MOVQ	CX, 0(SP)
	MOVQ	DX, 8(SP)

	LEAQ	sec+0(FP), DX
	MOVQ	-8(DX), CX	// Sets CX to function return address.
	MOVQ	CX, m_vdsoPC(BX)
	MOVQ	DX, m_vdsoSP(BX)

	CMPQ	R14, m_curg(BX)	// Only switch if on curg.
	JNE	noswitch

	MOVQ	m_g0(BX), DX
	MOVQ	(g_sched+gobuf_sp)(DX), SP	// Set SP to g0 stack

noswitch:
	SUBQ	$32, SP		// Space for two time results
	ANDQ	$~15, SP	// Align for C code

	MOVL	$0, DI // CLOCK_REALTIME

	BSFL	DX, DX	// Find first set bit.
	JZ	failure	// No set bit, failure.
	MOVQ	DX, (R8)
	RET

avx2:
#ifndef hasAVX2
	CMPB   internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JNE sse
#endif
	MOVD AX, X0
	LEAQ -32(SI)(BX*1), R11
	VPBROADCASTB  X0, Y1

	PCALIGN $32
avx2_loop:
	VMOVDQU (DI), Y2
	VPCMPEQB Y1, Y2, Y3
	VPTEST Y3, Y3
	JNZ avx2success
	ADDQ $32, DI

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

 */
	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

	// Directly return DX, we don't need to accumulate
	// since we have <16 bytes.
	POPCNTL DX, DX
	MOVQ	DX, (R8)
	RET

avx2:
#ifndef hasAVX2
	CMPB   internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JNE sse
#endif
	MOVD AX, X0
	LEAQ -64(SI)(BX*1), R11
	LEAQ (SI)(BX*1), R13
	VPBROADCASTB  X0, Y1
	PCALIGN $32
avx2_loop:
	VMOVDQU (DI), Y2
	VMOVDQU 32(DI), Y4
	VPCMPEQB Y1, Y2, Y3

	JGE     overflow
lastStep:
	SHLQ    $52, BX
	MOVQ    BX, X1
	MULSD   X1, X0
	MOVSD   X0, ret+8(FP)
	RET
notFinite:
	// test bits for -Inf
	MOVQ    $NegInf, AX
	CMPQ    AX, BX
	JNE     notNegInf
	// -Inf, return 0
underflow: // return 0
	MOVQ    $0, ret+8(FP)
	RET
overflow: // return +Inf
	MOVQ    $PosInf, BX
notNegInf: // NaN or +Inf, return x
	MOVQ    BX, ret+8(FP)
	RET

	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

	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