// DI:SI = v
	MOVL	val_lo+4(FP), SI
	MOVL	val_hi+8(FP), DI
	// DX:AX = *addr
	MOVL	0(BP), AX
	MOVL	4(BP), DX
casloop:
	// CX:BX = DX:AX (*addr) | DI:SI (mask)
	MOVL	AX, BX
	MOVL	DX, CX
	ORL	SI, BX
	ORL	DI, CX
	LOCK
	CMPXCHG8B	0(BP)
	JNZ casloop
	MOVL	AX, ret_lo+12(FP)
	MOVL	DX, ret_hi+16(FP)
	RET

// func Anduintptr(addr *uintptr, v uintptr) old uintptr
TEXT ·Anduintptr(SB), NOSPLIT, $0-12

		return 1
	}
	return 0
}

func biton32(a, b uint32) (n uint32) {
	// amd64:"BTSL"
	n += b | (1 << (a & 31))

	// amd64:"ORL\t[$]-2147483648"
	n += a | (1 << 31)

	// amd64:"ORL\t[$]268435456"
	n += a | (1 << 28)

	// amd64:"ORL\t[$]1"
	n += a | (1 << 0)

	return n
}

func bitoff32(a, b uint32) (n uint32) {
	// amd64:"BTRL"
	n += b &^ (1 << (a & 31))

	RORXL    $11, e, y1;                  \ // y1 = e >> 11				// S1B
	;                                     \
	ADDL     (disp + 0*4)(SP)(SRND*1), h; \ // h = k + w + h        // disp = k + w
	ORL      c, y3;                       \ // y3 = a|c				// MAJA
	VPALIGNR $4, XDWORD2, XDWORD3, XTMP0; \ // XTMP0 = W[-7]
	MOVL     f, y2;                       \ // y2 = f				// CH

	ROLL $shift, a; \
	MOVL c, BP; \
	ADDL b, a

#define ROUND2(a, b, c, d, index, const, shift) \
	LEAL	const(a)(DI*1),a; \
	MOVL	d,		DI; \
	ANDL	b,		DI; \
	MOVL	d,		BP; \
	NOTL	BP; \
	ANDL	c,		BP; \
	ORL	DI,		BP; \
	MOVL	(index*4)(SI),DI; \
	ADDL	BP,		a; \
	ROLL	$shift,	a; \
	ADDL	b,		a

#define ROUND3(a, b, c, d, index, const, shift) \
	LEAL	const(a)(DI*1),a; \
	MOVL	(index*4)(SI),DI; \
	XORL	d,		BP; \

	MOVQ	ptr+0(FP), AX
	MOVB	val+8(FP), BX
	LOCK
	ANDB	BX, (AX)
	RET

// func Or(addr *uint32, v uint32)
TEXT ·Or(SB), NOSPLIT, $0-12
	MOVQ	ptr+0(FP), AX
	MOVL	val+8(FP), BX
	LOCK
	ORL	BX, (AX)
	RET

// func And(addr *uint32, v uint32)
TEXT ·And(SB), NOSPLIT, $0-12
	MOVQ	ptr+0(FP), AX
	MOVL	val+8(FP), BX
	LOCK
	ANDL	BX, (AX)
	RET

	XORL	c, DI; \
	ANDL	b, DI; \
	XORL	d, DI

#define FUNC2(a, b, c, d, e) \
	MOVL	b, DI; \
	XORL	c, DI; \
	XORL	d, DI

#define FUNC3(a, b, c, d, e) \
	MOVL	b, SI; \
	ORL	c, SI; \
	ANDL	d, SI; \
	MOVL	b, DI; \
	ANDL	c, DI; \
	ORL	SI, DI

#define FUNC4 FUNC2

#define MIX(a, b, c, d, e, const) \
	ROLL	$30, b; \
	ADDL	DI, e; \
	MOVL	a, SI; \
	ROLL	$5, SI; \
	LEAL	const(e)(SI*1), e

	ROUND3(CX,DX,AX,BX, 2,0x1fa27cf8,16);
	ROUND3(BX,CX,DX,AX, 0,0xc4ac5665,23);

	MOVL	R11,	R9
	XORL	DX,		R9

#define ROUND4(a, b, c, d, index, const, shift) \
	ADDL	$const,	a; \
	ADDL	R8,		a; \
	ORL		b,		R9; \
	XORL	c,		R9; \
	ADDL	R9,		a; \
	MOVL	(index*4)(SI),R8; \
	MOVL	$0xffffffff,	R9; \
	ROLL	$shift,		a; \
	XORL	c,		R9; \
	ADDL	b,		a

	ROUND4(AX,BX,CX,DX, 7,0xf4292244, 6);

	XORL	c, R9; \
	ANDL	b, R9; \
	XORL	d, R9

#define FUNC2(a, b, c, d, e) \
	MOVL	b, R9; \
	XORL	c, R9; \
	XORL	d, R9

#define FUNC3(a, b, c, d, e) \
	MOVL	b, R8; \
	ORL	c, R8; \
	ANDL	d, R8; \
	MOVL	b, R9; \
	ANDL	c, R9; \
	ORL	R8, R9

#define FUNC4 FUNC2

#define MIX(a, b, c, d, e, const) \
	ROLL	$30, b; \
	ADDL	R9, e; \
	MOVL	a, R8; \
	ROLL	$5, R8; \
	LEAL	const(e)(R10*1), e; \
	ADDL	R8, e

	// s390x:"FLOGR","MOVWZ"
	// ppc64x/power8:"ANDN","POPCNTW"
	// ppc64x/power9: "CNTTZW"
	// wasm:"I64Ctz"
	return bits.TrailingZeros32(n)
}

func TrailingZeros16(n uint16) int {
	// amd64:"BSFL","ORL\\t\\$65536"
	// 386:"BSFL\t"
	// arm:"ORR\t\\$65536","CLZ",-"MOVHU\tR"
	// arm64:"ORR\t\\$65536","RBITW","CLZW",-"MOVHU\tR",-"RBIT\t",-"CLZ\t"
	// s390x:"FLOGR","OR\t\\$65536"
	// ppc64x/power8:"POPCNTD","ORIS\\t\\$1"

	// 386:`MOVWLZX\s\([A-Z]+\)\([A-Z]+`,-`ORL`,-`MOVB`
	// amd64:`MOVWLZX\s\([A-Z]+\)\([A-Z]+`,-`MOVB`,-`OR`
	// ppc64le:`MOVHZ`,-`MOVBZ`
	// ppc64:`MOVHBR`,-`MOVBZ`
	return uint16(s[idx]) | uint16(s[idx+1])<<8
}

func load_le_byte2_uint16_idx_inv(s []byte, idx int) uint16 {
	// arm64:`MOVHU\s\(R[0-9]+\)\(R[0-9]+\)`,-`ORR`,-`MOVB`
	// 386:`MOVWLZX\s\([A-Z]+\)\([A-Z]+`,-`ORL`,-`MOVB`