void
_nmod_poly_divrem_basecase_1(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;
mp_ptr ptrQ = Q - lenB + 1;
mp_ptr R1 = W;
flint_mpn_copyi(R1, A, lenA);
for (iR = lenA - 1; iR >= lenB - 1; iR--)
{
if (R1[iR] == 0)
{
ptrQ[iR] = WORD(0);
}
else
{
ptrQ[iR] = n_mulmod2_preinv(R1[iR], invL, mod.n, mod.ninv);
if (lenB > 1)
{
const mp_limb_t c = n_negmod(ptrQ[iR], mod.n);
mpn_addmul_1(R1 + iR - lenB + 1, B, lenB - 1, c);
}
}
}
if (lenB > 1)
_nmod_vec_reduce(R, R1, lenB - 1, mod);
}
void sample(void * arg, ulong count)
{
mp_limb_t n;
nmod_t mod;
info_t * info = (info_t *) arg;
mp_bitcnt_t bits = info->bits;
mp_ptr vec = _nmod_vec_init(1000);
mp_ptr vec2 = _nmod_vec_init(1000);
mp_size_t j;
long i;
flint_rand_t state;
flint_randinit(state);
for (j = 0; j < 1000; j++)
vec[j] = n_randlimb(state);
prof_start();
for (i = 0; i < count; i++)
{
n = n_randbits(state, bits);
if (n == 0UL) n++;
nmod_init(&mod, n);
_nmod_vec_reduce(vec2, vec, 1000, mod);
}
prof_stop();
flint_randclear(state);
_nmod_vec_clear(vec);
_nmod_vec_clear(vec2);
}
void _nmod_poly_rem_basecase_1(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;
mp_ptr R1 = W;
mpn_copyi(R1, A, lenA);
for (iR = lenA - 1; iR >= lenB - 1; iR--)
{
if (R1[iR] != 0)
{
const mp_limb_t q = n_mulmod2_preinv(R1[iR], invL, mod.n, mod.ninv);
const mp_limb_t c = n_negmod(q, mod.n);
mpn_addmul_1(R1 + iR - lenB + 1, B, lenB - 1, c);
}
}
_nmod_vec_reduce(R, R1, lenB - 1, mod);
}
}
void
nmod_mat_set(nmod_mat_t B, const nmod_mat_t A)
{
if (A->mod.n <= B->mod.n)
_nmod_vec_set(B->entries, A->entries, A->r*A->c);
else
_nmod_vec_reduce(B->entries, A->entries, A->r*A->c, B->mod);
}
/* Assumes poly1 and poly2 are not length 0 and 0 < trunc <= len1 + len2 - 1 */
void
_nmod_poly_mullow_classical(mp_ptr res, mp_srcptr poly1, slong len1,
mp_srcptr poly2, slong len2, slong trunc, nmod_t mod)
{
if (len1 == 1 || trunc == 1) /* Special case if the length of output is 1 */
{
res[0] = n_mulmod2_preinv(poly1[0], poly2[0], mod.n, mod.ninv);
}
else /* Ordinary case */
{
slong i;
slong bits = FLINT_BITS - (slong) mod.norm;
slong log_len = FLINT_BIT_COUNT(len2);
if (2 * bits + log_len <= FLINT_BITS)
{
/* Set res[i] = poly1[i]*poly2[0] */
mpn_mul_1(res, poly1, FLINT_MIN(len1, trunc), poly2[0]);
if (len2 != 1)
{
/* Set res[i+len1-1] = in1[len1-1]*in2[i] */
if (trunc > len1)
mpn_mul_1(res + len1, poly2 + 1, trunc - len1,
poly1[len1 - 1]);
/* out[i+j] += in1[i]*in2[j] */
for (i = 0; i < FLINT_MIN(len1, trunc) - 1; i++)
mpn_addmul_1(res + i + 1, poly2 + 1,
FLINT_MIN(len2, trunc - i) - 1, poly1[i]);
}
_nmod_vec_reduce(res, res, trunc, mod);
}
else
{
/* Set res[i] = poly1[i]*poly2[0] */
_nmod_vec_scalar_mul_nmod(res, poly1, FLINT_MIN(len1, trunc),
poly2[0], mod);
if (len2 == 1)
return;
/* Set res[i+len1-1] = in1[len1-1]*in2[i] */
if (trunc > len1)
_nmod_vec_scalar_mul_nmod(res + len1, poly2 + 1, trunc - len1,
poly1[len1 - 1], mod);
/* out[i+j] += in1[i]*in2[j] */
for (i = 0; i < FLINT_MIN(len1, trunc) - 1; i++)
_nmod_vec_scalar_addmul_nmod(res + i + 1, poly2 + 1,
FLINT_MIN(len2, trunc - i) - 1,
poly1[i], mod);
}
}
}
int
main(void)
{
int i, result;
flint_rand_t state;
flint_randinit(state);
printf("reduce....");
fflush(stdout);
for (i = 0; i < 10000; i++)
{
long j, len = n_randint(state, 100) + 1;
mp_ptr vec = _nmod_vec_init(len);
mp_ptr vec2 = _nmod_vec_init(len);
mp_limb_t n = n_randtest_not_zero(state);
nmod_t mod;
nmod_init(&mod, n);
for (j = 0; j < len; j++)
{
vec[j] = n_randtest(state);
vec2[j] = vec[j];
}
_nmod_vec_reduce(vec, vec, len, mod);
for (j = 0; j < len; j++)
vec2[j] = n_mod2_preinv(vec2[j], mod.n, mod.ninv);
result = _nmod_vec_equal(vec, vec2, len);
if (!_nmod_vec_equal(vec, vec2, len))
{
printf("FAIL:\n");
printf("len = %ld, n = %ld\n", len, n);
abort();
}
_nmod_vec_clear(vec);
_nmod_vec_clear(vec2);
}
flint_randclear(state);
printf("PASS\n");
return 0;
}
void _nmod_vec_scalar_mul_nmod(mp_ptr res, mp_srcptr vec,
slong len, mp_limb_t c, nmod_t mod)
{
if (mod.norm >= FLINT_BITS/2) /* products will fit in a limb */
{
mpn_mul_1(res, vec, len, c);
_nmod_vec_reduce(res, res, len, mod);
} else /* products may take two limbs */
{
slong i;
for (i = 0; i < len; i++)
{
mp_limb_t hi, lo;
umul_ppmm(hi, lo, vec[i], c);
NMOD_RED2(res[i], hi, lo, mod); /* hi already reduced mod n */
}
}
}
void _nmod_poly_divrem_q1(mp_ptr Q, mp_ptr R,
mp_srcptr A, long lenA, mp_srcptr B, long lenB,
nmod_t mod)
{
const mp_limb_t invL = (B[lenB-1] == 1) ? 1 : n_invmod(B[lenB-1], mod.n);
if (lenB == 1)
{
_nmod_vec_scalar_mul_nmod(Q, A, lenA, invL, mod);
}
else
{
mp_limb_t t;
Q[1] = n_mulmod2_preinv(A[lenA-1], invL, mod.n, mod.ninv);
t = n_mulmod2_preinv(Q[1], B[lenB-2], mod.n, mod.ninv);
t = n_submod(A[lenA-2], t, mod.n);
Q[0] = n_mulmod2_preinv(t, invL, mod.n, mod.ninv);
if (FLINT_BITS + 2 <= 2 * mod.norm)
{
mpn_mul_1(R, B, lenB - 1, Q[0]);
if (lenB > 2)
mpn_addmul_1(R + 1, B, lenB - 2, Q[1]);
_nmod_vec_reduce(R, R, lenB - 1, mod);
}
else
{
_nmod_vec_scalar_mul_nmod(R, B, lenB - 1, Q[0], mod);
if (lenB > 2)
_nmod_vec_scalar_addmul_nmod(R + 1, B, lenB - 2, Q[1], mod);
}
_nmod_vec_sub(R, A, R, lenB - 1, mod);
}
}