MULHDU	R5, R14, R11	// high x[i]*y
	ADDC	R15, R10
	ADDZE	R11
	ADDC	R9, R10
	ADDZE	R11, R9
	MULLD	R5, R16, R14	// low x[i+1]*y
	MULHDU	R5, R16, R15	// high x[i+1]*y
	ADDC	R17, R14
	ADDZE	R15
	ADDC	R9, R14
	ADDZE	R15, R9
	MULLD	R5, R18, R16	// low x[i+2]*y
	MULHDU	R5, R18, R17	// high x[i+2]*y
	ADDC	R19, R16
	ADDZE	R17
	ADDC	R9, R16
	ADDZE	R17, R9
	MULLD	R5, R20, R18	// low x[i+3]*y

	ADDC   h0, t1, t1;  \
	MULLD  h2, r1, t3;  \
	ADDZE  t4, h0;      \
	MULHDU r1, h1, t5;  \
	MULLD  r1, h1, t4;  \
	ADDC   t4, t2, t2;  \
	ADDE   t5, t3, t3;  \
	ADDC   h0, t2, t2;  \
	MOVD   $-4, t4;     \
	ADDZE  t3;          \
	RLDICL $0, t2, $62, h2; \
	AND    t2, t4, h0;  \
	ADDC   t0, h0, h0;  \
	ADDE   t3, t1, h1;  \
	SLD    $62, t3, t4; \
	SRD    $2, t2;      \
	ADDZE  h2;          \

		{name: "AND", argLength: 2, reg: gp21, asm: "AND", commutative: true},                           // arg0&arg1
		{name: "ANDN", argLength: 2, reg: gp21, asm: "ANDN"},                                            // arg0&^arg1
		{name: "ANDNCC", argLength: 2, reg: gp21, asm: "ANDNCC", typ: "(Int64,Flags)"},                  // arg0&^arg1 sets CC

package lifecycle

import (
	"encoding/xml"
)

var errDuplicateTagKey = Errorf("Duplicate Tag Keys are not allowed")

// And - a tag to combine a prefix and multiple tags for lifecycle configuration rule.
type And struct {
	XMLName               xml.Name `xml:"And"`
	ObjectSizeGreaterThan int64    `xml:"ObjectSizeGreaterThan,omitempty"`
	ObjectSizeLessThan    int64    `xml:"ObjectSizeLessThan,omitempty"`

	RORL	$2, DI; \
	MOVL	a, DX; \
	ANDL	b, BX; \
	RORL	$13, DX; \
	MOVL	a, CX; \
	ANDL	c, CX; \
	XORL	DX, DI; \
	XORL	CX, BX; \
	MOVL	a, DX; \
	MOVL	b, CX; \
	RORL	$22, DX; \
	ANDL	a, CX; \
	XORL	CX, BX; \
	XORL	DX, DI; \
	ADDL	DI, BX

// Calculate T1 and T2, then e = d + T1 and a = T1 + T2.
// The values for e and a are stored in d and h, ready for rotation.

(MOVWZreg ((OR|XOR|AND) <t> x (MOVWZreg y))) => (MOVWZreg ((OR|XOR|AND) <t> x y))
(MOVHZreg ((OR|XOR|AND) <t> x (MOVWZreg y))) => (MOVHZreg ((OR|XOR|AND) <t> x y))
(MOVHZreg ((OR|XOR|AND) <t> x (MOVHZreg y))) => (MOVHZreg ((OR|XOR|AND) <t> x y))
(MOVBZreg ((OR|XOR|AND) <t> x (MOVWZreg y))) => (MOVBZreg ((OR|XOR|AND) <t> x y))
(MOVBZreg ((OR|XOR|AND) <t> x (MOVHZreg y))) => (MOVBZreg ((OR|XOR|AND) <t> x y))

	RORL	$2, AX; \
	MOVL	(a*4)(DI), DX; \
	ANDL	(b*4)(DI), BX; \
	RORL	$13, DX; \
	MOVL	(a*4)(DI), CX; \
	ANDL	(c*4)(DI), CX; \
	XORL	DX, AX; \
	XORL	CX, BX; \
	MOVL	(a*4)(DI), DX; \
	MOVL	(b*4)(DI), CX; \
	RORL	$22, DX; \
	ANDL	(a*4)(DI), CX; \
	XORL	CX, BX; \
	XORL	DX, AX; \
	ADDL	AX, BX

// Calculate T1 and T2, then e = d + T1 and a = T1 + T2.

	LEAL	const(a)(DI*1), a; \
	ANDL	b, BP; \
	XORL d, BP; \
	MOVL (index*4)(SI), DI; \
	ADDL BP, a; \
	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

// Uses https://github.com/animetosho/md5-optimisation#dependency-shortcut-in-g-function

#define ROUND2(a, b, c, d, index, const, shift) \
	XORL	R11, R9; \
	ADDL	$const,	a; \
	ADDL	R8,	a; \
	ANDL	b,		R10; \
	ANDL	c,		R9; \
	MOVL	(index*4)(SI),R8; \
	ADDL	R9,	a; \
	ADDL	R10,	a; \
	MOVL	c,		R9; \
	MOVL	c,		R10; \
	ROLL	$shift,	a; \
	ADDL	b,		a

	ROUND2(AX,BX,CX,DX, 6,0xf61e2562, 5);

func BenchmarkAnd8(b *testing.B) {
	var x [512]uint8 // give byte its own cache line
	sink = &x
	for i := 0; i < b.N; i++ {
		atomic.And8(&x[255], uint8(i))
	}
}

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

func BenchmarkAnd8Parallel(b *testing.B) {
	var x [512]uint8 // give byte its own cache line