// 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

	MOVOU	(SI), X0
	MOVOU	(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff16

	MOVOU	16(SI), X0
	MOVOU	16(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff32

	MOVOU	32(SI), X0
	MOVOU	32(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff48

	MOVOU	48(SI), X0
	MOVOU	48(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX

func ptrEqual() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[:], x[:])
}

func ptrOneOffset() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[1:], x[:])
}

func ptrBothOffset() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[1:], x[2:])
}

	// will clobber the test pattern created by the caller
	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

aligned:
	MOVQ $0, AX // position in slices

	PCALIGN $16
loop16b:
	MOVOU (SI)(AX*1), X0   // XOR 16byte forwards.
	MOVOU (CX)(AX*1), X1
	PXOR  X1, X0
	MOVOU X0, (BX)(AX*1)
	ADDQ  $16, AX
	CMPQ  DX, AX
	JNE   loop16b
	RET

	PCALIGN $16
loop_1b:
	SUBQ  $1, DX           // XOR 1byte backwards.
	MOVB  (SI)(DX*1), DI
	MOVB  (CX)(DX*1), AX
	XORB  AX, DI
	MOVB  DI, (BX)(DX*1)

	MOVL	(_INTERRUPT_TIME+time_hi2), DI
	CMPL	AX, DI
	JNE	loop

	// w = DI:CX
	// multiply by 100
	MOVL	$100, AX
	MULL	CX
	IMULL	$100, DI
	ADDL	DI, DX
	// w*100 = DX:AX
	MOVL	AX, mono+12(FP)
	MOVL	DX, mono+16(FP)

wall:
	MOVL	(_SYSTEM_TIME+time_hi1), CX
	MOVL	(_SYSTEM_TIME+time_lo), AX
	MOVL	(_SYSTEM_TIME+time_hi2), DX
	CMPL	CX, DX
	JNE	wall

	// w = DX:AX

	CMPL	CX, $0;	JE	_0args
	CMPL	CX, $1;	JE	_1args
	CMPL	CX, $2;	JE	_2args
	CMPL	CX, $3;	JE	_3args
	CMPL	CX, $4;	JE	_4args

	// Check we have enough room for args.
	CMPL	CX, $const_maxArgs
	JLE	2(PC)
	INT	$3			// not enough room -> crash

	// Copy args to the stack.
	MOVQ	SP, DI
	CLD
	REP; MOVSQ
	MOVQ	SP, SI

	// Load first 4 args into correspondent registers.

	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

	MOVQ R12, 8(R10)(SI*8)
	MOVQ R13, 16(R10)(SI*8)
	MOVQ R14, 24(R10)(SI*8)
	SBBQ CX, CX		// save CF

	ADDQ $4, SI		// i += 4
	SUBQ $4, DI		// n -= 4
	JGE U1			// if n >= 0 goto U1

V1:	ADDQ $4, DI		// n += 4
	JLE E1			// if n <= 0 goto E1

L1:	// n > 0
	ADDQ CX, CX		// restore CF
	MOVQ 0(R8)(SI*8), R11
	ADCQ 0(R9)(SI*8), R11
	MOVQ R11, 0(R10)(SI*8)
	SBBQ CX, CX		// save CF

	ADDQ $1, SI		// i++
	SUBQ $1, DI		// n--

	// 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: