ADDQ	$(1024 * 8), SP

	// Ask signaller to setgid
	MOVL	$1, ·Baton(SB)

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

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

	JB	small
#ifdef GO386_softfloat
	JMP	mediumloop
#endif
largeloop:
	CMPL	BP, $16
	JB	mediumloop
	MOVOU	(SI), X0
	MOVOU	(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, BX
	XORL	$0xffff, BX	// convert EQ to NE
	JNE	diff16	// branch if at least one byte is not equal
	ADDL	$16, SI
	ADDL	$16, DI
	SUBL	$16, BP
	JMP	largeloop

diff16:
	BSFL	BX, BX	// index of first byte that differs
	XORL	DX, DX
	MOVB	(SI)(BX*1), CX

	// CX = size (want in BX)
	CMPQ	AX, BX
	JNE	neq
	MOVQ	$1, AX	// return 1
	RET
neq:
	MOVQ	AX, SI
	MOVQ	BX, DI
	MOVQ	CX, BX
	JMP	memeqbody<>(SB)

// memequal_varlen(a, b unsafe.Pointer) bool
TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT,$0-17
	// AX = a       (want in SI)
	// BX = b       (want in DI)
	// 8(DX) = size (want in BX)
	CMPQ	AX, BX
	JNE	neq
	MOVQ	$1, AX	// return 1
	RET
neq:

	JEQ	_0
	CMPL	BX, $2
	JBE	_1or2
	CMPL	BX, $4
	JB	_3
	JE	_4
	CMPL	BX, $8
	JBE	_5through8
	CMPL	BX, $16
	JBE	_9through16
	MOVL	BX, CX
	SHRL	$2, CX
	REP
	STOSL
	ANDL	$3, BX
	JNE	tail
	RET

_1or2:
	MOVB	AX, (DI)
	MOVB	AX, -1(DI)(BX*1)
	RET
_0:
	RET
_3:
	MOVW	AX, (DI)
	MOVB	AX, 2(DI)
	RET
_4:
	// We need a separate case for 4 to make sure we clear pointers atomically.

	// find out information about the processor we're on
	MOVL	$0, AX
	CPUID
	CMPL	AX, $0
	JE	nocpuinfo

	CMPL	BX, $0x756E6547  // "Genu"
	JNE	notintel
	CMPL	DX, $0x49656E69  // "ineI"
	JNE	notintel
	CMPL	CX, $0x6C65746E  // "ntel"
	JNE	notintel
	MOVB	$1, runtime·isIntel(SB)

notintel:
	// Load EAX=1 cpuid flags
	MOVL	$1, AX
	CPUID
	MOVL	AX, runtime·processorVersionInfo(SB)

has_cpuid:
	MOVL	$0, AX
	CPUID
	MOVL	AX, SI
	CMPL	AX, $0
	JE	nocpuinfo

	CMPL	BX, $0x756E6547  // "Genu"
	JNE	notintel
	CMPL	DX, $0x49656E69  // "ineI"
	JNE	notintel
	CMPL	CX, $0x6C65746E  // "ntel"
	JNE	notintel
	MOVB	$1, runtime·isIntel(SB)
notintel:

	// Load EAX=1 cpuid flags
	MOVL	$1, AX
	CPUID
	MOVL	CX, DI // Move to global variable clobbers CX when generating PIC

	MOVOU	X0, -64(DI)(BX*1)
	MOVOU	X0, -48(DI)(BX*1)
	MOVOU	X0, -32(DI)(BX*1)
	MOVOU	X0, -16(DI)(BX*1)
	RET
nosse2:
	MOVL	BX, CX
	SHRL	$2, CX
	REP
	STOSL
	ANDL	$3, BX
	JNE	tail

	MOVQ    BX, 24(SP)
	MOVQ    BP, 32(SP)
	MOVQ	R12, 40(SP)
	MOVQ	R13, 48(SP)
	MOVQ	R14, 56(SP)
	MOVQ	R15, 64(SP)

	get_tls(BX)
	// check that g exists
	MOVQ	g(BX), R10
	CMPQ	R10, $0
	JNE	allgood
	MOVQ	SI, 72(SP)
	MOVQ	DX, 80(SP)
	LEAQ	72(SP), AX
	MOVQ	DI, 0(SP)
	MOVQ	AX, 8(SP)
	MOVQ	$runtime·badsignal(SB), AX
	CALL	AX
	JMP	exit

allgood:

	FUCOMI  F0, F1       // compare F0 to F1
	JCC     2(PC)        // jump if F0 >= F1
	FXCHD   F0, F1       // F0=|p| (larger), F1=|q| (smaller)
	FTST                 // compare F0 to 0
	FSTSW	AX
	ANDW    $0x4000, AX
	JNE     10(PC)       // jump if F0 = 0
	FXCHD   F0, F1       // F0=q (smaller), F1=p (larger)
	FDIVD   F1, F0       // F0=q(=q/p), F1=p
	FMULD   F0, F0       // F0=q*q, F1=p
	FLD1                 // F0=1, F1=q*q, F2=p

	JBE	_17through32
	CMPQ	BX, $64
	JBE	_33through64
	CMPQ	BX, $128
	JBE	_65through128
	CMPQ	BX, $256
	JBE	_129through256

	CMPB	internal∕cpu·X86+const_offsetX86HasERMS(SB), $1 // enhanced REP MOVSB/STOSB
	JNE	skip_erms

	// If the size is less than 2kb, do not use ERMS as it has a big start-up cost.
	// Table 3-4. Relative Performance of Memcpy() Using ERMSB Vs. 128-bit AVX