void
_nmod_poly_divrem_basecase_3(mp_ptr Q, mp_ptr R, mp_ptr W,
mp_srcptr A, slong lenA, mp_srcptr B, slong lenB,
nmod_t mod)
{
const mp_limb_t invL = n_invmod(B[lenB - 1], mod.n);
slong iR, i;
mp_ptr B3 = W, R3 = W + 3*(lenB - 1), ptrQ = Q - lenB + 1;
for (i = 0; i < lenB - 1; i++)
{
B3[3 * i] = B[i];
B3[3 * i + 1] = 0;
B3[3 * i + 2] = 0;
}
for (i = 0; i < lenA; i++)
{
R3[3 * i] = A[i];
R3[3 * i + 1] = 0;
R3[3 * i + 2] = 0;
}
for (iR = lenA - 1; iR >= lenB - 1; )
{
mp_limb_t r =
n_lll_mod_preinv(R3[3 * iR + 2], R3[3 * iR + 1],
R3[3 * iR], mod.n, mod.ninv);
while ((iR + 1 >= lenB) && (r == WORD(0)))
{
ptrQ[iR--] = WORD(0);
if (iR + 1 >= lenB)
r = n_lll_mod_preinv(R3[3 * iR + 2], R3[3 * iR + 1],
R3[3 * iR], mod.n, mod.ninv);
}
if (iR + 1 >= lenB)
{
ptrQ[iR] = n_mulmod2_preinv(r, invL, mod.n, mod.ninv);
if (lenB > 1)
{
const mp_limb_t c = n_negmod(ptrQ[iR], mod.n);
mpn_addmul_1(R3 + 3 * (iR - lenB + 1), B3, 3 * lenB - 3, c);
}
iR--;
}
}
for (iR = 0; iR < lenB - 1; iR++)
R[iR] = n_lll_mod_preinv(R3[3 * iR + 2], R3[3 * iR + 1],
R3[3 * iR], mod.n, mod.ninv);
}
void _nmod_poly_rem_basecase_3(mp_ptr R, mp_ptr W,
mp_srcptr A, long lenA, mp_srcptr B, long lenB,
nmod_t mod)
{
if (lenB > 1)
{
const mp_limb_t invL = n_invmod(B[lenB - 1], mod.n);
long iR, i;
mp_ptr B3 = W, R3 = W + 3*(lenB - 1);
for (i = 0; i < lenB - 1; i++)
{
B3[3 * i] = B[i];
B3[3 * i + 1] = 0;
B3[3 * i + 2] = 0;
}
for (i = 0; i < lenA; i++)
{
R3[3 * i] = A[i];
R3[3 * i + 1] = 0;
R3[3 * i + 2] = 0;
}
for (iR = lenA - 1; iR >= lenB - 1; iR--)
{
const mp_limb_t r = n_lll_mod_preinv(R3[3*iR + 2], R3[3*iR + 1],
R3[3*iR], mod.n, mod.ninv);
if (r != 0)
{
const mp_limb_t q = n_mulmod2_preinv(r, invL, mod.n, mod.ninv);
const mp_limb_t c = n_negmod(q, mod.n);
mpn_addmul_1(R3 + 3 * (iR - lenB + 1), B3, 3 * lenB - 3, c);
}
}
for (iR = 0; iR < lenB - 1; iR++)
R[iR] = n_lll_mod_preinv(R3[3 * iR + 2], R3[3 * iR + 1],
R3[3 * iR], mod.n, mod.ninv);
}
}
void sample(void * arg, ulong count)
{
mp_limb_t n, d, dinv, r = 0, norm;
double dpre;
info_t * info = (info_t *) arg;
mp_bitcnt_t bits = info->bits;
ulong type = info->type;
ulong i;
flint_rand_t state;
flint_randinit(state);
mp_ptr arr = (mp_ptr) malloc(1024*sizeof(mp_limb_t));
mp_ptr arr2 = (mp_ptr) malloc(1024*sizeof(mp_limb_t));
for (i = 0; i < count; i++)
{
int j;
d = n_randbits(state, bits);
if (d == 0UL) d++;
dinv = n_preinvert_limb(d);
for (j = 0; j < 1024; j++)
{
arr[j] = n_randbits(state, FLINT_BITS);
arr2[j] = n_randint(state, n);
}
switch (type)
{
case 1:
prof_start();
for (mp_size_t j = 0; j < 10000UL; j++)
{
r += n_lll_mod_preinv(arr2[j&1023], arr[j&1023], arr[(j+1)&1023], d, dinv);
}
prof_stop();
break;
}
}
if (r == 9879875897UL) abort();
flint_randclear(state);
free(arr);
free(arr2);
}
int main(void)
{
int i, result;
flint_rand_t state;
printf("lll_mod_preinv....");
fflush(stdout);
flint_randinit(state);
for (i = 0; i < 1000000; i++)
{
mp_limb_t d, dinv, nh, nm, nl, r1, r2, m;
d = n_randtest_not_zero(state);
m = n_randtest(state);
nh = n_randint(state, d);
nm = n_randtest(state);
nl = n_randtest(state);
dinv = n_preinvert_limb(d);
r2 = n_lll_mod_preinv(nh, nm, nl, d, dinv);
nm = n_ll_mod_preinv(nh, nm, d, dinv);
r1 = n_ll_mod_preinv(nm, nl, d, dinv);
result = (r1 == r2);
if (!result)
{
printf("FAIL:\n");
printf("nh = %lu, nm = %ld, nl = %lu, d = %lu, dinv = %lu\n", nh, nm, nl, d, dinv);
printf("r1 = %lu, r2 = %lu\n", r1, r2);
abort();
}
}
flint_randclear(state);
printf("PASS\n");
return 0;
}
void
_nmod_poly_div_basecase_3(mp_ptr Q, mp_ptr W,
mp_srcptr A, long A_len, mp_srcptr B, long B_len,
nmod_t mod)
{
long coeff, i, len;
mp_limb_t lead_inv = n_invmod(B[B_len - 1], mod.n);
mp_limb_t r_coeff;
mp_ptr B3, R3;
mp_srcptr Btop;
B3 = W;
for (i = 0; i < B_len - 1; i++)
{
B3[3 * i] = B[i];
B3[3 * i + 1] = 0;
B3[3 * i + 2] = 0;
}
Btop = B3 + 3*(B_len - 1);
R3 = W + 3*(B_len - 1);
for (i = 0; i < A_len - B_len + 1; i++)
{
R3[3 * i] = A[B_len + i - 1];
R3[3 * i + 1] = 0;
R3[3 * i + 2] = 0;
}
coeff = A_len - B_len;
while (coeff >= 0)
{
r_coeff =
n_lll_mod_preinv(R3[3 * coeff + 2], R3[3 * coeff + 1],
R3[3 * coeff], mod.n, mod.ninv);
while (coeff >= 0 && r_coeff == 0L)
{
Q[coeff--] = 0L;
if (coeff >= 0)
r_coeff =
n_lll_mod_preinv(R3[3 * coeff + 2], R3[3 * coeff + 1],
R3[3 * coeff], mod.n, mod.ninv);
}
if (coeff >= 0)
{
mp_limb_t c, * R_sub;
Q[coeff] =
n_mulmod2_preinv(r_coeff, lead_inv, mod.n, mod.ninv);
c = n_negmod(Q[coeff], mod.n);
len = FLINT_MIN(B_len - 1, coeff);
R_sub = R3 + 3 * (coeff - len);
if (len > 0)
mpn_addmul_1(R_sub, Btop - 3*len, 3 * len, c);
coeff--;
}
}
}