TEXT polyHashADInternal<>(SB), NOSPLIT, $0
	// adp points to beginning of additional data
	// itr2 holds ad length
	XORQ acc0, acc0
	XORQ acc1, acc1
	XORQ acc2, acc2
	CMPQ itr2, $13
	JNE  hashADLoop

openFastTLSAD:
	// Special treatment for the TLS case of 13 bytes
	MOVQ (adp), acc0
	MOVQ 5(adp), acc1
	SHRQ $24, acc1
	MOVQ $1, acc2
	polyMul
	RET

hashADLoop:

		p.To.Type = obj.TYPE_REG
		p.To.Reg = regEntryTmp0
		if ctxt.Arch.Family == sys.I386 {
			p.As = ATESTL
		}

		// JNE checkargp (checkargp to be resolved later)
		jne := obj.Appendp(p, newprog)
		jne.As = AJNE
		jne.To.Type = obj.TYPE_BRANCH

		// end:
		//  NOP
		end := obj.Appendp(jne, newprog)
		end.As = obj.ANOP

		// Fast forward to end of function.
		var last *obj.Prog

	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
	MOVSD   $0.5, X0 // 0x3FE0000000000000
	ORPD    X0, X2 // X2= f1

 */
	CMPL	SI, DI
	JLS	back

/*
 * forward copy loop
 */
forward:
	// If REP MOVSB isn't fast, don't use it
	CMPB	internal∕cpu·X86+const_offsetX86HasERMS(SB), $1 // enhanced REP MOVSB/STOSB
	JNE	fwdBy4

	// Check alignment
	MOVL	SI, AX
	ORL	DI, AX
	TESTL	$3, AX
	JEQ	fwdBy4

	// Do 1 byte at a time
	MOVL	BX, CX
	REP;	MOVSB
	RET

fwdBy4:
	// Do 4 bytes at a time
	MOVL	BX, CX

	CALL	runtime·exits(SB)
	JMP	0(PC)

// void sigtramp(void *ureg, int8 *note)
TEXT runtime·sigtramp(SB),NOSPLIT,$0
	get_tls(AX)

	// check that g exists
	MOVL	g(AX), BX
	CMPL	BX, $0
	JNE	3(PC)
	CALL	runtime·badsignal2(SB) // will exit
	RET

	// save args
	MOVL	ureg+0(FP), CX
	MOVL	note+4(FP), DX

	// change stack
	MOVL	g_m(BX), BX
	MOVL	m_gsignal(BX), BP

	FMADD   F4, F2, F0
	FMOVD   F0, ret+8(FP)
	RET

L16:
	RISBGZ	$40, $55, $56, R3, R2
	RISBGZ	$57, $60, $51, R3, R3
	ORW     $0x45000000, R2
	BR      L4
L13:
	BGE     L18     //jnl .L18
	BVS     L18
	FMOVD   log10rodataL19<>+16(SB), F0
	BR      L1
L17:
	SRAW    $1, R2, R2
	SUBW    $0x40000000, R2
	BR      L8
L18:
	FMOVD   log10rodataL19<>+8(SB), F0

	LTDBR	F0, F0
	MOVD    $0x4086000000000000, R2
	MOVD    $0x4086000000000000, R3
	BLTU    L19
	FMOVD   F0, F4
L2:
	WORD    $0xED409018     //cdb %f4,.L24-.L23(%r9)
	BYTE    $0x00
	BYTE    $0x19
	BGE     L14     //jnl   .L14
	BVS     L14
	WFCEDBS V4, V4, V2
	BEQ     L20
L1:
	FMOVD   F0, ret+8(FP)
	RET

L14:
	WFCEDBS V4, V4, V2
	BVS     L1
	MOVD    $coshxlim1<>+0(SB), R1
	FMOVD   0(R1), F2

	JB loop33to63
fail_avx2:
	VZEROUPPER
fail:
	MOVQ $-1, (R11)
	RET
success_avx2:
	VZEROUPPER
	JMP success
sse42:
#ifndef hasSSE42
	CMPB internal∕cpu·X86+const_offsetX86HasSSE42(SB), $1
	JNE no_sse42
#endif
	CMPQ AX, $12
	// PCMPESTRI is slower than normal compare,
	// so using it makes sense only if we advance 4+ bytes per compare
	// This value was determined experimentally and is the ~same

TEXT runtime·settls(SB),NOSPLIT,$0
	RET

// void sigtramp(void *ureg, int8 *note)
TEXT runtime·sigtramp(SB),NOSPLIT|NOFRAME,$0
	get_tls(AX)

	// check that g exists
	MOVQ	g(AX), BX
	CMPQ	BX, $0
	JNE	3(PC)
	CALL	runtime·badsignal2(SB) // will exit
	RET

	// save args
	MOVQ	ureg+0(FP), CX
	MOVQ	note+8(FP), DX

	// change stack
	MOVQ	g_m(BX), BX
	MOVQ	m_gsignal(BX), R10

	MOVL	DX, 0(CX)(FS)
	RET

TEXT runtime·nanotime1(SB),NOSPLIT,$0-8
loop:
	MOVL	(_INTERRUPT_TIME+time_hi1), AX
	MOVL	(_INTERRUPT_TIME+time_lo), CX
	MOVL	(_INTERRUPT_TIME+time_hi2), DI
	CMPL	AX, DI
	JNE	loop

	// wintime = DI:CX, multiply by 100
	MOVL	$100, AX
	MULL	CX
	IMULL	$100, DI
	ADDL	DI, DX
	// wintime*100 = DX:AX
	MOVL	AX, ret_lo+0(FP)
	MOVL	DX, ret_hi+4(FP)
	RET