/*
 * forward copy loop
 */
forward:
	// If REP MOVSB isn't fast, don't use it
	CMPB	internal∕cpu·X86+const_offsetX86HasERMS(SB), $1 // enhanced REP MOVSB/STOSB
	JNE	fwdBy4

	// Check alignment
	MOVL	SI, AX
	ORL	DI, AX
	TESTL	$3, AX
	JEQ	fwdBy4

	// Do 1 byte at a time
	MOVL	BX, CX
	REP;	MOVSB
	RET

fwdBy4:
	// Do 4 bytes at a time
	MOVL	BX, CX
	SHRL	$2, CX
	ANDL	$3, BX
	REP;	MOVSL
	JMP	tail

	MOVQ	R10, g(BX)

	// call noted(AX)
	MOVQ	AX, 0(SP)
	CALL	runtime·noted(SB)
	RET

TEXT runtime·setfpmasks(SB),NOSPLIT,$8
	STMXCSR	0(SP)
	MOVL	0(SP), AX
	ANDL	$~0x3F, AX
	ORL	$(0x3F<<7), AX
	MOVL	AX, 0(SP)
	LDMXCSR	0(SP)
	RET

#define ERRMAX 128	/* from os_plan9.h */

// void errstr(int8 *buf, int32 len)
TEXT errstr<>(SB),NOSPLIT,$0
	MOVQ    $41, BP
	SYSCALL
	RET

	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`

	"MULSD",
	"MULSS",
	"MULW",
	"MULXL",
	"MULXQ",
	"MWAIT",
	"NEGB",
	"NEGL",
	"NEGQ",
	"NEGW",
	"NOPL",
	"NOPW",
	"NOTB",
	"NOTL",
	"NOTQ",
	"NOTW",
	"ORB",
	"ORL",
	"ORPD",
	"ORPS",
	"ORQ",
	"ORW",
	"OUTB",
	"OUTL",
	"OUTSB",
	"OUTSL",
	"OUTSW",
	"OUTW",
	"PABSB",
	"PABSD",
	"PABSW",
	"PACKSSLW",
	"PACKSSWB",
	"PACKUSDW",

	ORL DX, (BX)                            // 0913
	ORL R11, (BX)                           // 44091b
	ORL DX, (R11)                           // 410913
	ORL R11, (R11)                          // 45091b
	ORL DX, DX                              // 09d2 or 0bd2
	ORL R11, DX                             // 4409da or 410bd3
	ORL DX, R11                             // 4109d3 or 440bda

	JLS	move_256through2048

	// If REP MOVSB isn't fast, don't use it
	CMPB	internal∕cpu·X86+const_offsetX86HasERMS(SB), $1 // enhanced REP MOVSB/STOSB
	JNE	fwdBy8

	// Check alignment
	MOVL	SI, AX
	ORL	DI, AX
	TESTL	$7, AX
	JEQ	fwdBy8

	// Do 1 byte at a time
	MOVQ	BX, CX
	REP;	MOVSB
	RET

fwdBy8:
	// Do 8 bytes at a time
	MOVQ	BX, CX
	SHRQ	$3, CX
	ANDQ	$7, BX
	REP;	MOVSQ

			}
			v.reset(Op386ANDLmodify)
			v.AuxInt = int32ToAuxInt(off)
			v.Aux = symToAux(sym)
			v.AddArg3(ptr, x, mem)
			return true
		}
		break
	}
	// match: (MOVLstore {sym} [off] ptr y:(ORL l:(MOVLload [off] {sym} ptr mem) x) mem)
	// cond: y.Uses==1 && l.Uses==1 && clobber(y, l)
	// result: (ORLmodify [off] {sym} ptr x mem)
	for {
		off := auxIntToInt32(v.AuxInt)
		sym := auxToSym(v.Aux)