MOVL	SI, DI
	CLD; REPN; SCASB
	JZ 3(PC)
	MOVL	$-1, ret+16(FP)
	RET
	SUBL	SI, DI
	SUBL	$1, DI
	MOVL	DI, ret+16(FP)
	RET

TEXT ·IndexByteString(SB),NOSPLIT,$0-16
	MOVL	s_base+0(FP), SI
	MOVL	s_len+4(FP), CX
	MOVB	c+8(FP), AL
	MOVL	SI, DI
	CLD; REPN; SCASB
	JZ 3(PC)
	MOVL	$-1, ret+12(FP)
	RET
	SUBL	SI, DI
	SUBL	$1, DI
	MOVL	DI, ret+12(FP)

TEXT runtime·pthread_attr_init_trampoline(SB),NOSPLIT,$0
	PUSHL	BP
	MOVL	SP, BP
	SUBL	$4, SP
	MOVL	12(SP), DX		// pointer to args
	MOVL	0(DX), AX
	MOVL	AX, 0(SP)		// arg 1 - attr
	CALL	libc_pthread_attr_init(SB)
	MOVL	BP, SP
	POPL	BP
	RET

TEXT runtime·pthread_attr_destroy_trampoline(SB),NOSPLIT,$0
	PUSHL	BP
	MOVL	SP, BP
	SUBL	$4, SP
	MOVL	12(SP), DX		// pointer to args
	MOVL	0(DX), AX

	CLD

	// determine index into runtime·cbs table
	SUBL	$runtime·callbackasm(SB), AX
	MOVL	$0, DX
	MOVL	$5, BX	// divide by 5 because each call instruction in runtime·callbacks is 5 bytes long
	DIVL	BX
	SUBL	$1, AX	// subtract 1 because return PC is to the next slot

	// Create a struct callbackArgs on our stack.
	SUBL	$(12+callbackArgs__size), SP
	MOVL	AX, (12+callbackArgs_index)(SP)		// callback index

 * adjusted addresses
 */

	ADDL	BX, DI
	ADDL	BX, SI
	STD

/*
 * copy
 */
	MOVL	BX, CX
	SHRL	$2, CX
	ANDL	$3, BX

	SUBL	$4, DI
	SUBL	$4, SI
	REP;	MOVSL

	CLD
	ADDL	$4, DI
	ADDL	$4, SI
	SUBL	BX, DI
	SUBL	BX, SI
	JMP	tail

move_1or2:
	MOVB	(SI), AX
	MOVB	-1(SI)(BX*1), CX
	MOVB	AX, (DI)
	MOVB	CX, -1(DI)(BX*1)
	RET
move_0:
	RET
move_3:

 * adjusted addresses
 */

	ADDL	BX, DI
	ADDL	BX, SI
	STD

/*
 * copy
 */
	MOVL	BX, CX
	SHRL	$2, CX
	ANDL	$3, BX

	SUBL	$4, DI
	SUBL	$4, SI
	REP;	MOVSL

	CLD
	ADDL	$4, DI
	ADDL	$4, SI
	SUBL	BX, DI
	SUBL	BX, SI
	JMP	tail

move_1or2:
	MOVB	(SI), AX
	MOVB	-1(SI)(BX*1), CX
	MOVB	AX, (DI)
	MOVB	CX, -1(DI)(BX*1)
	RET
move_0:
	RET
move_3:

	LEAL	ret+16(FP), AX
	JMP	cmpbody<>(SB)

// input:
//   SI = a
//   DI = b
//   BX = alen
//   DX = blen
//   AX = address of return word (set to 1/0/-1)
TEXT cmpbody<>(SB),NOSPLIT,$0-0
	MOVL	DX, BP
	SUBL	BX, DX // DX = blen-alen
	JLE	2(PC)
	MOVL	BX, BP // BP = min(alen, blen)
	CMPL	SI, DI
	JEQ	allsame
	CMPL	BP, $4
	JB	small
#ifdef GO386_softfloat
	JMP	mediumloop
#endif
largeloop:
	CMPL	BP, $16

	MOVL $0, BX		// i = 0
	JMP E4

L4:	MOVL (SI)(BX*4), DX
	SUBL AX, DX
	MOVL DX, (DI)(BX*4)
	SBBL AX, AX		// save CF
	NEGL AX
	ADDL $1, BX		// i++

E4:	CMPL BX, BP		// i < n
	JL L4

	MOVL AX, c+28(FP)
	RET


// func shlVU(z, x []Word, s uint) (c Word)
TEXT ·shlVU(SB),NOSPLIT,$0
	MOVL z_len+4(FP), BX	// i = z
	SUBL $1, BX		// i--
	JL X8b			// i < 0	(n <= 0)

	// n > 0

	PAND	X2, X0
	PAND	X6, X4
	PAND	X4, X0
	PMOVMSKB X0, DX
	ADDL	$64, SI
	ADDL	$64, DI
	SUBL	$64, BX
	CMPL	DX, $0xffff
	JEQ	hugeloop
	MOVB	$0, (AX)
	RET

	// 4 bytes at a time using 32-bit register
bigloop:
	CMPL	BX, $4
	JBE	leftover
	MOVL	(SI), CX
	MOVL	(DI), DX
	ADDL	$4, SI
	ADDL	$4, DI
	SUBL	$4, BX
	CMPL	CX, DX
	JEQ	bigloop
	MOVB	$0, (AX)
	RET

	// remaining 0-4 bytes
leftover:

	MOVL	(_SYSTEM_TIME+time_lo), AX
	MOVL	(_SYSTEM_TIME+time_hi2), DX
	CMPL	CX, DX
	JNE	wall

	// w = DX:AX
	// convert to Unix epoch (but still 100ns units)
	#define delta 116444736000000000
	SUBL	$(delta & 0xFFFFFFFF), AX
	SBBL $(delta >> 32), DX

	// nano/100 = DX:AX
	// split into two decimal halves by div 1e9.
	// (decimal point is two spots over from correct place,
	// but we avoid overflow in the high word.)

	ADDL	$(1024 * 8), SP

	// Ask signaller to setgid
	MOVL	$1, (BX)

	// Wait for setgid completion
loop:
	PAUSE
	MOVL	(BX), AX
	CMPL	AX, $0
	JNE	loop

	// Restore stack
	SUBL	$(1024 * 8), SP