CMPQ    AX, DX
	JLE     notFinite
	// check if argument will overflow
	MOVQ    BX, X0
	MOVSD   $Overflow, X1
	COMISD  X1, X0
	JA      overflow
	MOVSD   $LOG2E, X1
	MULSD   X0, X1
	CVTSD2SL X1, BX // BX = exponent
	CVTSL2SD BX, X1
	CMPB ·useFMA(SB), $1
	JE   avxfma
	MOVSD   $LN2U, X2
	MULSD   X1, X2
	SUBSD   X2, X0
	MOVSD   $LN2L, X2
	MULSD   X1, X2
	SUBSD   X2, X0
	// reduce argument
	MULSD   $0.0625, X0

	MOVSS	x_real+0(FP), X0
	MOVSS	x_imag+4(FP), X0
	MOVSD	y_real+8(FP), X0
	MOVSD	y_imag+16(FP), X0
	// Loading both parts of a complex is ok: see issue 35264.
	MOVSD	x+0(FP), X0
	MOVO	y+8(FP), X0
	MOVOU	y+8(FP), X0

	// These are not ok.
	MOVO	x+0(FP), X0 // ERROR "\[amd64\] cpx: invalid MOVO of x\+0\(FP\); complex64 is 8-byte value containing x_real\+0\(FP\) and x_imag\+4\(FP\)"

	MOVSD   $1.0, X3 // x0= 0 or ^0, x1= k, x2 = f1, x3= 1
	ANDPD   X0, X3 // x0= 0 or ^0, x1= k, x2 = f1, x3= 0 or 1
	SUBSD   X3, X1 // x0= 0 or ^0, x1= k, x2 = f1, x3= 0 or 1
	MOVSD   $1.0, X0 // x0= 1, x1= k, x2= f1, x3= 0 or 1
	ADDSD   X0, X3 // x0= 1, x1= k, x2= f1, x3= 1 or 2
	MULSD   X3, X2 // x0= 1, x1= k, x2= f1
	// f := f1 - 1
	SUBSD   X0, X2 // x1= k, x2= f
	// s := f / (2 + f)
	MOVSD   $2.0, X0
	ADDSD   X2, X0
	MOVAPD  X2, X3

	JAE     isBig_floor
	MOVQ	AX, X0 // X0 = x
	CVTTSD2SQ	X0, AX
	CVTSQ2SD	AX, X1 // X1 = float(int(x))
	CMPSD	X1, X0, 1 // compare LT; X0 = 0xffffffffffffffff or 0
	MOVSD	$(-1.0), X2
	ANDPD	X2, X0 // if x < float(int(x)) {X0 = -1} else {X0 = 0}
	ADDSD	X1, X0
	MOVSD	X0, ret+8(FP)
	RET
isBig_floor:
	MOVQ    AX, ret+8(FP) // return x
	RET

// func archCeil(x float64) float64
TEXT ·archCeil(SB),NOSPLIT,$0
	MOVQ	x+0(FP), AX

	MOVQ    R9, CX
	ANDQ    DX, CX // y = |y|
	CMPQ    AX, CX
	JLT     isMaxNaN
	// ±0 special cases
	ORQ     CX, BX
	JEQ     isMaxZero

	MOVQ    R8, X0
	MOVQ    R9, X1
	MAXSD   X1, X0
	MOVSD   X0, ret+16(FP)
	RET
isMaxNaN: // return NaN
	MOVQ	$NaN, AX
isPosInf: // return +Inf
	MOVQ    AX, ret+16(FP)
	RET
isMaxZero:
	MOVQ    $(1<<63), AX // -0.0
	CMPQ    AX, R8
	JEQ     +3(PC)

}

func idxFloat64(x, y []float64, i int) {
	var t float64
	//    amd64: `MOVSD\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), X[0-9]+`
	// 386/sse2: `MOVSD\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), X[0-9]+`
	//    arm64: `FMOVD\t\(R[0-9]*\)\(R[0-9]*<<3\), F[0-9]+`
	t = x[i+1]
	//    amd64: `MOVSD\tX[0-9]+, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`
	// 386/sse2: `MOVSD\tX[0-9]+, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`

	MOVQ    BX, X0
	MOVQ    CX, X1
	ORQ     CX, BX
	JEQ     isZero
	MOVAPD  X0, X2
	MAXSD   X1, X0
	MINSD   X2, X1
	DIVSD   X0, X1
	MULSD   X1, X1
	ADDSD   $1.0, X1
	SQRTSD  X1, X1
	MULSD   X1, X0
	MOVSD   X0, ret+16(FP)
	RET
isInfOrNaN:
	CMPQ    AX, BX
	JEQ     isInf
	CMPQ    AX, CX
	JEQ     isInf
	MOVQ    $NaN, AX
	MOVQ    AX, ret+16(FP) // return NaN
	RET
isInf:
	MOVQ    AX, ret+16(FP) // return +Inf

		}
		switch inst.Op {
		case INVLPG:
			prefix = "byte "
		case STOSB, MOVSB, CMPSB, LODSB, SCASB:
			prefix = "byte "
		case STOSW, MOVSW, CMPSW, LODSW, SCASW:
			prefix = "word "
		case STOSD, MOVSD, CMPSD, LODSD, SCASD:
			prefix = "dword "
		case STOSQ, MOVSQ, CMPSQ, LODSQ, SCASQ:
			prefix = "qword "
		case LAR:
			prefix = "word "
		case BOUND:
			if inst.Mode == 32 {
				prefix = "qword "

		{name: "MOVSDload", argLength: 2, reg: fpload, asm: "MOVSD", aux: "SymOff", faultOnNilArg0: true, symEffect: "Read"}, // fp64 load
		{name: "MOVSSconst", reg: fp01, asm: "MOVSS", aux: "Float32", rematerializeable: true},                               // fp32 constant
		{name: "MOVSDconst", reg: fp01, asm: "MOVSD", aux: "Float64", rematerializeable: true},                               // fp64 constant

	CQO:       "cqto",
	INSD:      "insl",
	IRET:      "iretw",
	IRETD:     "iret",
	IRETQ:     "iretq",
	LODSB:     "lods",
	LODSD:     "lods",
	LODSQ:     "lods",
	LODSW:     "lods",
	MOVSD:     "movsl",
	MOVSD_XMM: "movsd",
	OUTSD:     "outsl",
	POPA:      "popaw",
	POPAD:     "popa",
	POPF:      "popfw",
	POPFD:     "popf",
	PUSHA:     "pushaw",
	PUSHAD:    "pusha",
	PUSHF:     "pushfw",