JEQ     isPosInf
	// NaN special cases
	MOVQ    $~(1<<63), DX // bit mask
	MOVQ    $PosInf, AX
	MOVQ    R8, BX
	ANDQ    DX, BX // x = |x|
	CMPQ    AX, BX
	JLT     isMaxNaN
	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

	RET

// func archCeil(x float64) float64
TEXT ·archCeil(SB),NOSPLIT,$0
	MOVQ	x+0(FP), AX
	MOVQ	$~(1<<63), DX // sign bit mask
	MOVQ	AX, BX // BX = copy of x
	ANDQ    DX, BX // BX = |x|
	MOVQ    $Big, CX // if |x| >= 2**52 or IsNaN(x), return x
	CMPQ    BX, CX
	JAE     isBig_ceil
	MOVQ	AX, X0 // X0 = x
	MOVQ	DX, X2 // X2 = sign bit mask
	CVTTSD2SQ	X0, AX
	ANDNPD	X0, X2 // X2 = sign

multiply:
	POLY1305_MUL(R8, R9, R10, R11, R12, BX, CX, R13, R14)
	SUBQ $16, R15
	CMPQ R15, $16
	JAE  loop

bytes_between_0_and_15:
	TESTQ R15, R15
	JZ    done
	MOVQ  $1, BX
	XORQ  CX, CX
	XORQ  R13, R13
	ADDQ  R15, SI

flush_buffer:
	SHLQ $8, BX, CX
	SHLQ $8, BX
	MOVB -1(SI), R13
	XORQ R13, BX
	DECQ SI
	DECQ R15
	JNZ  flush_buffer

	ADDQ BX, R8
	ADCQ CX, R9
	ADCQ $0, R10

	// Create a new thread to do the runtime initialization and return.
	MOVQ	BX, 32(SP) // callee-saved, preserved across the CALL
	MOVQ	SP, BX
	ANDQ	$~15, SP // alignment as per Windows requirement
	MOVQ	_cgo_sys_thread_create(SB), AX
	MOVQ	$_rt0_amd64_windows_lib_go(SB), CX
	MOVQ	$0, DX
	CALL	AX
	MOVQ	BX, SP
	MOVQ	32(SP), BX
	RET

TEXT _rt0_amd64_windows_lib_go(SB),NOSPLIT|NOFRAME,$0
	MOVQ  $0, DI
	MOVQ	$0, SI

TEXT ·archLog(SB),NOSPLIT,$0
	// test bits for special cases
	MOVQ    x+0(FP), BX
	MOVQ    $~(1<<63), AX // sign bit mask
	ANDQ    BX, AX
	JEQ     isZero
	MOVQ    $0, AX
	CMPQ    AX, BX
	JGT     isNegative
	MOVQ    $PosInf, AX
	CMPQ    AX, BX
	JLE     isInfOrNaN
	// f1, ki := math.Frexp(x); k := float64(ki)
	MOVQ    BX, X0
	MOVQ    $0x000FFFFFFFFFFFFF, AX
	MOVQ    AX, X2
	ANDPD   X0, X2

	DIVB	foo+4(SB)
	PUSHL	$foo+4(SB)
	POPL		AX

// LTYPE3 rimrem	{ outcode(int($1), &$2); }
	SUBB	$1, AX
	SUBB	$1, foo+4(SB)
	SUBB	BX, AX
	SUBB	BX, foo+4(SB)

// LTYPE4 remrim	{ outcode(int($1), &$2); }
	CMPB	AX, $1
	CMPB	foo+4(SB), $4
	CMPB	BX, AX
	CMPB	foo+4(SB), BX

// LTYPER nonrel	{ outcode(int($1), &$2); }
label:
	JC	label // JCS
	JC	-1(PC) // JCS -1(PC)

// LTYPEC spec3	{ outcode(int($1), &$2); }

// Use a local trampoline, to avoid taking the address of a dynamically exported
// function.
TEXT ·set_crosscall2(SB),NOSPLIT,$0-0
	MOVL	_crosscall2_ptr(SB), AX
	MOVL	$crosscall2_trampoline<>(SB), BX
	MOVL	BX, (AX)
	RET

TEXT crosscall2_trampoline<>(SB),NOSPLIT,$0-0
	JMP	crosscall2(SB)

// Called by C code generated by cmd/cgo.
// func crosscall2(fn, a unsafe.Pointer, n int32, ctxt uintptr)

	return a << big
}

func main() {
	f(1)
}

/*
main·f: doasm: notfound from=75 to=13 (82)    SHLQ    $-9223372036854775808,BX
main·f: doasm: notfound from=75 to=13 (82)    SHLQ    $-9223372036854775808,BX
main·f: doasm: notfound from=75 to=13 (82)    SHLQ    $-9223372036854775808,BX

// func memclrNoHeapPointers(ptr unsafe.Pointer, n uintptr)
TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-16
	MOVQ	ptr+0(FP), DI
	MOVQ	n+8(FP), CX
	MOVQ	CX, BX
	ANDQ	$7, BX
	SHRQ	$3, CX
	MOVQ	$0, AX
	CLD
	REP
	STOSQ
	MOVQ	BX, CX
	REP
	STOSB

	VMOVNTDQ Y26, 15(DX)(BX*4)                         // 62617d28e7949a0f000000
	VMOVNTDQ Z18, -15(R14)(R15*1)                      // 62817d48e7943ef1ffffff
	VMOVNTDQ Z24, -15(R14)(R15*1)                      // 62017d48e7843ef1ffffff
	VMOVNTDQ Z18, -15(BX)                              // 62e17d48e793f1ffffff
	VMOVNTDQ Z24, -15(BX)                              // 62617d48e783f1ffffff