MOVWBR	0(R6), R8
	MOVWZ	R5, R9

#define ROUND1(a, b, c, d, index, const, shift) \
	XOR	c, R9; \
	ADD	$const, a; \
	ADD	R8, a; \
	MOVWBR	(index*4)(R6), R8; \
	AND	b, R9; \
	XOR	d, R9; \
	ADD	R9, a; \
	RLL	$shift, a; \
	MOVWZ	c, R9; \
	ADD	b, a

	ROUND1(R2,R3,R4,R5, 1,0xd76aa478, 7);
	ROUND1(R5,R2,R3,R4, 2,0xe8c7b756,12);
	ROUND1(R4,R5,R2,R3, 3,0x242070db,17);
	ROUND1(R3,R4,R5,R2, 4,0xc1bdceee,22);

	a += x<<7 | x>>25

	// amd64:`ROLL\t[$]8`
	// arm:"MOVW\tR\\d+@>24"
	// arm64:"RORW\t[$]24"
	// s390x:"RLL\t[$]8"
	// ppc64x:"ROTLW\t[$]8"
	// loong64: "ROTR\t[$]24"
	// riscv64: "RORIW\t[$]24"
	a += x<<8 + x>>24

	// amd64:"ROLL\t[$]9"
	// arm:"MOVW\tR\\d+@>23"
	// arm64:"RORW\t[$]23"
	// s390x:"RLL\t[$]9"
	// ppc64x:"ROTLW\t[$]9"
	// loong64: "ROTR\t[$]23"
	// riscv64: "RORIW\t[$]23"
	a += x<<9 ^ x>>23

	// wasm:"I64Rotl"
	return bits.RotateLeft64(n, 37)
}

func RotateLeft32(n uint32) uint32 {
	// amd64:"ROLL" 386:"ROLL"
	// arm:`MOVW\tR[0-9]+@>23`
	// arm64:"RORW"
	// ppc64x:"ROTLW"
	// s390x:"RLL"
	// wasm:"I32Rotl"
	return bits.RotateLeft32(n, 9)
}

func RotateLeft16(n uint16, s int) uint16 {
	// amd64:"ROLW" 386:"ROLW"
	// arm64:"RORW",-"CSEL"
	return bits.RotateLeft16(n, s)
}

	ORW	$~0xff, R4           // R4 = uint32(v) | 0xffffff00
	MOVD	$(3<<3), R5
	RXSBG	$59, $60, $3, R3, R5 // R5 = 24 - ((addr % 4) * 8) = ((addr & 3) << 3) ^ (3 << 3)
	ANDW	$~3, R3              // R3 = floor(addr, 4) = addr &^ 3
	RLL	R5, R4, R4           // R4 = rotl(R4, R5)
	LAN	R4, R6, 0(R3)        // R6 = *R3; *R3 &= R4; (atomic)
	RET

// func Or(addr *uint32, v uint32)
TEXT ·Or(SB), NOSPLIT, $0-12
	MOVD	ptr+0(FP), R3
	MOVW	val+8(FP), R4

				{0, 23551}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R11 R12 R14
			},
			outputs: []outputInfo{
				{0, 23551}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R11 R12 R14
			},
		},
	},
	{
		name:   "RLL",
		argLen: 2,
		asm:    s390x.ARLL,
		reg: regInfo{
			inputs: []inputInfo{
				{1, 23550}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R11 R12 R14
				{0, 23551}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R11 R12 R14