// Convert to nSec
	MOVW    $100, R2
	MULLU   R0, R2, (R4, R3)    // R4:R3 = R1:R0 * R2
	MULA    R1, R2, R4, R4
	// w = R2:R1 in nSec
	MOVW    R3, R1	      // R4:R3 -> R2:R1
	MOVW    R4, R2

	// multiply nanoseconds by reciprocal of 10**9 (scaled by 2**61)
	// to get seconds (96 bit scaled result)
	MOVW	$0x89705f41, R3		// 2**61 * 10**-9
	MULLU	R1,R3,(R6,R5)		// R7:R6:R5 = R2:R1 * R3
	MOVW	$0,R7

	"NEGD",
	"SRL",
	"SRA",
	"SLL",
	"MULU",
	"DIVU",
	"MUL",
	"MMUL",
	"DIV",
	"MOD",
	"MODU",
	"DIVHW",
	"DIVUHW",
	"MOVB",
	"MOVBS",
	"MOVBU",
	"MOVH",
	"MOVHS",
	"MOVHU",
	"MOVW",
	"MOVM",
	"SWPBU",
	"SWPW",
	"RFE",
	"SWI",
	"MULA",
	"MULS",
	"MMULA",
	"MMULS",
	"WORD",
	"MULL",
	"MULAL",
	"MULLU",
	"MULALU",
	"BX",
	"BXRET",

	JMP		addMulVVWx(SB)

TEXT addMulVVWx(SB), NOFRAME|NOSPLIT, $0
	MOVW	$0, R0
	MOVW	z+0(FP), R1
	MOVW	x+4(FP), R2
	MOVW	y+8(FP), R3
	ADD	R5<<2, R1, R5
	MOVW	$0, R4
	B E9

L9:	MOVW.P	4(R2), R6
	MULLU	R6, R3, (R7, R6)
	ADD.S	R4, R6
	ADC	R0, R7
	MOVW	0(R1), R4
	ADD.S	R4, R6
	ADC	R0, R7
	MOVW.P	R6, 4(R1)
	MOVW	R7, R4

E9:	TEQ	R1, R5
	BNE	L9

	MOVW	R4, c+12(FP)

	MOVW	$0x89705f41, R3		// 2**61 * 10**-9
	MULLU	R1,R3,(R6,R5)		// R5:R6:R7 = R1:R2 * R3
	MOVW	$0,R7
	MULALU	R2,R3,(R7,R6)

	// unscale by discarding low 32 bits, shifting the rest by 29
	MOVW	R6>>29,R6		// R6:R7 = (R5:R6:R7 >> 61)
	ORR	R7<<3,R6
	MOVW	R7>>29,R7

	// subtract (10**9 * sec) from nsec to get nanosecond remainder
	MOVW	$1000000000, R5		// 10**9
	MULLU	R6,R5,(R9,R8)		// R8:R9 = R6:R7 * R5
	MULA	R7,R5,R9,R9

	MOVW	$0, R0
	MOVW	z+0(FP), R1
	MOVW	z_len+4(FP), R5
	MOVW	x+12(FP), R2
	MOVW	y+24(FP), R3
	MOVW	r+28(FP), R4
	ADD	R5<<2, R1, R5
	B E8

	// word loop
L8:
	MOVW.P	4(R2), R6
	MULLU	R6, R3, (R7, R6)
	ADD.S	R4, R6
	ADC	R0, R7
	MOVW.P	R6, 4(R1)
	MOVW	R7, R4
E8:
	TEQ	R1, R5
	BNE	L8

	MOVW	R4, c+32(FP)
	RET


// func addMulVVW(z, x []Word, y Word) (c Word)

	TEQ 	RM->1, RM // check for d=0 or d=1

	// 2nd Newton iteration
	MUL.NE	RM, Rq, Ra
	MOVW.NE	$0, Rs
	MULAL.NE Rq, Ra, (Rq,Rs)
	BEQ 	udiv_by_0_or_1

	// q now accurate enough for a remainder r, 0<=r<3*d
	MULLU	Rq, Rr, (Rq,Rs) // q = (r * q) >> 32
	ADD 	RM, Rr, Rr // r = n - d
	MULA	RM, Rq, Rr, Rr // r = n - (q+1)*d

	// since 0 <= n-q*d < 3*d; thus -d <= r < 2*d
	CMN 	RM, Rr // t = r-d

	case fm == 0: // 0 * x = 0 (with sign adjusted)
		return f ^ gs

	case gm == 0: // x * 0 = 0 (with sign adjusted)
		return g ^ fs
	}

	// 53-bit * 53-bit = 107- or 108-bit
	lo, hi := mullu(fm, gm)
	shift := mantbits64 - 1
	trunc := lo & (1<<shift - 1)
	mant := hi<<(64-shift) | lo>>shift
	return fpack64(fs^gs, mant, fe+ge-1, trunc)
}

func fdiv64(f, g uint64) uint64 {

		{name: "RSCconst", argLength: 2, reg: gp1flags1, asm: "RSC", aux: "Int32"},  // auxInt - arg0 - carry, arg1=flags

		{name: "MULLU", argLength: 2, reg: gp22, asm: "MULLU", commutative: true}, // arg0 * arg1, high 32 bits in out0, low 32 bits in out1
		{name: "MULA", argLength: 3, reg: gp31, asm: "MULA"},                      // arg0 * arg1 + arg2

loop:
	MOVW	time_hi1(R3), R1
	DMB	MB_ISH
	MOVW	time_lo(R3), R0
	DMB	MB_ISH
	MOVW	time_hi2(R3), R2
	CMP	R1, R2
	BNE	loop

	// wintime = R1:R0, multiply by 100
	MOVW	$100, R2
	MULLU	R0, R2, (R4, R3)    // R4:R3 = R1:R0 * R2
	MULA	R1, R2, R4, R4

	// wintime*100 = R4:R3
	MOVW	R3, ret_lo+0(FP)
	MOVW	R4, ret_hi+4(FP)
	RET

// save_g saves the g register (R10) into thread local memory

	MOVW $4, R0 // CLOCK_MONOTONIC
	MOVW $8(R13), R1
	MOVW $SYS_clock_gettime, R7
	SWI $0

	MOVW 8(R13), R0 // sec.low
	MOVW 12(R13), R4 // sec.high
	MOVW 16(R13), R2 // nsec

	MOVW $1000000000, R3
	MULLU R0, R3, (R1, R0)
	MUL R3, R4
	ADD.S R2, R0
	ADC R4, R1

	MOVW R0, ret_lo+0(FP)
	MOVW R1, ret_hi+4(FP)
	RET

TEXT runtime·asmSigaction(SB),NOSPLIT|NOFRAME,$0
	MOVW sig+0(FP), R0		// arg 1 sig