// It is restored with ADDQ Rx, Rx: if Rx was -1 the carry is set, otherwise it is cleared.
// This is faster than using rotate instructions.

// func addVV(z, x, y []Word) (c Word)
TEXT ·addVV(SB),NOSPLIT,$0
	MOVQ z_len+8(FP), DI
	MOVQ x+24(FP), R8
	MOVQ y+48(FP), R9
	MOVQ z+0(FP), R10

	MOVQ $0, CX		// c = 0
	MOVQ $0, SI		// i = 0

// arithmetic operations on vectors implemented in arith.go.

// DI = R3, CX = R4, SI = r10, r8 = r8, r9=r9, r10 = r2, r11 = r5, r12 = r6, r13 = r7, r14 = r1 (R0 set to 0) + use R11
// func addVV(z, x, y []Word) (c Word)

TEXT ·addVV(SB), NOSPLIT, $0
	MOVD addvectorfacility+0x00(SB), R1
	BR   (R1)

TEXT ·addVV_check(SB), NOSPLIT, $0
	MOVB   ·hasVX(SB), R1

		return z.add(y, x)
	case m == 0:
		// n == 0 because m >= n; result is 0
		return z[:0]
	case n == 0:
		// result is x
		return z.set(x)
	}
	// m > 0

	z = z.make(m + 1)
	c := addVV(z[0:n], x, y)
	if m > n {
		c = addVW(z[n:m], x[n:], c)
	}
	z[m] = c

	return z.norm()
}

func (z nat) sub(x, y nat) nat {
	m := len(x)
	n := len(y)

	switch {
	case m < n:

// arithmetic operations on vectors implemented in arith.go.

// TODO: Consider re-implementing using Advanced SIMD
// once the assembler supports those instructions.

// func addVV(z, x, y []Word) (c Word)
TEXT ·addVV(SB),NOSPLIT,$0
	MOVD	z_len+8(FP), R0
	MOVD	x+24(FP), R8
	MOVD	y+48(FP), R9
	MOVD	z+0(FP), R10
	ADDS	$0, R0		// clear carry flag
	TBZ	$0, R0, two
	MOVD.P	8(R8), R11
	MOVD.P	8(R9), R15

#include "textflag.h"

// This file provides fast assembly versions for the elementary
// arithmetic operations on vectors implemented in arith.go.

// func addVV(z, y, y []Word) (c Word)
// z[i] = x[i] + y[i] for all i, carrying
TEXT ·addVV(SB), NOSPLIT, $0
	MOVD  z_len+8(FP), R7   // R7 = z_len
	MOVD  x+24(FP), R8      // R8 = x[]
	MOVD  y+48(FP), R9      // R9 = y[]
	MOVD  z+0(FP), R10      // R10 = z[]

		// If it underflows, q̂·v > u, which we fix up
		// by decrementing q̂ and adding v back.
		c := subVV(u[j:j+qhl], u[j:], qhatv)
		if c != 0 {
			c := addVV(u[j:j+n], u[j:], v)
			// If n == qhl, the carry from subVV and the carry from addVV
			// cancel out and don't affect u[j+n].
			if n < qhl {
				u[j+n] += c
			}
			qhat--
		}

		// Save quotient digit.