name: "MOVSDloadidx8", argLength: 3, reg: fploadidx, asm: "MOVSD", aux: "SymOff"},     // fp64 load indexed by 8*i
   180  
   181  		{name: "MOVSSstore", argLength: 3, reg: fpstore, asm: "MOVSS", aux: "SymOff"},        // fp32 store
   182  		{name: "MOVSDstore", argLength: 3, reg: fpstore, asm: "MOVSD", aux: "SymOff"},        // fp64 store
   183  		{name: "MOVSSstoreidx1", argLength: 4, reg: fpstoreidx, asm: "MOVSS", aux: "SymOff"}, // fp32 indexed by i store
   184  		{name: "MOVSSstoreidx4", argLength: 4, reg: fpstoreidx, asm: "MOVSS", aux: "SymOff"}, // fp32 indexed by 4i store
   185  		{name: "MOVSDstoreidx1", argLength: 4, reg: fpstoreidx, asm: "MOVSD", aux: "SymOff"}, // fp64 indexed by i store
   186  		{name: "MOVSDstoreidx8", argLength: 4, reg: fpstoreidx, asm: "MOVSD", aux: "SymOff"}, // fp64 indexed by 8i store
   187  
   188  		// binary ops
   189  		{name: "ADDQ", argLength: 2, reg: gp21, asm: "ADDQ", commutative: true, resultInArg0: true},                // arg0 + arg1
   190  		{name: "ADDL", argLength: 2, reg: gp21, asm: "ADDL", commutative: true, resultInArg0: true},                // arg0 + arg1
   191  		{name: "ADDW", argLength: 2, reg: gp21, asm: "ADDL", commutative: true, resultInArg0: true},                // arg0 + arg1
   192  		{name: "ADDB", argLength: 2, reg: gp21, asm: "ADDL", commutative: true, resultInArg0: true},                // arg0 + arg1
   193  		{name: "ADDQconst", argLength: 1, reg: gp11, asm: "ADDQ", aux: "Int64", resultInArg0: true, typ: "UInt64"}, // arg0 + auxint
   194  		{name: "ADDLconst", argLength: 1, reg: gp11, asm: "ADDL", aux: "Int32", resultInArg0: true},                // arg0 + auxint
   195  		{name: "ADDWconst", argLength: 1, reg: gp11, asm: "ADDL", aux: "Int16", resultInArg0: true},                // arg0 + auxint
   196  		{name: "ADDBconst", argLength: 1, reg: gp11, asm: "ADDL", aux: "Int8", resultInArg0: true},                 // arg0 + auxint
   197  
   198  		{name: "SUBQ", argLength: 2, reg: gp21, asm: "SUBQ", resultInArg0: true},                    // arg0 - arg1
   199  		{name: "SUBL", argLength: 2, reg: gp21, asm: "SUBL", resultInArg0: true},                    // arg0 - arg1
   200  		{name: "SUBW", argLength: 2, reg: gp21, asm: "SUBL", resultInArg0: true},                    // arg0 - arg1
   201  		{name: "SUBB", argLength: 2, reg: gp21, asm: "SUBL", resultInArg0: true},                    // arg0 - arg1
   202  		{name: "SUBQconst", argLength: 1, reg: gp11, asm: "SUBQ", aux: "Int64", resultInArg0: true}, // arg0 - auxint
   203  		{name: "SUBLconst", argLength: 1, reg: gp11, asm: "SUBL", aux: "Int32", resultInArg0: true}, // arg0 - auxint
   204  		{name: "SUBWconst", argLength: 1, reg: gp11, asm: "SUBL", aux: "Int16", resultInArg0: true}, // arg0 - auxint
   205  		{name: "SUBBconst", argLength: 1, reg: gp11, asm: "SUBL", aux: "Int8", resultInArg0: true},  // arg0 - auxint
   206  
   207  		{name: "MULQ", argLength: 2, reg: gp21, asm: "IMULQ", commutative: true, resultInArg0: true}, // arg0 * arg1
   208  		{name: "MULL", argLength: 2, reg: gp21, asm: "IMULL", commutative: true, resultInArg0: true}, // arg0 * arg1
   209  		{name: "MULW", argLength: 2, reg: gp21, asm: "IMULW", commutative: true, resultInArg0: true}, // arg0 * arg1
   210  		{name: "MULB", argLength: 2, reg: gp21, asm: "IMULW", commutative: true, resultInArg0: true}, // arg0 * arg1
   211  		{name: "MULQconst", argLength: 1, reg: gp11, asm: "IMULQ", aux: "Int64", resultInArg0: true}, // arg0 * auxint
   212  		{name: "MULLconst", argLength: 1, reg: gp11, asm: "IMULL", aux: "Int32", resultInArg0: true}, // arg0 * auxint
   213  		{name: "MULWconst", argLength: 1, reg: gp11, asm: "IMULW", aux: "Int16", resultInArg0: true}, // arg0 * auxint
   214  		{name: "MULBconst", argLength: 1, reg: gp11, asm: "IMULW", aux: "Int8", resultInArg0: true},  // arg0 * auxint
   215  
   216  		{name: "HMULQ", argLength: 2, reg: gp11hmul, asm: "IMULQ"}, // (arg0 * arg1) >> width
   217  		{name: "HMULL", argLength: 2, reg: gp11hmul, asm: "IMULL"}, // (arg0 * arg1) >> width
   218  		{name: "HMULW", argLength: 2, reg: gp11hmul, asm: "IMULW"}, // (arg0 * arg1) >> width
   219  		{name: "HMULB", argLength: 2, reg: gp11hmul, asm: "IMULB"}, // (arg0 * arg1) >> width
   220  		{name: "HMULQU", argLength: 2, reg: gp11hmul, asm: "MULQ"}, // (arg0 * arg1) >> width
   221  		{name: "HMULLU", argLength: 2, reg: gp11hmul, asm: "MULL"}, // (arg0 * arg1) >> width
   222  		{name: "HMULWU", argLength: 2, reg: gp11hmul, asm: "MULW"}, // (arg0 * arg1) >> width
   223  		{name: "HMULBU", argLength: 2, reg: gp11hmul, asm: "MULB"}, // (arg0 * arg1) >> width
   224  
   225  		{name: "AVGQU", argLength: 2, reg: gp21, commutative: true, resultInArg0: true}, // (arg0 + arg1) / 2 as unsigned, all 64 result bits
   226  
   227  		{name: "DIVQ", argLength: 2, reg: gp11div, asm: "IDIVQ"}, // arg0 / arg1
   228  		{name: "DIVL", argLength: 2, reg: gp11div, asm: "IDIVL"}, // arg0 / arg1
   229  		{name: "DIVW", argLength: 2, reg: gp11div, asm: "IDIVW"}, // arg0 / arg1
   230  		{name: "DIVQU", argLength: 2, reg: gp11div, asm: "DIVQ"}, // arg0 / arg1
   231  		{name: "DIVLU", argLength: 2, reg: gp11div, asm: "DIVL"}, // arg0 / arg1
   232  		{name: "DIVWU", argLength: 2, reg: gp11div, asm: "DIVW"}, // arg0 / arg1
   233  
   234  		{name: "MODQ", argLength: 2, reg: gp11mod, asm: "IDIVQ"}, // arg0 % arg1
   235  		{name: "MODL", argLength: 2, reg: gp11mod, asm: "IDIVL"}, // arg0 % arg1
   236  		{name: "MODW", argLength: 2, reg: gp11mod, asm: "IDIVW"}, // arg0 % arg1
   237  		{name: "MODQU", argLength: 2, reg: gp11mod, asm: "DIVQ"}, // arg0 % arg1
   238  		{name: "MODLU", argLength: 2, reg: gp11mod, asm: "DIVL"}, // arg0 % arg1
   239  		{name: "MODWU", argLength: 2, reg: gp11mod, asm: "DIVW"}, // arg0 % arg1
   240  
   241  		{name: "ANDQ", argLength: 2, reg: gp21, asm: "ANDQ", commutative: true, resultInArg0: true}, // arg0 & arg1
   242  		{name: "ANDL", argLength: 2, reg: gp21, asm: "ANDL", commutative: true, resultInArg0: true}, // arg0 & arg1
   243  		{name: "ANDW", argLength: 2, reg: gp21, asm: "ANDL", commutative: true, resultInArg0: true}, // arg0 & arg1
   244  		{name: "ANDB", argLength: 2, reg: gp21, asm: "ANDL", commutative: true, resultInArg0: true}, // arg0 & arg1
   245  		{name: "ANDQconst", argLength: 1, reg: gp11, asm: "ANDQ", aux: "Int64", resultInArg0: true}, // arg0 & auxint
   246  		{name: "ANDLconst", argLength: 1, reg: gp11, asm: "ANDL", aux: "Int32", resultInArg0: true}, // arg0 & auxint
   247  		{name: "ANDWconst", argLength: 1, reg: gp11, asm: "ANDL", aux: "Int16", resultInArg0: true}, // arg0 & auxint
   248  		{name: "ANDBconst", argLength: 1, reg: gp11, asm: "ANDL", aux: "Int8", resultInArg0: true},  // arg0 & auxint
   249  
   250  		{name: "ORQ", argLength: 2, reg: gp21, asm: "ORQ", commutative: true, resultInArg0: true}, // arg0 | arg1
   251  		{name: "ORL", argLength: 2, reg: gp21, asm: "ORL", commutative: true, resultInArg0: true}, // arg0 | arg1
   252  		{name: "ORW", argLength: 2, reg: gp21, asm: "ORL", commutative: true, resultInArg0: true}, // arg0 | arg1
   253  		{name: "ORB", argLength: 2, reg: gp21, asm: "ORL", commutative: true, resultInArg0: true}, // arg0 | arg1
   254  		{name: "ORQconst", argLength: 1, reg: gp11, asm: "ORQ", aux: "Int64", resultInArg0: true}, // arg0 | auxint
   255  		{name: "ORLconst", argLength: 1, reg: gp11, asm: "ORL", aux: "Int32", resultInArg0: true}, // arg0 | auxint
   256  		{name: "ORWconst", argLength: 1, reg: gp11, asm: "ORL", aux: "Int16", resultInArg0: true}, // arg0 | auxint
   257  		{name: "ORBconst", argLength: 1, reg: gp11, asm: "ORL", aux: "Int8", resultInArg0: true},  // arg0 | auxint
   258  
   259  		{name: "XORQ", argLength: 2, reg: gp21, asm: "XORQ", commutative: true, resultInArg0: true}, // arg0 ^ arg1
   260  		{name: "XORL", argLength: 2, reg: gp21, asm: "XORL", commutative: true, resultInArg0: true}, // arg0 ^ arg1
   261  		{name: "XORW", argLength: 2, reg: gp21, asm: "XORL", commutative: true, resultInArg0: true}, // arg0 ^ arg1
   262  		{name: "XORB", argLength: 2, reg: gp21, asm: "XORL", commutative: true, resultInArg0: true}, // arg0 ^ arg1
   263  		{name: "XORQconst", argLength: 1, reg: gp11, asm: "XORQ", aux: "Int64", resultInArg0: true}, // arg0 ^ auxint
   264  		{name: "XORLconst", argLength: 1, reg: gp11, asm: "XORL", aux: "Int32", resultInArg0: true}, // arg0 ^ auxint
   265  		{name: "XORWconst", argLength: 1, reg: gp11, asm: "XORL", aux: "Int16", resultInArg0: true}, // arg0 ^ auxint
   266  		{name: "XORBconst", argLength: 1, reg: gp11, asm: "XORL", aux: "Int8", resultInArg0: true},  // arg0 ^ auxint
   267  
   268  		{name: "CMPQ", argLength: 2, reg: gp2flags, asm: "CMPQ", typ: "Flags"},                    // arg0 compare to arg1
   269  		{name: "CMPL", argLength: 2, reg: gp2flags, asm: "CMPL", typ: "Flags"},                    // arg0 compare to arg1
   270  		{name: "CMPW", argLength: 2, reg: gp2flags, asm: "CMPW", typ: "Flags"},                    // arg0 compare to arg1
   271  		{name: "CMPB", argLength: 2, reg: gp2flags, asm: "CMPB", typ: "Flags"},                    // arg0 compare to arg1
   272  		{name: "CMPQconst", argLength: 1, reg: gp1flags, asm: "CMPQ", typ: "Flags", aux: "Int64"}, // arg0 compare to auxint
   273  		{name: "CMPLconst", argLength: 1, reg: gp1flags, asm: "CMPL", typ: "Flags", aux: "Int32"}, // arg0 compare to auxint
   274  		{name: "CMPWconst", argLength: 1, reg: gp1flags, asm: "CMPW", typ: "Flags", aux: "Int16"}, // arg0 compare to auxint
   275  		{name: "CMPBconst", argLength: 1, reg: gp1flags, asm: "CMPB", typ: "Flags", aux: "Int8"},  // arg0 compare to auxint
   276  
   277  		{name: "UCOMISS", argLength: 2, reg: fp2flags, asm: "UCOMISS", typ: "Flags"}, // arg0 compare to arg1, f32
   278  		{name: "UCOMISD", argLength: 2, reg: fp2flags, asm: "UCOMISD", typ: "Flags"}, // arg0 compare to arg1, f64
   279  
   280  		{name: "TESTQ", argLength: 2, reg: gp2flags, asm: "TESTQ", typ: "Flags"},                    // (arg0 & arg1) compare to 0
   281  		{name: "TESTL", argLength: 2, reg: gp2flags, asm: "TESTL", typ: "Flags"},                    // (arg0 & arg1) compare to 0
   282  		{name: "TESTW", argLength: 2, reg: gp2flags, asm: "TESTW", typ: "Flags"},                    // (arg0 & arg1) compare to 0
   283  		{name: "TESTB", argLength: 2, reg: gp2flags, asm: "TESTB", typ: "Flags"},                    // (arg0 & arg1) compare to 0
   284  		{name: "TESTQconst", argLength: 1, reg: gp1flags, asm: "TESTQ", typ: "Flags", aux: "Int64"}, // (arg0 & auxint) compare to 0
   285  		{name: "TESTLconst", argLength: 1, reg: gp1flags, asm: "TESTL", typ: "Flags", aux: "Int32"}, // (arg0 & auxint) compare to 0
   286  		{name: "TESTWconst", argLength: 1, reg: gp1flags, asm: "TESTW", typ: "Flags", aux: "Int16"}, // (arg0 & auxint) compare to 0
   287  		{name: "TESTBconst", argLength: 1, reg: gp1flags, asm: "TESTB", typ: "Flags", aux: "Int8"},  // (arg0 & auxint) compare to 0
   288  
   289  		{name: "SHLQ", argLength: 2, reg: gp21shift, asm: "SHLQ", resultInArg0: true},               // arg0 << arg1, shift amount is mod 64
   290  		{name: "SHLL", argLength: 2, reg: gp21shift, asm: "SHLL", resultInArg0: true},               // arg0 << arg1, shift amount is mod 32
   291  		{name: "SHLW", argLength: 2, reg: gp21shift, asm: "SHLL", resultInArg0: true},               // arg0 << arg1, shift amount is mod 32
   292  		{name: "SHLB", argLength: 2, reg: gp21shift, asm: "SHLL", resultInArg0: true},               // arg0 << arg1, shift amount is mod 32
   293  		{name: "SHLQconst", argLength: 1, reg: gp11, asm: "SHLQ", aux: "Int64", resultInArg0: true}, // arg0 << auxint, shift amount 0-63
   294  		{name: "SHLLconst", argLength: 1, reg: gp11, asm: "SHLL", aux: "Int32", resultInArg0: true}, // arg0 << auxint, shift amount 0-31
   295  		{name: "SHLWconst", argLength: 1, reg: gp11, asm: "SHLL", aux: "Int16", resultInArg0: true}, // arg0 << auxint, shift amount 0-31
   296  		{name: "SHLBconst", argLength: 1, reg: gp11, asm: "SHLL", aux: "Int8", resultInArg0: true},  // arg0 << auxint, shift amount 0-31
   297  		// Note: x86 is weird, the 16 and 8 byte shifts still use all 5 bits of shift amount!
   298  
   299  		{name: "SHRQ", argLength: 2, reg: gp21shift, asm: "SHRQ", resultInArg0: true},               // unsigned arg0 >> arg1, shift amount is mod 64
   300  		{name: "SHRL", argLength: 2, reg: gp21shift, asm: "SHRL", resultInArg0: true},               // unsigned arg0 >> arg1, shift amount is mod 32
   301  		{name: "SHRW", argLength: 2, reg: gp21shift, asm: "SHRW", resultInArg0: true},               // unsigned arg0 >> arg1, shift amount is mod 32
   302  		{name: "SHRB", argLength: 2, reg: gp21shift, asm: "SHRB", resultInArg0: true},               // unsigned arg0 >> arg1, shift amount is mod 32
   303  		{name: "SHRQconst", argLength: 1, reg: gp11, asm: "SHRQ", aux: "Int64", resultInArg0: true}, // unsigned arg0 >> auxint, shift amount 0-63
   304  		{name: "SHRLconst", argLength: 1, reg: gp11, asm: "SHRL", aux: "Int32", resultInArg0: true}, // unsigned arg0 >> auxint, shift amount 0-31
   305  		{name: "SHRWconst", argLength: 1, reg: gp11, asm: "SHRW", aux: "Int16", resultInArg0: true}, // unsigned arg0 >> auxint, shift amount 0-31
   306  		{name: "SHRBconst", argLength: 1, reg: gp11, asm: "SHRB", aux: "Int8", resultInArg0: true},  // unsigned arg0 >> auxint, shift amount 0-31
   307  
   308  		{name: "SARQ", argLength: 2, reg: gp21shift, asm: "SARQ", resultInArg0: true},               // signed arg0 >> arg1, shift amount is mod 64
   309  		{name: "SARL", argLength: 2, reg: gp21shift, asm: "SARL", resultInArg0: true},               // signed arg0 >> arg1, shift amount is mod 32
   310  		{name: "SARW", argLength: 2, reg: gp21shift, asm: "SARW", resultInArg0: true},               // signed arg0 >> arg1, shift amount is mod 32
   311  		{name: "SARB", argLength: 2, reg: gp21shift, asm: "SARB", resultInArg0: true},               // signed arg0 >> arg1, shift amount is mod 32
   312  		{name: "SARQconst", argLength: 1, reg: gp11, asm: "SARQ", aux: "Int64", resultInArg0: true}, // signed arg0 >> auxint, shift amount 0-63
   313  		{name: "SARLconst", argLength: 1, reg: gp11, asm: "SARL", aux: "Int32", resultInArg0: true}, // signed arg0 >> auxint, shift amount 0-31
   314  		{name: "SARWconst", argLength: 1, reg: gp11, asm: "SARW", aux: "Int16", resultInArg0: true}, // signed arg0 >> auxint, shift amount 0-31
   315  		{name: "SARBconst", argLength: 1, reg: gp11, asm: "SARB", aux: "Int8", resultInArg0: true},  // signed arg0 >> auxint, shift amount 0-31
   316  
   317  		{name: "ROLQconst", argLength: 1, reg: gp11, asm: "ROLQ", aux: "Int64", resultInArg0: true}, // arg0 rotate left auxint, rotate amount 0-63
   318  		{name: "ROLLconst", argLength: 1, reg: gp11, asm: "ROLL", aux: "Int32", resultInArg0: true}, // arg0 rotate left auxint, rotate amount 0-31
   319  		{name: "ROLWconst", argLength: 1, reg: gp11, asm: "ROLW", aux: "Int16", resultInArg0: true}, // arg0 rotate left auxint, rotate amount 0-15
   320  		{name: "ROLBconst", argLength: 1, reg: gp11, asm: "ROLB", aux: "Int8", resultInArg0: true},  // arg0 rotate left auxint, rotate amount 0-7
   321  
   322  		// unary ops
   323  		{name: "NEGQ", argLength: 1, reg: gp11, asm: "NEGQ", resultInArg0: true}, // -arg0
   324  		{name: "NEGL", argLength: 1, reg: gp11, asm: "NEGL", resultInArg0: true}, // -arg0
   325  		{name: "NEGW", argLength: 1, reg: gp11, asm: "NEGL", resultInArg0: true}, // -arg0
   326  		{name: "NEGB", argLength: 1, reg: gp11, asm: "NEGL", resultInArg0: true}, // -arg0
   327  
   328  		{name: "NOTQ", argLength: 1, reg: gp11, asm: "NOTQ", resultInArg0: true}, // ^arg0
   329  		{name: "NOTL", argLength: 1, reg: gp11, asm: "NOTL", resultInArg0: true}, // ^arg0
   330  		{name: "NOTW", argLength: 1, reg: gp11, asm: "NOTL", resultInArg0: true}, // ^arg0
   331  		{name: "NOTB", argLength: 1, reg: gp11, asm: "NOTL", resultInArg0: true}, // ^arg0
   332  
   333  		{name: "BSFQ", argLength: 1, reg: gp11, asm: "BSFQ"}, // arg0 # of low-order zeroes ; undef if zero
   334  		{name: "BSFL", argLength: 1, reg: gp11, asm: "BSFL"}, // arg0 # of low-order zeroes ; undef if zero
   335  		{name: "BSFW", argLength: 1, reg: gp11, asm: "BSFW"}, // arg0 # of low-order zeroes ; undef if zero
   336  
   337  		{name: "BSRQ", argLength: 1, reg: gp11, asm: "BSRQ"}, // arg0 # of high-order zeroes ; undef if zero
   338  		{name: "BSRL", argLength: 1, reg: gp11, asm: "BSRL"}, // arg0 # of high-order zeroes ; undef if zero
   339  		{name: "BSRW", argLength: 1, reg: gp11, asm: "BSRW"}, // arg0 # of high-order zeroes ; undef if zero
   340  
   341  		// Note ASM for ops moves whole register
   342  		{name: "CMOVQEQconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVQEQ", typ: "UInt64", aux: "Int64", resultInArg0: true}, // replace arg0 w/ constant if Z set
   343  		{name: "CMOVLEQconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVLEQ", typ: "UInt32", aux: "Int32", resultInArg0: true}, // replace arg0 w/ constant if Z set
   344  		{name: "CMOVWEQconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVLEQ", typ: "UInt16", aux: "Int16", resultInArg0: true}, // replace arg0 w/ constant if Z set
   345  		{name: "CMOVQNEconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVQNE", typ: "UInt64", aux: "Int64", resultInArg0: true}, // replace arg0 w/ constant if Z not set
   346  		{name: "CMOVLNEconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVLNE", typ: "UInt32", aux: "Int32", resultInArg0: true}, // replace arg0 w/ constant if Z not set
   347  		{name: "CMOVWNEconst", argLength: 2, reg: gp1flagsgp, asm: "CMOVLNE", typ: "UInt16", aux: "Int16", resultInArg0: true}, // replace arg0 w/ constant if Z not set
   348  
   349  		{name: "BSWAPQ", argLength: 1, reg: gp11, asm: "BSWAPQ", resultInArg0: true}, // arg0 swap bytes
   350  		{name: "BSWAPL", argLength: 1, reg: gp11, asm: "BSWAPL", resultInArg0: true}, // arg0 swap bytes
   351  
   352  		{name: "SQRTSD", argLength: 1, reg: fp11, asm: "SQRTSD"}, // sqrt(arg0)
   353  
   354  		{name: "SBBQcarrymask", argLength: 1, reg: flagsgp, asm: "SBBQ"}, // (int64)(-1) if carry is set, 0 if carry is clear.
   355  		{name: "SBBLcarrymask", argLength: 1, reg: flagsgp, asm: "SBBL"}, // (int32)(-1) if carry is set, 0 if carry is clear.
   356  		// Note: SBBW and SBBB are subsumed by SBBL
   357  
   358  		{name: "SETEQ", argLength: 1, reg: readflags, asm: "SETEQ"}, // extract == condition from arg0
   359  		{name: "SETNE", argLength: 1, reg: readflags, asm: "SETNE"}, // extract != condition from arg0
   360  		{name: "SETL", argLength: 1, reg: readflags, asm: "SETLT"},  // extract signed < condition from arg0
   361  		{name: "SETLE", argLength: 1, reg: readflags, asm: "SETLE"}, // extract signed <= condition from arg0
   362  		{name: "SETG", argLength: 1, reg: readflags, asm: "SETGT"},  // extract signed > condition from arg0
   363  		{name: "SETGE", argLength: 1, reg: readflags, asm: "SETGE"}, // extract signed >= condition from arg0
   364  		{name: "SETB", argLength: 1, reg: readflags, asm: "SETCS"},  // extract unsigned < condition from arg0
   365  		{name: "SETBE", argLength: 1, reg: readflags, asm: "SETLS"}, // extract unsigned <= condition from arg0
   366  		{name: "SETA", argLength: 1, reg: readflags, asm: "SETHI"},  // extract unsigned > condition from arg0
   367  		{name: "SETAE", argLength: 1, reg: readflags, asm: "SETCC"}, // extract unsigned >= condition from arg0
   368  		// Need different opcodes for floating point conditions because
   369  		// any comparison involving a NaN is always FALSE and thus
   370  		// the patterns for inverting conditions cannot be used.
   371  		{name: "SETEQF", argLength: 1, reg: flagsgpax, asm: "SETEQ"}, // extract == condition from arg0
   372  		{name: "SETNEF", argLength: 1, reg: flagsgpax, asm: "SETNE"}, // extract != condition from arg0
   373  		{name: "SETORD", argLength: 1, reg: flagsgp, asm: "SETPC"},   // extract "ordered" (No Nan present) condition from arg0
   374  		{name: "SETNAN", argLength: 1, reg: flagsgp, asm: "SETPS"},   // extract "unordered" (Nan present) condition from arg0
   375  
   376  		{name: "SETGF", argLength: 1, reg: flagsgp, asm: "SETHI"},  // extract floating > condition from arg0
   377  		{name: "SETGEF", argLength: 1, reg: flagsgp, asm: "SETCC"}, // extract floating >= condition from arg0
   378  
   379  		{name: "MOVBQSX", argLength: 1, reg: gp11nf, asm: "MOVBQSX"}, // sign extend arg0 from int8 to int64
   380  		{name: "MOVBQZX", argLength: 1, reg: gp11nf, asm: "MOVBQZX"}, // zero extend arg0 from int8 to int64
   381  		{name: "MOVWQSX", argLength: 1, reg: gp11nf, asm: "MOVWQSX"}, // sign extend arg0 from int16 to int64
   382  		{name: "MOVWQZX", argLength: 1, reg: gp11nf, asm: "MOVWQZX"}, // zero extend arg0 from int16 to int64
   383  		{name: "MOVLQSX", argLength: 1, reg: gp11nf, asm: "MOVLQSX"}, // sign extend arg0 from int32 to int64
   384  		{name: "MOVLQZX", argLength: 1, reg: gp11nf, asm: "MOVLQZX"}, // zero extend arg0 from int32 to int64
   385  
   386  		{name: "MOVBconst", reg: gp01, asm: "MOVB", typ: "UInt8", aux: "Int8", rematerializeable: true},   // 8 low bits of auxint
   387  		{name: "MOVWconst", reg: gp01, asm: "MOVW", typ: "UInt16", aux: "Int16", rematerializeable: true}, // 16 low bits of auxint
   388  		{name: "MOVLconst", reg: gp01, asm: "MOVL", typ: "UInt32", aux: "Int32", rematerializeable: true}, // 32 low bits of auxint
   389  		{name: "MOVQconst", reg: gp01, asm: "MOVQ", typ: "UInt64", aux: "Int64", rematerializeable: true}, // auxint
   390  
   391  		{name: "CVTTSD2SL", argLength: 1, reg: fpgp, asm: "CVTTSD2SL"}, // convert float64 to int32
   392  		{name: "CVTTSD2SQ", argLength: 1, reg: fpgp, asm: "CVTTSD2SQ"}, // convert float64 to int64
   393  		{name: "CVTTSS2SL", argLength: 1, reg: fpgp, asm: "CVTTSS2SL"}, // convert float32 to int32
   394  		{name: "CVTTSS2SQ", argLength: 1, reg: fpgp, asm: "CVTTSS2SQ"}, // convert float32 to int64
   395  		{name: "CVTSL2SS", argLength: 1, reg: gpfp, asm: "CVTSL2SS"},   // convert int32 to float32
   396  		{name: "CVTSL2SD", argLength: 1, reg: gpfp, asm: "CVTSL2SD"},   // convert int32 to float64
   397  		{name: "CVTSQ2SS", argLength: 1, reg: gpfp, asm: "CVTSQ2SS"},   // convert int64 to float32
   398  		{name: "CVTSQ2SD", argLength: 1, reg: gpfp, asm: "CVTSQ2SD"},   // convert int64 to float64
   399  		{name: "CVTSD2SS", argLength: 1, reg: fp11, asm: "CVTSD2SS"},   // convert float64 to float32
   400  		{name: "CVTSS2SD", argLength: 1, reg: fp11, asm: "CVTSS2SD"},   // convert float32 to float64
   401  
   402  		{name: "PXOR", argLength: 2, reg: fp21, asm: "PXOR", commutative: true, resultInArg0: true}, // exclusive or, applied to X regs for float negation.
   403  
   404  		{name: "LEAQ", argLength: 1, reg: gp11sb, aux: "SymOff", rematerializeable: true}, // arg0 + auxint + offset encoded in aux
   405  		{name: "LEAQ1", argLength: 2, reg: gp21sb, aux: "SymOff"},                         // arg0 + arg1 + auxint + aux
   406  		{name: "LEAQ2", argLength: 2, reg: gp21sb, aux: "SymOff"},                         // arg0 + 2*arg1 + auxint + aux
   407  		{name: "LEAQ4", argLength: 2, reg: gp21sb, aux: "SymOff"},                         // arg0 + 4*arg1 + auxint + aux
   408  		{name: "LEAQ8", argLength: 2, reg: gp21sb, aux: "SymOff"},                         // arg0 + 8*arg1 + auxint + aux
   409  		// Note: LEAQ{1,2,4,8} must not have OpSB as either argument.
   410  
   411  		// auxint+aux == add auxint and the offset of the symbol in aux (if any) to the effective address
   412  		{name: "MOVBload", argLength: 2, reg: gpload, asm: "MOVBLZX", aux: "SymOff", typ: "UInt8"},  // load byte from arg0+auxint+aux. arg1=mem.  Zero extend.
   413  		{name: "MOVBQSXload", argLength: 2, reg: gpload, asm: "MOVBQSX", aux: "SymOff"},             // ditto, sign extend to int64
   414  		{name: "MOVWload", argLength: 2, reg: gpload, asm: "MOVWLZX", aux: "SymOff", typ: "UInt16"}, // load 2 bytes from arg0+auxint+aux. arg1=mem.  Zero extend.
   415  		{name: "MOVWQSXload", argLength: 2, reg: gpload, asm: "MOVWQSX", aux: "SymOff"},             // ditto, sign extend to int64
   416  		{name: "MOVLload", argLength: 2, reg: gpload, asm: "MOVL", aux: "SymOff", typ: "UInt32"},    // load 4 bytes from arg0+auxint+aux. arg1=mem.  Zero extend.
   417  		{name: "MOVLQSXload", argLength: 2, reg: gpload, asm: "MOVLQSX", aux: "SymOff"},             // ditto, sign extend to int64
   418  		{name: "MOVQload", argLength: 2, reg: gpload, asm: "MOVQ", aux: "SymOff", typ: "UInt64"},    // load 8 bytes from arg0+auxint+aux. arg1=mem
   419  		{name: "MOVBstore", argLength: 3, reg: gpstore, asm: "MOVB", aux: "SymOff", typ: "Mem"},     // store byte in arg1 to arg0+auxint+aux. arg2=mem
   420  		{name: "MOVWstore", argLength: 3, reg: gpstore, asm: "MOVW", aux: "SymOff", typ: "Mem"},     // store 2 bytes in arg1 to arg0+auxint+aux. arg2=mem
   421  		{name: "MOVLstore", argLength: 3, reg: gpstore, asm: "MOVL", aux: "SymOff", typ: "Mem"},     // store 4 bytes in arg1 to arg0+auxint+aux. arg2=mem
   422  		{name: "MOVQstore", argLength: 3, reg: gpstore, asm: "MOVQ", aux: "SymOff", typ: "Mem"},     // store 8 bytes in arg1 to arg0+auxint+aux. arg2=mem
   423  		{name: "MOVOload", argLength: 2, reg: fpload, asm: "MOVUPS", aux: "SymOff", typ: "Int128"},  // load 16 bytes from arg0+auxint+aux. arg1=mem
   424  		{name: "MOVOstore", argLength: 3, reg: fpstore, asm: "MOVUPS", aux: "SymOff", typ: "Mem"},   // store 16 bytes in arg1 to arg0+auxint+aux. arg2=mem
   425  
   426  		// indexed loads/stores
   427  		{name: "MOVBloadidx1", argLength: 3, reg: gploadidx, asm: "MOVBLZX", aux: "SymOff"}, // load a byte from arg0+arg1+auxint+aux. arg2=mem
   428  		{name: "MOVWloadidx1", argLength: 3, reg: gploadidx, asm: "MOVWLZX", aux: "SymOff"}, // load 2 bytes from arg0+arg1+auxint+aux. arg2=mem
   429  		{name: "MOVWloadidx2", argLength: 3, reg: gploadidx, asm: "MOVWLZX", aux: "SymOff"}, // load 2 bytes from arg0+2*arg1+auxint+aux. arg2=mem
   430  		{name: "MOVLloadidx1", argLength: 3, reg: gploadidx, asm: "MOVL", aux: "SymOff"},    // load 4 bytes from arg0+arg1+auxint+aux. arg2=mem
   431  		{name: "MOVLloadidx4", argLength: 3, reg: gploadidx, asm: "MOVL", aux: "SymOff"},    // load 4 bytes from arg0+4*arg1+auxint+aux. arg2=mem
   432  		{name: "MOVQloadidx1", argLength: 3, reg: gploadidx, asm: "MOVQ", aux: "SymOff"},    // load 8 bytes from arg0+arg1+auxint+aux. arg2=mem
   433  		{name: "MOVQloadidx8", argLength: 3, reg: gploadidx, asm: "MOVQ", aux: "SymOff"},    // load 8 bytes from arg0+8*arg1+auxint+aux. arg2=mem
   434  		// TODO: sign-extending indexed loads
   435  		{name: "MOVBstoreidx1", argLength: 4, reg: gpstoreidx, asm: "MOVB", aux: "SymOff"}, // store byte in arg2 to arg0+arg1+auxint+aux. arg3=mem
   436  		{name: "MOVWstoreidx1", argLength: 4, reg: gpstoreidx, asm: "MOVW", aux: "SymOff"}, // store 2 bytes in arg2 to arg0+arg1+auxint+aux. arg3=mem
   437  		{name: "MOVWstoreidx2", argLength: 4, reg: gpstoreidx, asm: "MOVW", aux: "SymOff"}, // store 2 bytes in arg2 to arg0+2*arg1+auxint+aux. arg3=mem
   438  		{name: "MOVLstoreidx1", argLength: 4, reg: gpstoreidx, asm: "MOVL", aux: "SymOff"}, // store 4 bytes in arg2 to arg0+arg1+auxint+aux. arg3=mem
   439  		{name: "MOVLstoreidx4", argLength: 4, reg: gpstoreidx, asm: "MOVL", aux: "SymOff"}, // store 4 bytes in arg2 to arg0+4*arg1+auxint+aux. arg3=mem
   440  		{name: "MOVQstoreidx1", argLength: 4, reg: gpstoreidx, asm: "MOVQ", aux: "SymOff"}, // store 8 bytes in arg2 to arg0+arg1+auxint+aux. arg3=mem
   441  		{name: "MOVQstoreidx8", argLength: 4, reg: gpstoreidx, asm: "MOVQ", aux: "SymOff"}, // store 8 bytes in arg2 to arg0+8*arg1+auxint+aux. arg3=mem
   442  		// TODO: add size-mismatched indexed loads, like MOVBstoreidx4.
   443  
   444  		// For storeconst ops, the AuxInt field encodes both
   445  		// the value to store and an address offset of the store.
   446  		// Cast AuxInt to a ValAndOff to extract Val and Off fields.
   447  		{name: "MOVBstoreconst", argLength: 2, reg: gpstoreconst, asm: "MOVB", aux: "SymValAndOff", typ: "Mem"}, // store low byte of ValAndOff(AuxInt).Val() to arg0+ValAndOff(AuxInt).Off()+aux.  arg1=mem
   448  		{name: "MOVWstoreconst", argLength: 2, reg: gpstoreconst, asm: "MOVW", aux: "SymValAndOff", typ: "Mem"}, // store low 2 bytes of ...
   449  		{name: "MOVLstoreconst", argLength: 2, reg: gpstoreconst, asm: "MOVL", aux: "SymValAndOff", typ: "Mem"}, // store low 4 bytes of ...
   450  		{name: "MOVQstoreconst", argLength: 2, reg: gpstoreconst, asm: "MOVQ", aux: "SymValAndOff", typ: "Mem"}, // store 8 bytes of ...
   451  
   452  		{name: "MOVBstoreconstidx1", argLength: 3, reg: gpstoreconstidx, asm: "MOVB", aux: "SymValAndOff", typ: "Mem"}, // store low byte of ValAndOff(AuxInt).Val() to arg0+1*arg1+ValAndOff(AuxInt).Off()+aux.  arg2=mem
   453  		{name: "MOVWstoreconstidx1", argLength: 3, reg: gpstoreconstidx, asm: "MOVW", aux: "SymValAndOff", typ: "Mem"}, // store low 2 bytes of ... arg1 ...
   454  		{name: "MOVWstoreconstidx2", argLength: 3, reg: gpstoreconstidx, asm: "MOVW", aux: "SymValAndOff", typ: "Mem"}, // store low 2 bytes of ... 2*arg1 ...
   455  		{name: "MOVLstoreconstidx1", argLength: 3, reg: gpstoreconstidx, asm: "MOVL", aux: "SymValAndOff", typ: "Mem"}, // store low 4 bytes of ... arg1 ...
   456  		{name: "MOVLstoreconstidx4", argLength: 3, reg: gpstoreconstidx, asm: "MOVL", aux: "SymValAndOff", typ: "Mem"}, // store low 4 bytes of ... 4*arg1 ...
   457  		{name: "MOVQstoreconstidx1", argLength: 3, reg: gpstoreconstidx, asm: "MOVQ", aux: "SymValAndOff", typ: "Mem"}, // store 8 bytes of ... arg1 ...
   458  		{name: "MOVQstoreconstidx8", argLength: 3, reg: gpstoreconstidx, asm: "MOVQ", aux: "SymValAndOff", typ: "Mem"}, // store 8 bytes of ... 8*arg1 ...
   459  
   460  		// arg0 = (duff-adjusted) pointer to start of memory to zero
   461  		// arg1 = value to store (will always be zero)
   462  		// arg2 = mem
   463  		// auxint = offset into duffzero code to start executing
   464  		// returns mem
   465  		{
   466  			name:      "DUFFZERO",
   467  			aux:       "Int64",
   468  			argLength: 3,
   469  			reg: regInfo{
   470  				inputs:   []regMask{buildReg("DI"), buildReg("X0")},
   471  				clobbers: buildReg("DI FLAGS"),
   472  			},
   473  		},
   474  		{name: "MOVOconst", reg: regInfo{nil, 0, []regMask{fp}}, typ: "Int128", rematerializeable: true},
   475  
   476  		// arg0 = address of memory to zero
   477  		// arg1 = # of 8-byte words to zero
   478  		// arg2 = value to store (will always be zero)
   479  		// arg3 = mem
   480  		// returns mem
   481  		{
   482  			name:      "REPSTOSQ",
   483  			argLength: 4,
   484  			reg: regInfo{
   485  				inputs:   []regMask{buildReg("DI"), buildReg("CX"), buildReg("AX")},
   486  				clobbers: buildReg("DI CX"),
   487  			},
   488  		},
   489  
   490  		{name: "CALLstatic", argLength: 1, reg: regInfo{clobbers: callerSave}, aux: "SymOff"},                                // call static function aux.(*gc.Sym).  arg0=mem, auxint=argsize, returns mem
   491  		{name: "CALLclosure", argLength: 3, reg: regInfo{[]regMask{gpsp, buildReg("DX"), 0}, callerSave, nil}, aux: "Int64"}, // call function via closure.  arg0=codeptr, arg1=closure, arg2=mem, auxint=argsize, returns mem
   492  		{name: "CALLdefer", argLength: 1, reg: regInfo{clobbers: callerSave}, aux: "Int64"},                                  // call deferproc.  arg0=mem, auxint=argsize, returns mem
   493  		{name: "CALLgo", argLength: 1, reg: regInfo{clobbers: callerSave}, aux: "Int64"},                                     // call newproc.  arg0=mem, auxint=argsize, returns mem
   494  		{name: "CALLinter", argLength: 2, reg: regInfo{inputs: []regMask{gp}, clobbers: callerSave}, aux: "Int64"},           // call fn by pointer.  arg0=codeptr, arg1=mem, auxint=argsize, returns mem
   495  
   496  		// arg0 = destination pointer
   497  		// arg1 = source pointer
   498  		// arg2 = mem
   499  		// auxint = offset from duffcopy symbol to call
   500  		// returns memory
   501  		{
   502  			name:      "DUFFCOPY",
   503  			aux:       "Int64",
   504  			argLength: 3,
   505  			reg: regInfo{
   506  				inputs:   []regMask{buildReg("DI"), buildReg("SI")},
   507  				clobbers: buildReg("DI SI X0 FLAGS"), // uses X0 as a temporary
   508  			},
   509  		},
   510  
   511  		// arg0 = destination pointer
   512  		// arg1 = source pointer
   513  		// arg2 = # of 8-byte words to copy
   514  		// arg3 = mem
   515  		// returns memory
   516  		{
   517  			name:      "REPMOVSQ",
   518  			argLength: 4,
   519  			reg: regInfo{
   520  				inputs:   []regMask{buildReg("DI"), buildReg("SI"), buildReg("CX")},
   521  				clobbers: buildReg("DI SI CX"),
   522  			},
   523  		},
   524  
   525  		// (InvertFlags (CMPQ a b)) == (CMPQ b a)
   526  		// So if we want (SETL (CMPQ a b)) but we can't do that because a is a constant,
   527  		// then we do (SETL (InvertFlags (CMPQ b a))) instead.
   528  		// Rewrites will convert this to (SETG (CMPQ b a)).
   529  		// InvertFlags is a pseudo-op which can't appear in assembly output.
   530  		{name: "InvertFlags", argLength: 1}, // reverse direction of arg0
   531  
   532  		// Pseudo-ops
   533  		{name: "LoweredGetG", argLength: 1, reg: gp01}, // arg0=mem
   534  		// Scheduler ensures LoweredGetClosurePtr occurs only in entry block,
   535  		// and sorts it to the very beginning of the block to prevent other
   536  		// use of DX (the closure pointer)
   537  		{name: "LoweredGetClosurePtr", reg: regInfo{outputs: []regMask{buildReg("DX")}}},
   538  		//arg0=ptr,arg1=mem, returns void.  Faults if ptr is nil.
   539  		{name: "LoweredNilCheck", argLength: 2, reg: regInfo{inputs: []regMask{gpsp}, clobbers: flags}},
   540  
   541  		// MOVQconvert converts between pointers and integers.
   542  		// We have a special op for this so as to not confuse GC
   543  		// (particularly stack maps).  It takes a memory arg so it
   544  		// gets correctly ordered with respect to GC safepoints.
   545  		// arg0=ptr/int arg1=mem, output=int/ptr
   546  		{name: "MOVQconvert", argLength: 2, reg: gp11nf, asm: "MOVQ"},
   547  
   548  		// Constant flag values. For any comparison, there are 5 possible
   549  		// outcomes: the three from the signed total order (<,==,>) and the
   550  		// three from the unsigned total order. The == cases overlap.
   551  		// Note: there's a sixth "unordered" outcome for floating-point
   552  		// comparisons, but we don't use such a beast yet.
   553  		// These ops are for temporary use by rewrite rules. They
   554  		// cannot appear in the generated assembly.
   555  		{name: "FlagEQ"},     // equal
   556  		{name: "FlagLT_ULT"}, // signed < and unsigned <
   557  		{name: "FlagLT_UGT"}, // signed < and unsigned >
   558  		{name: "FlagGT_UGT"}, // signed > and unsigned <
   559  		{name: "FlagGT_ULT"}, // signed > and unsigned >
   560  	}
   561  
   562  	var AMD64blocks = []blockData{
   563  		{name: "EQ"},
   564  		{name: "NE"},
   565  		{name: "LT"},
   566  		{name: "LE"},
   567  		{name: "GT"},
   568  		{name: "GE"},
   569  		{name: "ULT"},
   570  		{name: "ULE"},
   571  		{name: "UGT"},
   572  		{name: "UGE"},
   573  		{name: "EQF"},
   574  		{name: "NEF"},
   575  		{name: "ORD"}, // FP, ordered comparison (parity zero)
   576  		{name: "NAN"}, // FP, unordered comparison (parity one)
   577  	}
   578  
   579  	archs = append(archs, arch{
   580  		name:     "AMD64",
   581  		pkg:      "cmd/internal/obj/x86",
   582  		genfile:  "../../amd64/ssa.go",
   583  		ops:      AMD64ops,
   584  		blocks:   AMD64blocks,
   585  		regnames: regNamesAMD64,
   586  	})
   587  }