Пример #1
0
void
mpn_toom2_sqr (mp_ptr pp,
	       mp_srcptr ap, mp_size_t an,
	       mp_ptr scratch)
{
  const int __gmpn_cpuvec_initialized = 1;
  mp_size_t n, s;
  mp_limb_t cy, cy2;
  mp_ptr asm1;

#define a0  ap
#define a1  (ap + n)

  s = an >> 1;
  n = an - s;

  ASSERT (0 < s && s <= n && s >= n - 1);

  asm1 = pp;

  /* Compute asm1.  */
  if (s == n)
    {
      if (mpn_cmp (a0, a1, n) < 0)
	{
	  mpn_sub_n (asm1, a1, a0, n);
	}
      else
	{
	  mpn_sub_n (asm1, a0, a1, n);
	}
    }
  else /* n - s == 1 */
    {
      if (a0[s] == 0 && mpn_cmp (a0, a1, s) < 0)
	{
	  mpn_sub_n (asm1, a1, a0, s);
	  asm1[s] = 0;
	}
      else
	{
	  asm1[s] = a0[s] - mpn_sub_n (asm1, a0, a1, s);
	}
    }

#define v0	pp				/* 2n */
#define vinf	(pp + 2 * n)			/* s+s */
#define vm1	scratch				/* 2n */
#define scratch_out	scratch + 2 * n

  /* vm1, 2n limbs */
  TOOM2_SQR_REC (vm1, asm1, n, scratch_out);

  /* vinf, s+s limbs */
  TOOM2_SQR_REC (vinf, a1, s, scratch_out);

  /* v0, 2n limbs */
  TOOM2_SQR_REC (v0, ap, n, scratch_out);

  /* H(v0) + L(vinf) */
  cy = mpn_add_n (pp + 2 * n, v0 + n, vinf, n);

  /* L(v0) + H(v0) */
  cy2 = cy + mpn_add_n (pp + n, pp + 2 * n, v0, n);

  /* L(vinf) + H(vinf) */
  cy += mpn_add (pp + 2 * n, pp + 2 * n, n, vinf + n, s + s - n);

  cy -= mpn_sub_n (pp + n, pp + n, vm1, 2 * n);

  ASSERT (cy + 1  <= 3);
  ASSERT (cy2 <= 2);

  MPN_INCR_U (pp + 2 * n, s + s, cy2);
  if (LIKELY (cy <= 2))
    MPN_INCR_U (pp + 3 * n, s + s - n, cy);
  else
    MPN_DECR_U (pp + 3 * n, s + s - n, 1);
}
Пример #2
0
void
mpn_toom2_sqr (mp_ptr pp,
	       mp_srcptr ap, mp_size_t an,
	       mp_ptr scratch)
{
  mp_size_t n, s;
  mp_limb_t cy, cy2;
  mp_ptr asm1;

#define a0  ap
#define a1  (ap + n)

  s = an >> 1;
  n = an - s;

  ASSERT (0 < s && s <= n);

  asm1 = pp;

  /* Compute asm1.  */
  if (s == n)
    {
      if (mpn_cmp (a0, a1, n) < 0)
	{
	  mpn_sub_n (asm1, a1, a0, n);
	}
      else
	{
	  mpn_sub_n (asm1, a0, a1, n);
	}
    }
  else
    {
      if (mpn_zero_p (a0 + s, n - s) && mpn_cmp (a0, a1, s) < 0)
	{
	  mpn_sub_n (asm1, a1, a0, s);
	  MPN_ZERO (asm1 + s, n - s);
	}
      else
	{
	  mpn_sub (asm1, a0, n, a1, s);
	}
    }

#define v0	pp				/* 2n */
#define vinf	(pp + 2 * n)			/* s+s */
#define vm1	scratch				/* 2n */
#define scratch_out	scratch + 2 * n

  /* vm1, 2n limbs */
  TOOM2_SQR_REC (vm1, asm1, n, scratch_out);

  /* vinf, s+s limbs */
  TOOM2_SQR_REC (vinf, a1, s, scratch_out);

  /* v0, 2n limbs */
  TOOM2_SQR_REC (v0, ap, n, scratch_out);

  /* H(v0) + L(vinf) */
  cy = mpn_add_n (pp + 2 * n, v0 + n, vinf, n);

  /* L(v0) + H(v0) */
  cy2 = cy + mpn_add_n (pp + n, pp + 2 * n, v0, n);

  /* L(vinf) + H(vinf) */
  cy += mpn_add (pp + 2 * n, pp + 2 * n, n, vinf + n, s + s - n);

  cy -= mpn_sub_n (pp + n, pp + n, vm1, 2 * n);

  ASSERT (cy + 1  <= 3);
  ASSERT (cy2 <= 2);

  mpn_incr_u (pp + 2 * n, cy2);
  if (LIKELY (cy <= 2))
    mpn_incr_u (pp + 3 * n, cy);
  else
    mpn_decr_u (pp + 3 * n, 1);
}