MOVD	b+16(FP), R5	// R5 = b
	MOVD	n+24(FP), R6	// R6 = n

	CMPU	R6, $64, CR7	// Check if n ≥ 64 bytes
	MOVD	R0, R8		// R8 = index
	CMPU	R6, $8, CR6	// Check if 8 ≤ n < 64 bytes
	BLE	CR6, small	// <= 8
	BLT	CR7, xor32	// Case for 32 ≤ n < 64 bytes

	// Case for n ≥ 64 bytes
preloop64:
	SRD	$6, R6, R7	// Set up loop counter
	MOVD	R7, CTR
	MOVD	$16, R10
	MOVD	$32, R14
	MOVD	$48, R15

((Add32|Or32|Xor32) left:(Lsh32x64 x y) right:(Rsh32Ux64 x (Sub64 (Const64 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)
((Add32|Or32|Xor32) left:(Lsh32x32 x y) right:(Rsh32Ux32 x (Sub32 (Const32 [32]) y))) && (shiftIsBounded(left) || shiftIsBounded(right)) && canRotate(config, 32) => (RotateLeft32 x y)

					continue
				}
				y := v_1_1
				v.copyOf(y)
				return true
			}
		}
		break
	}
	// match: (Xor32 (Xor32 i:(Const32 <t>) z) x)
	// cond: (z.Op != OpConst32 && x.Op != OpConst32)
	// result: (Xor32 i (Xor32 <t> z x))
	for {
		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
			if v_0.Op != OpXor32 {
				continue
			}
			_ = v_0.Args[1]

	{name: "Or16", argLength: 2, commutative: true},
	{name: "Or32", argLength: 2, commutative: true},
	{name: "Or64", argLength: 2, commutative: true},

	{name: "Xor8", argLength: 2, commutative: true}, // arg0 ^ arg1
	{name: "Xor16", argLength: 2, commutative: true},
	{name: "Xor32", argLength: 2, commutative: true},
	{name: "Xor64", argLength: 2, commutative: true},

//
//go:linkname And32
//go:linkname Or32
//go:linkname And64
//go:linkname Or64
//go:linkname Anduintptr
//go:linkname Oruintptr

package atomic

import _ "unsafe" // For linkname

//go:nosplit
func And32(ptr *uint32, val uint32) uint32 {
	for {
		old := *ptr
		if Cas(ptr, old, old&val) {
			return old
		}
	}
}

//go:nosplit
func Or32(ptr *uint32, val uint32) uint32 {
	for {

	var x [128]uint32 // give x its own cache line
	sink = &x
	for i := 0; i < b.N; i++ {
		atomic.Or32(&x[63], uint32(i))
	}
}

func BenchmarkOr32Parallel(b *testing.B) {
	var x [128]uint32 // give x its own cache line
	sink = &x
	b.RunParallel(func(pb *testing.PB) {
		i := uint32(0)
		for pb.Next() {
			atomic.Or32(&x[63], i)
			i++
		}
	})
}

func BenchmarkOr64(b *testing.B) {

TEXT ·And(SB), NOSPLIT, $0-8
	MOVW	ptr+0(FP), R1
	MOVW	val+4(FP), R2

	SYNC
	LL	(R1), R3
	AND	R2, R3
	SC	R3, (R1)
	BEQ	R3, -4(PC)
	SYNC
	RET

// func Or32(addr *uint32, v uint32) old uint32
TEXT ·Or32(SB), NOSPLIT, $0-12
	MOVW	ptr+0(FP), R1
	MOVW	val+4(FP), R2

	SYNC
	LL	(R1), R3
	OR	R2, R3, R4
	SC	R4, (R1)
	BEQ	R4, -4(PC)
	SYNC
	MOVW	R3, ret+8(FP)
	RET

//go:noescape
func And(ptr *uint32, val uint32)

//go:noescape
func Or(ptr *uint32, val uint32)

//go:noescape
func And32(ptr *uint32, val uint32) uint32

//go:noescape
func Or32(ptr *uint32, val uint32) uint32

//go:noescape
func And64(ptr *uint64, val uint64) uint64

//go:noescape
func Or64(ptr *uint64, val uint64) uint64

//go:noescape

	MOVL	val+4(FP), CX
casloop:
	MOVL 	CX, DX
	MOVL	(BX), AX
	ANDL	AX, DX
	LOCK
	CMPXCHGL	DX, (BX)
	JNZ casloop
	MOVL 	AX, ret+8(FP)
	RET

// func Or32(addr *uint32, v uint32) old uint32
TEXT ·Or32(SB), NOSPLIT, $0-12
	MOVL	ptr+0(FP), BX
	MOVL	val+4(FP), CX
casloop:
	MOVL 	CX, DX
	MOVL	(BX), AX
	ORL	AX, DX
	LOCK
	CMPXCHGL	DX, (BX)
	JNZ casloop
	MOVL 	AX, ret+8(FP)

	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

// func Or32(addr *uint32, v uint32) old uint32
TEXT ·Or32(SB), NOSPLIT, $0-20
	MOVQ	ptr+0(FP), BX
	MOVL	val+8(FP), CX
casloop:
	MOVL 	CX, DX
	MOVL	(BX), AX
	ORL	AX, DX
	LOCK
	CMPXCHGL	DX, (BX)
	JNZ casloop
	MOVL 	AX, ret+16(FP)