FMOVDP  F0, ret+16(FP)
	RET
not_finite:
// test bits for -Inf or +Inf
	MOVL    p_hi+4(FP), AX  // high word p
	ORL     p_lo+0(FP), AX  // low word p
	ANDL    $0x7fffffff, AX
	CMPL    AX, $0x7ff00000
	JEQ     is_inf
	MOVL    q_hi+12(FP), AX  // high word q
	ORL     q_lo+8(FP), AX   // low word q
	ANDL    $0x7fffffff, AX
	CMPL    AX, $0x7ff00000
	JEQ     is_inf

	// 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

	s *= x[i+3]
	// 386: `ANDL\t16\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	// amd64: `ANDL\t16\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`
	s &= x[i+4]
	// 386: `ORL\t20\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	// amd64: `ORL\t20\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`
	s |= x[i+5]
	// 386: `XORL\t24\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	// amd64: `XORL\t24\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`

	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

		{name: "ANDLconst", argLength: 1, reg: gp11, asm: "ANDL", aux: "Int32", resultInArg0: true, clobberFlags: true}, // arg0 & auxint

		{name: "ORL", argLength: 2, reg: gp21, asm: "ORL", commutative: true, resultInArg0: true, clobberFlags: true}, // arg0 | arg1
		{name: "ORLconst", argLength: 1, reg: gp11, asm: "ORL", aux: "Int32", resultInArg0: true, clobberFlags: true}, // arg0 | auxint

	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

		{name: "ORLloadidx1", argLength: 4, reg: gp21loadidx, asm: "ORL", scale: 1, aux: "SymOff", resultInArg0: true, clobberFlags: true, symEffect: "Read"},
		{name: "ORLloadidx4", argLength: 4, reg: gp21loadidx, asm: "ORL", scale: 4, aux: "SymOff", resultInArg0: true, clobberFlags: true, symEffect: "Read"},
		{name: "ORLloadidx8", argLength: 4, reg: gp21loadidx, asm: "ORL", scale: 8, aux: "SymOff", resultInArg0: true, clobberFlags: true, symEffect: "Read"},