void fmpz_invert(fmpz_t res, fmpz_t x, fmpz_t m)
{
if (m[0] == 0)
{
printf("Error: division by zero!\n");
abort();
}
fmpz_t s0, U, V, temp;
unsigned long size = fmpz_size(m);
// U may use fmpz_size(m) + 1 limbs after the sum, and gmp requires +1
U = fmpz_init(size + 2);
V = fmpz_init(size + 2);
s0 = fmpz_init(size + 2);
temp = fmpz_init(size + 2);
// U := (x%m) + abs(m)
// V := abs(m)
fmpz_abs(V, m);
fmpz_mod(U, x, V);
fmpz_add(U, U, V);
// Compute s0 such that 1 = s0 * x % m
mpn_gcdext(temp+1, s0+1, (long *) s0, U+1, fmpz_size(U), V+1, fmpz_size(V));
fmpz_mod(res, s0, m);
fmpz_clear(temp);
fmpz_clear(s0);
fmpz_clear(V);
fmpz_clear(U);
}
void fmpz_multi_mod_ui(unsigned long * out, fmpz_t in, fmpz_comb_t comb, fmpz_t ** temp)
{
ulong i, j, k;
ulong n = comb->n;
long log_comb;
ulong size;
mp_limb_t * ptr;
ulong num;
unsigned long num_primes = comb->num_primes;
if (num_primes == 1) // we are reducing modulo a single prime which is assumed to be big enough
{
if ((long)in[0] > 0L) out[0] = in[1];
else if ((long)in[0] < 0L) out[0] = comb->primes[0] - in[1];
else out[0] = 0L;
return;
}
log_comb = n - 1;
// find level in comb with entries bigger than the input integer
log_comb = 0;
if ((long) in[0] < 0L)
while ((fmpz_bits(in) >= fmpz_bits(comb->comb[log_comb][0]) - 1) && (log_comb < comb->n - 1)) log_comb++;
else
while (fmpz_cmpabs(in, comb->comb[log_comb][0]) >= 0) log_comb++;
num = (1L<<(n - log_comb - 1));
// set each entry of this level of temp to the input integer
for (i = 0; i < num; i++)
{
fmpz_set(temp[log_comb][i], in);
}
log_comb--;
num *= 2;
// fill in other entries of temp by taking entries of temp at higher level mod pairs from comb
while (log_comb > FLINT_LOG_MULTI_MOD_CUTOFF) // keep going until we reach the basecase
{
for (i = 0, j = 0; i < num; i += 2, j++)
{
fmpz_mod(temp[log_comb][i], temp[log_comb + 1][j], comb->comb[log_comb][i]);
fmpz_mod(temp[log_comb][i+1], temp[log_comb + 1][j], comb->comb[log_comb][i+1]);
}
num *= 2;
log_comb--;
}
// do basecase
num /= 2;
log_comb++;
ulong stride = (1L << (log_comb + 1));
for (i = 0, j = 0; j < num_primes; i++, j += stride)
{
fmpz_multi_mod_ui_basecase(out + j, temp[log_comb][i], comb->primes + j, FLINT_MIN(stride, num_primes - j));
}
}
static void dsum_p(
fmpz_t rop,
const fmpz *dinv, const fmpz *mu, long M, const long *C, long lenC,
const fmpz_t a, long ui, long vi, long n, long d, long p, long N)
{
long m, r, idx;
fmpz_t apm1, apow, f, g, P, PN;
fmpz_init(apm1);
fmpz_init(apow);
fmpz_init(f);
fmpz_init(g);
fmpz_init_set_ui(P, p);
fmpz_init(PN);
fmpz_pow_ui(PN, P, N);
fmpz_zero(rop);
r = 0;
m = (p * (ui + 1) - (vi + 1)) / d;
if (m <= M) /* Step {r = 0} */
{
idx = _bsearch(C, 0, lenC, m % p);
fmpz_powm_ui(apm1, a, p - 1, PN);
fmpz_one(apow);
fmpz_one(f);
fmpz_mod(rop, mu + idx + lenC * (m / p), PN);
}
for (r = 1, m += p; m <= M; r++, m += p)
{
idx = _bsearch(C, 0, lenC, m % p);
fmpz_mul(apow, apow, apm1);
fmpz_mod(apow, apow, PN);
fmpz_mul_ui(f, f, ui + 1 + (r - 1) * d);
fmpz_mod(f, f, PN);
fmpz_mul(g, f, dinv + r);
fmpz_mul(g, g, apow);
fmpz_mul(g, g, mu + idx + lenC * (m / p));
fmpz_mod(g, g, PN);
fmpz_add(rop, rop, g);
}
fmpz_mod(rop, rop, PN);
fmpz_clear(apm1);
fmpz_clear(apow);
fmpz_clear(f);
fmpz_clear(g);
fmpz_clear(P);
fmpz_clear(PN);
}
void
_nf_elem_mod_fmpz(nf_elem_t res, const nf_elem_t a, const fmpz_t mod, const nf_t nf)
{
if (nf_elem_is_zero(a, nf))
{
nf_elem_zero(res, nf);
return;
}
if (nf->flag & NF_LINEAR)
{
fmpz_mod(LNF_ELEM_NUMREF(res), LNF_ELEM_NUMREF(a), mod);
fmpz_one(LNF_ELEM_DENREF(res));
}
else if (nf->flag & NF_QUADRATIC)
{
_fmpz_vec_scalar_mod_fmpz(QNF_ELEM_NUMREF(res), QNF_ELEM_NUMREF(a), 3, mod);
fmpz_one(QNF_ELEM_DENREF(res));
}
else
{
fmpq_poly_fit_length(NF_ELEM(res), fmpq_poly_length(NF_ELEM(a)));
_fmpq_poly_set_length(NF_ELEM(res), fmpq_poly_length(NF_ELEM(a)));
_fmpz_vec_scalar_mod_fmpz(NF_ELEM(res)->coeffs, NF_ELEM(a)->coeffs, fmpq_poly_length(NF_ELEM(a)), mod);
fmpz_one(NF_ELEM_DENREF(res));
}
nf_elem_canonicalise(res, nf);
}
static void precompute_dinv_p(fmpz *list, long M, long d, long p, long N)
{
fmpz_one(list + 0);
if (M >= p)
{
fmpz_t P, PN;
long r;
fmpz_init_set_ui(P, p);
fmpz_init(PN);
fmpz_pow_ui(PN, P, N);
fmpz_set_ui(list + 1, d);
_padic_inv(list + 1, list + 1, P, N);
for (r = 2; r <= M / p; r++)
{
fmpz_mul(list + r, list + (r - 1), list + 1);
fmpz_mod(list + r, list + r, PN);
}
fmpz_clear(P);
fmpz_clear(PN);
}
}
void _fmpz_mod_poly_evaluate_fmpz(fmpz_t res, const fmpz *poly, slong len,
const fmpz_t a, const fmpz_t p)
{
if (len == 0)
{
fmpz_zero(res);
}
else if (len == 1 || fmpz_is_zero(a))
{
fmpz_set(res, poly);
}
else
{
slong i = len - 1;
fmpz_t t;
fmpz_init(t);
fmpz_set(res, poly + i);
for (i = len - 2; i >= 0; i--)
{
fmpz_mul(t, res, a);
fmpz_mod(t, t, p);
fmpz_add(res, poly + i, t);
}
fmpz_clear(t);
if (fmpz_cmpabs(res, p) >= 0)
fmpz_sub(res, res, p);
}
}
// Compute a*b mod m
void fmpz_mulmod(fmpz_t res, fmpz_t a, fmpz_t b, fmpz_t m)
{
fmpz_t ab = fmpz_init(fmpz_size(a) + fmpz_size(b));
fmpz_mul(ab, a, b);
fmpz_mod(res, ab, m);
fmpz_clear(ab);
}
static void entry(fmpz_t rop_u, long *rop_v,
const long *u, const long *v, const fmpz *a, const fmpz *dinv,
const fmpz **mu, long M, const long **C, const long *lenC,
long n, long d, long p, long N, long N2)
{
const long ku = diagfrob_k(u, n, d);
const long kv = diagfrob_k(v, n, d);
fmpz_t f, g, P;
fmpz_init(f);
fmpz_init(g);
fmpz_init_set_ui(P, p);
/*
Compute $g := (u'-1)! \alpha_{u+1,v+1}$ to precision $N2$.
*/
fmpz_fac_ui(f, ku - 1);
alpha(g, u, v, a, dinv, mu, M, C, lenC, n, d, p, N2);
fmpz_mul(g, f, g);
/*
Compute $f := (-1)^{u'+v'} (v'-1)!$ exactly.
*/
fmpz_fac_ui(f, kv - 1);
if ((ku + kv) % 2 != 0)
{
fmpz_neg(f, f);
}
/*
Set rop to the product of $f$ and $g^{-1} mod $p^N$.
*/
*rop_v = fmpz_remove(f, f, P) + n - fmpz_remove(g, g, P);
if (*rop_v >= N)
{
fmpz_zero(rop_u);
*rop_v = 0;
}
else
{
_padic_inv(g, g, P, N - *rop_v);
fmpz_mul(rop_u, f, g);
fmpz_pow_ui(f, P, N - *rop_v);
fmpz_mod(rop_u, rop_u, f);
}
fmpz_clear(f);
fmpz_clear(g);
fmpz_clear(P);
}
void
_padic_log_balanced(fmpz_t z, const fmpz_t y, long v, const fmpz_t p, long N)
{
fmpz_t pv, pN, r, t, u;
long val;
padic_inv_t S;
fmpz_init(pv);
fmpz_init(pN);
fmpz_init(r);
fmpz_init(t);
fmpz_init(u);
_padic_inv_precompute(S, p, N);
fmpz_set(t, y);
fmpz_set(pv, p);
fmpz_pow_ui(pN, p, N);
fmpz_zero(z);
val = 1;
/*
TODO: Abort earlier if larger than $p^N$, possible
with variable precision?
*/
while (!fmpz_is_zero(t))
{
fmpz_mul(pv, pv, pv);
fmpz_fdiv_qr(t, r, t, pv);
if (!fmpz_is_zero(t))
{
fmpz_mul(t, t, pv);
fmpz_add_ui(u, r, 1);
_padic_inv_precomp(u, u, S);
fmpz_mul(t, t, u);
fmpz_mod(t, t, pN);
}
if (!fmpz_is_zero(r))
{
fmpz_neg(r, r);
_padic_log_bsplit(r, r, val, p, N);
fmpz_sub(z, z, r);
}
val *= 2;
}
fmpz_clear(pv);
fmpz_clear(pN);
fmpz_clear(r);
fmpz_clear(t);
fmpz_clear(u);
_padic_inv_clear(S);
}
// Compute a/b mod m
void fmpz_divmod(fmpz_t res, fmpz_t a, fmpz_t b, fmpz_t m)
{
fmpz_t b_inv = fmpz_init(fmpz_size(m));
fmpz_invert(b_inv, b, m);
fmpz_t ab = fmpz_init(fmpz_size(a) + fmpz_size(b_inv));
fmpz_mul(ab, a, b_inv);
fmpz_mod(res, ab, m);
fmpz_clear(ab);
fmpz_clear(b_inv);
}
static void alpha(fmpz_t rop, const long *u, const long *v,
const fmpz *a, const fmpz *dinv, const fmpz **mu, long M, const long **C, const long *lenC,
long n, long d, long p, long N)
{
const long ku = diagfrob_k(u, n, d);
long i;
fmpz_t f, g, P, PN;
fmpz_init(f);
fmpz_init(g);
fmpz_init_set_ui(P, p);
fmpz_init(PN);
fmpz_pow_ui(PN, P, N);
fmpz_pow_ui(rop, P, ku);
fmpz_mod(rop, rop, PN);
for (i = 0; i <= n; i++)
{
long e = (p * (u[i] + 1) - (v[i] + 1)) / d;
fmpz_powm_ui(f, a + i, e, PN);
dsum(g, dinv, mu[i], M, C[i], lenC[i], a + i, u[i], v[i], n, d, p, N);
fmpz_mul(rop, rop, f);
fmpz_mul(rop, rop, g);
fmpz_mod(rop, rop, PN);
}
if (ku % 2 != 0 && !fmpz_is_zero(rop))
{
fmpz_sub(rop, PN, rop);
}
fmpz_clear(f);
fmpz_clear(g);
fmpz_clear(P);
fmpz_clear(PN);
}
void
elem_set_si(elem_ptr elem, long v, const ring_t ring)
{
switch (ring->type)
{
case TYPE_FMPZ:
fmpz_set_si(elem, v);
break;
case TYPE_LIMB:
*((mp_ptr) elem) = v;
break;
case TYPE_POLY:
elem_poly_set_si(elem, v, ring);
break;
case TYPE_MOD:
{
switch (RING_PARENT(ring)->type)
{
case TYPE_FMPZ:
fmpz_set_si(elem, v);
fmpz_mod(elem, elem, RING_MODULUS(ring));
break;
case TYPE_LIMB:
*((mp_ptr) elem) = nmod_set_si(v, ring->nmod);
break;
default:
NOT_IMPLEMENTED("set_si (mod)", ring);
}
}
break;
case TYPE_FRAC:
elem_set_si(NUMER(elem, ring), v, RING_NUMER(ring));
elem_one(DENOM(elem, ring), RING_DENOM(ring));
break;
case TYPE_COMPLEX:
elem_set_si(REALPART(elem, ring), v, ring->parent);
elem_zero(IMAGPART(elem, ring), ring->parent);
break;
default:
NOT_IMPLEMENTED("set_si", ring);
}
}
void _qadic_trace(fmpz_t rop, const fmpz *op, slong len,
const fmpz *a, const slong *j, slong lena, const fmpz_t pN)
{
const slong d = j[lena - 1];
slong i, l;
fmpz *t;
t = _fmpz_vec_init(d);
fmpz_set_ui(t + 0, d);
for (i = 1; i < d; i++)
{
for (l = lena - 2; l >= 0 && j[l] >= d - (i - 1); l--)
{
fmpz_addmul(t + i, t + (j[l] + i - d), a + l);
}
if (l >= 0 && j[l] == d - i)
{
fmpz_addmul_ui(t + i, a + l, i);
}
fmpz_neg(t + i, t + i);
fmpz_mod(t + i, t + i, pN);
}
fmpz_zero(rop);
for (i = 0; i < d; i++)
{
fmpz_addmul(rop, op + i, t + i);
}
fmpz_mod(rop, rop, pN);
_fmpz_vec_clear(t, d);
}
void
fmpz_randtest_mod(fmpz_t f, flint_rand_t state, const fmpz_t m)
{
fmpz_t t;
fmpz_init(t);
fmpz_randtest_unsigned(t, state, fmpz_bits(m) + 2);
fmpz_mod(t, t, m);
if (n_randlimb(state) & UWORD(1))
{
fmpz_sub(t, m, t);
fmpz_sub_ui(t, t, UWORD(1));
}
fmpz_set(f, t);
fmpz_clear(t);
}
void _fmpz_mod_poly_radix_init(fmpz **Rpow, fmpz **Rinv,
const fmpz *R, long lenR, long k,
const fmpz_t invL, const fmpz_t p)
{
const long degR = lenR - 1;
long i;
fmpz_t invLP;
fmpz *W;
fmpz_init_set(invLP, invL);
W = flint_malloc((1L << (k - 1)) * degR * sizeof(fmpz));
_fmpz_vec_set(Rpow[0], R, lenR);
for (i = 1; i < k; i++)
{
_fmpz_mod_poly_sqr(Rpow[i], Rpow[i - 1], degR * (1L << (i - 1)) + 1, p);
}
for (i = 0; i < k; i++)
{
const long lenQ = (1L << i) * degR;
long j;
/* W := rev{Rpow[i], lenQ} */
for (j = 0; j < lenQ; j++)
{
W[j] = Rpow[i][lenQ - j];
}
_fmpz_mod_poly_inv_series_newton(Rinv[i], W, lenQ, invLP, p);
/* invLP := inv{lead{R^{2^i}}} */
if (i != k - 1)
{
fmpz_mul(invLP, invLP, invLP);
fmpz_mod(invLP, invLP, p);
}
}
fmpz_clear(invLP);
flint_free(W);
}
void _fmpz_vec_scalar_smod_fmpz(fmpz *res, const fmpz *vec, slong len, const fmpz_t p)
{
slong i;
fmpz_t pdiv2;
fmpz_init(pdiv2);
fmpz_fdiv_q_2exp(pdiv2, p, 1);
for (i = 0; i < len; i++)
{
fmpz_mod(res + i, vec + i, p);
if (fmpz_cmp(res + i, pdiv2) > 0)
{
fmpz_sub(res + i, res + i, p);
}
}
fmpz_clear(pdiv2);
}
void _padic_reduce(padic_t rop, const padic_ctx_t ctx)
{
if (!fmpz_is_zero(padic_unit(rop)))
{
if (padic_val(rop) < ctx->N)
{
int alloc;
fmpz_t pow;
alloc = _padic_ctx_pow_ui(pow, ctx->N - padic_val(rop), ctx);
fmpz_mod(padic_unit(rop), padic_unit(rop), pow);
if (alloc)
fmpz_clear(pow);
}
else
{
fmpz_zero(padic_unit(rop));
padic_val(rop) = 0;
}
}
}
void _fmpz_mod_poly_div_basecase(fmpz *Q, fmpz *R,
const fmpz *A, long lenA, const fmpz *B, long lenB,
const fmpz_t invB, const fmpz_t p)
{
const long alloc = (R == NULL) ? lenA : 0;
long lenR = lenB - 1, iQ;
if (alloc)
R = _fmpz_vec_init(alloc);
if (R != A)
_fmpz_vec_set(R + lenR, A + lenR, lenA - lenR);
for (iQ = lenA - lenB; iQ >= 0; iQ--)
{
if (fmpz_is_zero(R + lenA - 1))
{
fmpz_zero(Q + iQ);
}
else
{
fmpz_mul(Q + iQ, R + lenA - 1, invB);
fmpz_mod(Q + iQ, Q + iQ, p);
_fmpz_vec_scalar_submul_fmpz(R + lenA - lenR - 1, B, lenR, Q + iQ);
_fmpz_vec_scalar_mod_fmpz(R + lenA - lenR - 1, R + lenA - lenR - 1, lenR, p);
}
if (lenR - 1 >= iQ)
{
B++;
lenR--;
}
lenA--;
}
if (alloc)
_fmpz_vec_clear(R, alloc);
}
int main()
{
long i, j;
int sign;
fmpz_t input;
fmpz_t result;
fmpz_t r1;
fmpz_t m1;
fmpz_t mprod;
ulong r2, m2;
flint_rand_t state;
printf("CRT_ui....");
fflush(stdout);
fmpz_init(input);
fmpz_init(result);
fmpz_init(r1);
fmpz_init(m1);
fmpz_init(mprod);
flint_randinit(state);
for (i = 0; i < 10000; i++)
{
long nprimes;
m2 = n_randtest_prime(state, 0);
nprimes = 1 + n_randint(state, 4);
fmpz_set_ui(m1, 1UL);
for (j = 0; j < nprimes; )
{
ulong t = n_randtest_prime(state, 0);
if (t != m2)
{
fmpz_mul_ui(m1, m1, t);
j++;
}
}
fmpz_mul_ui(mprod, m1, m2);
sign = n_randint(state, 2);
if (sign)
fmpz_randtest_mod_signed(input, state, mprod);
else
fmpz_randtest_mod(input, state, mprod);
fmpz_mod(r1, input, m1);
r2 = fmpz_fdiv_ui(input, m2);
if (sign)
fmpz_CRT_ui(result, r1, m1, r2, m2);
else
fmpz_CRT_ui_unsigned(result, r1, m1, r2, m2);
if (!fmpz_equal(result, input))
{
printf("FAIL:\n");
printf("m1: "); fmpz_print(m1); printf("\n");
printf("m2: %lu\n", m2);
printf("m1*m2: "); fmpz_print(mprod); printf("\n");
printf("input: "); fmpz_print(input); printf("\n");
printf("r1: "); fmpz_print(r1); printf("\n");
printf("r2: %lu\n", r2);
printf("result: "); fmpz_print(result); printf("\n");
printf("%ld Equalness: %d\n", i, fmpz_equal(result, input));
printf("\n");
abort();
}
}
fmpz_clear(input);
fmpz_clear(result);
fmpz_clear(r1);
fmpz_clear(m1);
fmpz_clear(mprod);
flint_randclear(state);
_fmpz_cleanup();
printf("PASS\n");
return 0;
}
char * _padic_get_str(char * str, const padic_t op, const padic_ctx_t ctx)
{
const fmpz * u = padic_unit(op);
const long v = padic_val(op);
if (fmpz_is_zero(u))
{
if (!str)
{
str = flint_malloc(2);
}
str[0] = '0';
str[1] = '\0';
return str;
}
if (ctx->mode == PADIC_TERSE)
{
if (v == 0)
{
str = fmpz_get_str(str, 10, u);
}
else if (v > 0)
{
fmpz_t t;
fmpz_init(t);
fmpz_pow_ui(t, ctx->p, v);
fmpz_mul(t, t, u);
str = fmpz_get_str(str, 10, t);
fmpz_clear(t);
}
else /* v < 0 */
{
fmpz_t t;
fmpz_init(t);
fmpz_pow_ui(t, ctx->p, -v);
str = _fmpq_get_str(str, 10, u, t);
fmpz_clear(t);
}
}
else if (ctx->mode == PADIC_SERIES)
{
char *s;
fmpz_t x;
fmpz_t d;
long j, N;
if (fmpz_sgn(u) < 0)
{
printf("ERROR (_padic_get_str). u < 0 in SERIES mode.\n");
abort();
}
N = fmpz_clog(u, ctx->p) + v;
if (!str)
{
long b = (N - v) * (2 * fmpz_sizeinbase(ctx->p, 10)
+ z_sizeinbase(FLINT_MAX(FLINT_ABS(v), FLINT_ABS(N)), 10)
+ 5) + 1;
str = flint_malloc(b);
if (!str)
{
printf("ERROR (_padic_get_str). Memory allocation failed.\n");
abort();
}
}
s = str;
fmpz_init(d);
fmpz_init(x);
fmpz_set(x, u);
/* Unroll first step */
j = 0;
{
fmpz_mod(d, x, ctx->p); /* d = u mod p^{j+1} */
fmpz_sub(x, x, d); /* x = x - d */
fmpz_divexact(x, x, ctx->p); /* x = x / p */
if (!fmpz_is_zero(d))
{
if (j + v != 0)
{
fmpz_get_str(s, 10, d);
while (*++s != '\0') ;
*s++ = '*';
fmpz_get_str(s, 10, ctx->p);
while (*++s != '\0') ;
*s++ = '^';
sprintf(s, "%ld", j + v);
while (*++s != '\0') ;
}
else
{
fmpz_get_str(s, 10, d);
while (*++s != '\0') ;
}
}
j++;
}
for ( ; !fmpz_is_zero(x); j++)
{
fmpz_mod(d, x, ctx->p); /* d = u mod p^{j+1} */
fmpz_sub(x, x, d); /* x = x - d */
fmpz_divexact(x, x, ctx->p); /* x = x / p */
if (!fmpz_is_zero(d))
{
if (j + v != 0)
{
*s++ = ' ';
*s++ = '+';
*s++ = ' ';
fmpz_get_str(s, 10, d);
while (*++s != '\0') ;
*s++ = '*';
fmpz_get_str(s, 10, ctx->p);
while (*++s != '\0') ;
*s++ = '^';
sprintf(s, "%ld", j + v);
while (*++s != '\0') ;
}
else
{
*s++ = ' ';
*s++ = '+';
*s++ = ' ';
fmpz_get_str(s, 10, d);
while (*++s != '\0') ;
}
}
}
fmpz_clear(x);
fmpz_clear(d);
}
else /* ctx->mode == PADIC_VAL_UNIT */
{
if (!str)
{
long b = fmpz_sizeinbase(u, 10) + fmpz_sizeinbase(ctx->p, 10)
+ z_sizeinbase(v, 10) + 4;
str = flint_malloc(b);
if (!str)
{
printf("ERROR (_padic_get_str). Memory allocation failed.\n");
abort();
}
}
if (v == 0)
{
str = fmpz_get_str(str, 10, u);
}
else if (v == 1)
{
char *s = str;
fmpz_get_str(s, 10, u);
while (*++s != '\0') ;
*s++ = '*';
fmpz_get_str(s, 10, ctx->p);
}
else
{
char *s = str;
fmpz_get_str(s, 10, u);
while (*++s != '\0') ;
*s++ = '*';
fmpz_get_str(s, 10, ctx->p);
while (*++s != '\0') ;
*s++ = '^';
sprintf(s, "%ld", v);
}
}
return str;
}
int test_jigsaw(const size_t lambda, const size_t kappa, int symmetric, aes_randstate_t randstate) {
printf("λ: %4zu, κ: %2zu, symmetric: %d …", lambda, kappa, symmetric);
gghlite_sk_t self;
gghlite_flag_t flags = GGHLITE_FLAGS_QUIET;
if (symmetric)
gghlite_init(self, lambda, kappa, kappa, 0x0, flags | GGHLITE_FLAGS_GOOD_G_INV, randstate);
else
gghlite_jigsaw_init(self, lambda, kappa, flags, randstate);
fmpz_t p; fmpz_init(p);
fmpz_poly_oz_ideal_norm(p, self->g, self->params->n, 0);
fmpz_t a[kappa];
fmpz_t acc; fmpz_init(acc);
fmpz_set_ui(acc, 1);
for(size_t k=0; k<kappa; k++) {
fmpz_init(a[k]);
fmpz_randm_aes(a[k], randstate, p);
fmpz_mul(acc, acc, a[k]);
fmpz_mod(acc, acc, p);
}
gghlite_clr_t e[kappa];
gghlite_enc_t u[kappa];
for(size_t k=0; k<kappa; k++) {
gghlite_clr_init(e[k]);
gghlite_enc_init(u[k], self->params);
}
gghlite_enc_t left;
gghlite_enc_init(left, self->params);
gghlite_enc_set_ui0(left, 1, self->params);
for(size_t k=0; k<kappa; k++) {
fmpz_poly_set_coeff_fmpz(e[k], 0, a[k]);
const int i = (gghlite_sk_is_symmetric(self)) ? 0 : k;
int group[GAMMA];
memset(group, 0, GAMMA * sizeof(int));
group[i] = 1;
gghlite_enc_set_gghlite_clr(u[k], self, e[k], 1, group, 1);
gghlite_enc_mul(left, self->params, left, u[k]);
}
gghlite_enc_t rght;
gghlite_enc_init(rght, self->params);
gghlite_enc_set_ui0(rght, 1, self->params);
fmpz_poly_t tmp; fmpz_poly_init(tmp);
fmpz_poly_set_coeff_fmpz(tmp, 0, acc);
gghlite_enc_set_gghlite_clr0(rght, self, tmp);
for(size_t k=0; k<kappa; k++) {
const int i = (gghlite_sk_is_symmetric(self)) ? 0 : k;
gghlite_enc_mul(rght, self->params, rght, self->z_inv[i]);
}
gghlite_enc_sub(rght, self->params, rght, left);
int status = 1 - gghlite_enc_is_zero(self->params, rght);
for(size_t i=0; i<kappa; i++) {
fmpz_clear(a[i]);
gghlite_clr_clear(e[i]);
gghlite_enc_clear(u[i]);
}
gghlite_enc_clear(left);
gghlite_enc_clear(rght);
gghlite_clr_clear(tmp);
fmpz_clear(p);
gghlite_sk_clear(self, 1);
if (status == 0)
printf(" PASS\n");
else
printf(" FAIL\n");
return status;
}
/**
* Testing the flexibility of large index sets with a smaller kappa
*/
int test_jigsaw_indices(const size_t lambda, const size_t kappa, const size_t gamma, aes_randstate_t randstate) {
printf("lambda: %d, kappa: %d, gamma: %d", (int) lambda, (int) kappa, (int) gamma);
gghlite_sk_t self;
gghlite_flag_t flags = GGHLITE_FLAGS_QUIET;
gghlite_jigsaw_init_gamma(self, lambda, kappa, gamma, flags, randstate);
fmpz_t p; fmpz_init(p);
fmpz_poly_oz_ideal_norm(p, self->g, self->params->n, 0);
/* partitioning of the universe set */
int partition[GAMMA];
for (int i = 0; i < gamma; i++) {
partition[i] = rand() % kappa;
}
fmpz_t a[kappa];
fmpz_t acc; fmpz_init(acc);
fmpz_set_ui(acc, 1);
for(size_t k=0; k<kappa; k++) {
fmpz_init(a[k]);
fmpz_randm_aes(a[k], randstate, p);
fmpz_mul(acc, acc, a[k]);
fmpz_mod(acc, acc, p);
}
gghlite_clr_t e[kappa];
gghlite_enc_t u[kappa];
for(size_t k=0; k<kappa; k++) {
gghlite_clr_init(e[k]);
gghlite_enc_init(u[k], self->params);
}
gghlite_enc_t left;
gghlite_enc_init(left, self->params);
gghlite_enc_set_ui0(left, 1, self->params);
for(size_t k=0; k<kappa; k++) {
fmpz_poly_set_coeff_fmpz(e[k], 0, a[k]);
const int i = (gghlite_sk_is_symmetric(self)) ? 0 : k;
int group[GAMMA];
memset(group, 0, GAMMA * sizeof(int));
for (int j = 0; j < gamma; j++) {
if (partition[j] == k) {
group[j] = 1;
}
}
gghlite_enc_set_gghlite_clr(u[k], self, e[k], 1, group, 1);
gghlite_enc_mul(left, self->params, left, u[k]);
}
gghlite_enc_t rght;
gghlite_enc_init(rght, self->params);
gghlite_enc_set_ui0(rght, 1, self->params);
fmpz_poly_t tmp; fmpz_poly_init(tmp);
fmpz_poly_set_coeff_fmpz(tmp, 0, acc);
gghlite_enc_set_gghlite_clr0(rght, self, tmp);
for(size_t k=0; k<gamma; k++) {
gghlite_enc_mul(rght, self->params, rght, self->z_inv[k]);
}
gghlite_enc_sub(rght, self->params, rght, left);
int status = 1 - gghlite_enc_is_zero(self->params, rght);
for(size_t i=0; i<kappa; i++) {
fmpz_clear(a[i]);
gghlite_clr_clear(e[i]);
gghlite_enc_clear(u[i]);
}
gghlite_enc_clear(left);
gghlite_enc_clear(rght);
gghlite_clr_clear(tmp);
fmpz_clear(p);
gghlite_sk_clear(self, 1);
if (status == 0)
printf(" PASS\n");
else
printf(" FAIL\n");
return status;
}
int main(int argc, char *argv[]) {
cmdline_params_t cmdline_params;
const char *name = "Jigsaw Puzzles";
parse_cmdline(cmdline_params, argc, argv, name, NULL);
print_header(name, cmdline_params);
aes_randstate_t randstate;
aes_randinit_seed(randstate, cmdline_params->shaseed, NULL);
uint64_t t = ggh_walltime(0);
uint64_t t_total = ggh_walltime(0);
uint64_t t_gen = 0;
gghlite_sk_t self;
gghlite_jigsaw_init(self,
cmdline_params->lambda,
cmdline_params->kappa,
cmdline_params->flags,
randstate);
printf("\n");
gghlite_params_print(self->params);
printf("\n---\n\n");
t_gen = ggh_walltime(t);
printf("1. GGH InstGen wall time: %8.2f s\n", ggh_seconds(t_gen));
t = ggh_walltime(0);
fmpz_t p; fmpz_init(p);
fmpz_poly_oz_ideal_norm(p, self->g, self->params->n, 0);
fmpz_t a[5];
for(long k=0; k<5; k++)
fmpz_init(a[k]);
fmpz_t acc; fmpz_init(acc);
fmpz_set_ui(acc, 1);
// reducing these elements to something small mod g is expensive, for benchmarketing purposes we
// hence avoid this costly step most of the time by doing it only twice and by then computing the
// remaining elements from those two elements using cheap operations. The reported times still
// refer to the expensive step for fairness.
fmpz_randm_aes(a[0], randstate, p);
fmpz_mul(acc, acc, a[0]);
fmpz_mod(acc, acc, p);
fmpz_randm_aes(a[1], randstate, p);
fmpz_mul(acc, acc, a[1]);
fmpz_mod(acc, acc, p);
fmpz_set(a[3], a[0]);
fmpz_set(a[4], a[1]);
for(long k=2; k<cmdline_params->kappa; k++) {
fmpz_mul(a[2], a[0], a[1]);
fmpz_add_ui(a[2], a[2], k);
fmpz_mod(a[2], a[2], p);
fmpz_mul(acc, acc, a[2]);
fmpz_mod(acc, acc, p);
fmpz_set(a[1], a[0]);
fmpz_set(a[0], a[2]);
}
printf("2. Sampling from U(Z_p) wall time: %8.2f s\n", ggh_seconds(ggh_walltime(t)));
gghlite_clr_t e[3];
gghlite_clr_init(e[0]);
gghlite_clr_init(e[1]);
gghlite_clr_init(e[2]);
gghlite_enc_t u_k;
gghlite_enc_init(u_k, self->params);
gghlite_enc_t left;
gghlite_enc_init(left, self->params);
gghlite_enc_set_ui0(left, 1, self->params);
fmpz_poly_set_coeff_fmpz(e[0], 0, a[3]);
int GAMMA = 20;
uint64_t t_enc = ggh_walltime(0);
int group0[GAMMA];
memset(group0, 0, GAMMA * sizeof(int));
group0[0] = 1;
gghlite_enc_set_gghlite_clr(u_k, self, e[0], 1, group0, 1, randstate);
t_enc = ggh_walltime(t_enc);
printf("3. Encoding wall time: %8.2f s (per elem)\n", ggh_seconds(t_enc));
fflush(0);
uint64_t t_mul = 0;
t = ggh_walltime(0);
gghlite_enc_mul(left, self->params, left, u_k);
t_mul += ggh_walltime(t);
fmpz_poly_set_coeff_fmpz(e[1], 0, a[4]);
int group1[GAMMA];
memset(group1, 0, GAMMA * sizeof(int));
group1[1] = 1;
gghlite_enc_set_gghlite_clr(u_k, self, e[1], 1, group1, 1, randstate);
t = ggh_walltime(0);
gghlite_enc_mul(left, self->params, left, u_k);
t_mul += ggh_walltime(t);
const mp_bitcnt_t prec = (self->params->n/4 < 8192) ? 8192 : self->params->n/4;
const oz_flag_t flags = (self->params->flags & GGHLITE_FLAGS_VERBOSE) ? OZ_VERBOSE : 0;
_fmpz_poly_oz_rem_small_iter(e[0], e[0], self->g, self->params->n, self->g_inv, prec, flags);
_fmpz_poly_oz_rem_small_iter(e[1], e[1], self->g, self->params->n, self->g_inv, prec, flags);
for(long k=2; k<cmdline_params->kappa; k++) {
fmpz_poly_oz_mul(e[2], e[0], e[1], self->params->n);
assert(fmpz_poly_degree(e[2])>=0);
fmpz_add_ui(e[2]->coeffs, e[2]->coeffs, k);
_fmpz_poly_oz_rem_small_iter(e[2], e[2], self->g, self->params->n, self->g_inv, prec, flags);
int groupk[GAMMA];
memset(groupk, 0, GAMMA * sizeof(int));
groupk[k] = 1;
gghlite_enc_set_gghlite_clr(u_k, self, e[2], 1, groupk, 1, randstate);
t = ggh_walltime(0);
gghlite_enc_mul(left, self->params, left, u_k);
t_mul += ggh_walltime(t);
fmpz_poly_set(e[1], e[0]);
fmpz_poly_set(e[0], e[2]);
}
printf("4. Multiplication wall time: %8.4f s\n", ggh_seconds(t_mul));
fflush(0);
t = ggh_walltime(0);
gghlite_enc_t rght;
gghlite_enc_init(rght, self->params);
gghlite_enc_set_ui0(rght, 1, self->params);
fmpz_poly_t tmp; fmpz_poly_init(tmp);
fmpz_poly_set_coeff_fmpz(tmp, 0, acc);
gghlite_enc_set_gghlite_clr0(rght, self, tmp, randstate);
for(long k=0; k<cmdline_params->kappa; k++) {
gghlite_enc_mul(rght, self->params, rght, self->z_inv[k]);
}
printf("5. RHS generation wall time: %8.2f s\n", ggh_seconds(ggh_walltime(t)));
t = ggh_walltime(0);
gghlite_enc_sub(rght, self->params, rght, left);
int status = 1 - gghlite_enc_is_zero(self->params, rght);
gghlite_clr_t clr; gghlite_clr_init(clr);
gghlite_enc_extract(clr, self->params, rght);
double size = fmpz_poly_2norm_log2(clr);
gghlite_clr_clear(clr);
printf("6. Checking correctness wall time: %8.2f s\n", ggh_seconds(ggh_walltime(t)));
printf(" Correct: %8s (%8.2f)\n\n", (status == 0) ? "TRUE" : "FALSE", size);
for(long i=0; i<5; i++) {
fmpz_clear(a[i]);
}
gghlite_clr_clear(e[0]);
gghlite_clr_clear(e[1]);
gghlite_clr_clear(e[2]);
gghlite_enc_clear(u_k);
fmpz_clear(acc);
gghlite_enc_clear(left);
gghlite_enc_clear(rght);
gghlite_clr_clear(tmp);
fmpz_clear(p);
gghlite_sk_clear(self, 1);
t_total = ggh_walltime(t_total);
printf("λ: %3ld, κ: %2ld, n: %6ld, seed: 0x%08lx, success: %d, gen: %10.2fs, enc: %8.2fs, mul: %8.4fs, time: %10.2fs\n",
cmdline_params->lambda, cmdline_params->kappa, self->params->n, cmdline_params->seed,
status==0,
ggh_seconds(t_gen), ggh_seconds(t_enc), ggh_seconds(t_mul), ggh_seconds(t_total));
aes_randclear(randstate);
mpfr_free_cache();
flint_cleanup();
return status;
}
void fmpz_multi_CRT_ui(fmpz_t output, unsigned long * residues, fmpz_comb_t comb, fmpz_t ** comb_temp)
{
ulong i, j, k;
ulong n = comb->n;
ulong num;
ulong log_res;
mp_limb_t * ptr;
ulong size;
ulong num_primes = comb->num_primes;
if (num_primes == 1) // the output is less than a single prime, so just output the result
{
unsigned long p = comb->primes[0];
if ((p - residues[0]) < residues[0]) fmpz_set_si(output, (long) (residues[0] - p));
else fmpz_set_ui(output, residues[0]);
return;
}
// first layer of reconstruction
num = (1L<<n);
mp_limb_t temps[3];
mp_limb_t temp2s[3];
for (i = 0, j = 0; i + 2 <= num_primes; i += 2, j++)
{
fmpz_set_ui(temps, residues[i]);
fmpz_set_ui(temp2s, fmpz_mod_ui(temps, comb->primes[i+1]));
fmpz_sub_ui_inplace(temp2s, residues[i + 1]);
temp2s[0] = -temp2s[0];
fmpz_mul(temps, temp2s, comb->res[0][j]);
fmpz_set_ui(temp2s, fmpz_mod_ui(temps, comb->primes[i+1]));
fmpz_mul_ui(temps, temp2s, comb->primes[i]);
fmpz_add_ui(comb_temp[0][j], temps, residues[i]);
}
if (i < num_primes) fmpz_set_ui(comb_temp[0][j], residues[i]);
// compute other layers of reconstruction
fmpz_t temp = (fmpz_t) flint_heap_alloc(2*num + 1);
fmpz_t temp2 = (fmpz_t) flint_heap_alloc(2*num + 1);
num /= 2;
log_res = 1;
while (log_res < n)
{
for (i = 0, j = 0; i < num; i += 2, j++)
{
if (fmpz_is_one(comb->comb[log_res-1][i+1]))
{
if (!fmpz_is_one(comb->comb[log_res-1][i])) fmpz_set(comb_temp[log_res][j], comb_temp[log_res-1][i]);
} else
{
fmpz_mod(temp2, comb_temp[log_res-1][i], comb->comb[log_res-1][i+1]);
fmpz_sub(temp, temp2, comb_temp[log_res-1][i+1]);
temp[0] = -temp[0];
fmpz_mul(temp2, temp, comb->res[log_res][j]);
fmpz_mod(temp, temp2, comb->comb[log_res-1][i+1]);
fmpz_mul(temp2, temp, comb->comb[log_res-1][i]);
fmpz_add(comb_temp[log_res][j], temp2, comb_temp[log_res-1][i]);
}
}
log_res++;
num /= 2;
}
// write out the output
__fmpz_multi_CRT_sign(comb_temp[log_res - 1][0], comb_temp[log_res - 1][0], comb);
fmpz_set(output, comb_temp[log_res - 1][0]);
flint_heap_free(temp2); //temp2
flint_heap_free(temp); //temp
}
int
main(void)
{
long l, len = 20;
long runs[] = {
100000000, 1000000, 1000000, 1000000, 100000,
100000, 10000, 10000, 10000, 1000,
100, 100, 10, 1, 1,
1, 1, 1, 1, 1
};
long N[] = {
1, 2, 4, 8, 16,
32, 64, 128, 256, 512,
1024, WORD(1) << 11, WORD(1) << 12, WORD(1) << 13, WORD(1) << 14,
WORD(1) << 15, WORD(1) << 16, WORD(1) << 17, WORD(1) << 18, WORD(1) << 19
};
long T[20] = {0};
flint_printf("Benchmark for p-adic exponential (rectangular).\n");
fflush(stdout);
for (l = 0; l < FLINT_MIN(17, len); l++)
{
FLINT_TEST_INIT(state);
long n = N[l], r;
clock_t c0, c1;
long double cputime;
fmpz_t p;
padic_ctx_t ctx;
padic_t d, z;
fmpz_init_set_ui(p, 17);
padic_ctx_init(ctx, p, n, n, PADIC_VAL_UNIT);
padic_init(d);
padic_init(z);
if (n > 1)
{
fmpz_t f = {WORD(3)}, pow;
fmpz_init(pow);
fmpz_pow_ui(pow, p, n - 1);
fmpz_pow_ui(padic_unit(d), f, 3 * n);
fmpz_mod(padic_unit(d), padic_unit(d), pow);
padic_val(d) = 1;
fmpz_clear(pow);
}
c0 = clock();
for (r = runs[l]; (r); r--)
{
padic_exp_rectangular(z, d, ctx);
padic_zero(z);
}
c1 = clock();
cputime = (long double) (c1 - c0) / (long double) CLOCKS_PER_SEC;
T[l] = (slong) (cputime * (1000000000 / runs[l]));
flint_printf("%2ld, %4XYXYXYXY, %9ld, %wd\n",
l, cputime, runs[l], T[l]);
padic_clear(d);
padic_clear(z);
fmpz_clear(p);
padic_ctx_clear(ctx);
flint_randclear(state);
}
flint_printf("Output as a list:\n");
for (l = 0; l < len; l++)
flint_printf("%wd, ", T[l]);
flint_printf("\n");
}
long _fmpz_poly_hensel_start_lift(fmpz_poly_factor_t lifted_fac, long *link,
fmpz_poly_t *v, fmpz_poly_t *w, const fmpz_poly_t f,
const nmod_poly_factor_t local_fac, long N)
{
const long r = local_fac->num;
long i, preve;
fmpz_t p, P;
fmpz_poly_t monic_f;
fmpz_init(p);
fmpz_init(P);
fmpz_poly_init(monic_f);
fmpz_set_ui(p, (local_fac->p + 0)->mod.n);
fmpz_pow_ui(P, p, N);
if (fmpz_is_one(fmpz_poly_lead(f)))
{
fmpz_poly_set(monic_f, f);
}
else if (fmpz_cmp_si(fmpz_poly_lead(f), -1) == 0)
{
fmpz_poly_neg(monic_f, f);
}
else
{
fmpz_t t;
fmpz_init(t);
fmpz_mod(t, fmpz_poly_lead(f), P);
if (fmpz_invmod(t, t, P) == 0)
{
printf("Exception in fmpz_poly_start_hensel_lift.\n");
abort();
}
fmpz_poly_scalar_mul_fmpz(monic_f, f, t);
fmpz_poly_scalar_mod_fmpz(monic_f, monic_f, P);
fmpz_clear(t);
}
fmpz_poly_hensel_build_tree(link, v, w, local_fac);
{
long *e, n = FLINT_CLOG2(N) + 1;
e = flint_malloc(n * sizeof(long));
for (e[i = 0] = N; e[i] > 1; i++)
e[i + 1] = (e[i] + 1) / 2;
for (i--; i > 0; i--)
{
fmpz_poly_hensel_lift_tree(link, v, w, monic_f, r,
p, e[i+1], e[i], 1);
}
if (N > 1)
{
fmpz_poly_hensel_lift_tree(link, v, w, monic_f, r,
p, e[i+1], e[i], 0);
}
preve = e[i+1];
flint_free(e);
}
/*
Now everything is lifted to p^N, we just need to
insert the factors into their correct places in lifted_fac.
*/
fmpz_poly_factor_fit_length(lifted_fac, r);
for (i = 0; i < 2*r - 2; i++)
{
if (link[i] < 0)
{
fmpz_poly_scalar_smod_fmpz(lifted_fac->p + (- link[i] - 1), v[i], P);
lifted_fac->exp[- link[i] - 1] = 1L;
}
}
lifted_fac->num = r;
fmpz_clear(p);
fmpz_clear(P);
fmpz_poly_clear(monic_f);
return preve;
}
int main(void)
{
int i, j, result;
ulong count = UWORD(0);
gmp_randstate_t st;
FLINT_TEST_INIT(state);
gmp_randinit_default(st);
flint_printf("factor_pp1....");
fflush(stdout);
for (i = 0; i < 50 * flint_test_multiplier(); i++) /* Test random numbers */
{
mp_bitcnt_t bits;
mpz_t m, n;
fmpz_t n1, n2, r;
mpz_init(n);
mpz_init(m);
fmpz_init(n1);
fmpz_init(n2);
fmpz_init(r);
do {
mpz_urandomb(n, st, n_randint(state, 128) + 2);
} while (flint_mpz_cmp_ui(n, 2) < 0);
do {
mpz_urandomb(m, st, n_randint(state, 50) + 2);
} while (!mpz_probab_prime_p(m, 20));
mpz_mul(n, n, m);
fmpz_set_mpz(n1, n);
bits = FLINT_MIN(fmpz_bits(n1), FLINT_BITS);
for (j = 0; j < 20; j++)
{
fmpz_factor_pp1(n2, n1, 10000, 10000, n_randbits(state, bits - 2) + 3);
if (fmpz_cmp_ui(n2, 1) > 0) break;
}
if (fmpz_cmp_ui(n2, 1) > 0)
{
count++;
fmpz_mod(r, n1, n2);
result = (fmpz_is_zero(r));
if (!result)
{
flint_printf("FAIL:\n");
flint_printf("n1 = ");
fmpz_print(n1);
flint_printf(", n2 = ");
fmpz_print(n2);
flint_printf("\n");
fmpz_print(r); flint_printf("\n");
abort();
}
}
fmpz_clear(n1);
fmpz_clear(n2);
fmpz_clear(r);
mpz_clear(m);
mpz_clear(n);
}
if (count < 49 * flint_test_multiplier())
{
flint_printf("FAIL:\n");
flint_printf("Only %wu numbers factored\n", count);
abort();
}
gmp_randclear(st);
FLINT_TEST_CLEANUP(state);
flint_printf("PASS\n");
return 0;
}
void _padic_teichmuller(fmpz_t rop, const fmpz_t op, const fmpz_t p, slong N)
{
if (fmpz_equal_ui(p, 2))
{
fmpz_one(rop);
}
else if (N == 1)
{
fmpz_mod(rop, op, p);
}
else
{
slong *a, i, n;
fmpz *pow, *u;
fmpz_t s, t;
fmpz_t inv;
fmpz_t pm1;
a = _padic_lifts_exps(&n, N);
pow = _fmpz_vec_init(2 * n);
u = pow + n;
_padic_lifts_pows(pow, a, n, p);
fmpz_init(s);
fmpz_init(t);
fmpz_init(inv);
fmpz_init(pm1);
fmpz_sub_ui(pm1, p, 1);
/* Compute reduced units for (p-1) */
{
fmpz_mod(u + 0, pm1, pow + 0);
}
for (i = 1; i < n; i++)
{
fmpz_mod(u + i, u + (i - 1), pow + i);
}
/* Run Newton iteration */
i = n - 1;
{
fmpz_mod(rop, op, pow + i);
fmpz_set(inv, pm1);
}
for (i--; i >= 0; i--)
{
/* Lift rop */
fmpz_powm(s, rop, p, pow + i);
fmpz_sub(s, s, rop);
fmpz_mul(t, s, inv);
fmpz_sub(rop, rop, t);
fmpz_mod(rop, rop, pow + i);
/* Lift inv */
if (i > 0)
{
fmpz_mul(s, inv, inv);
fmpz_mul(t, u + i, s);
fmpz_mul_2exp(inv, inv, 1);
fmpz_sub(inv, inv, t);
fmpz_mod(inv, inv, pow + i);
}
}
fmpz_clear(s);
fmpz_clear(t);
fmpz_clear(inv);
fmpz_clear(pm1);
_fmpz_vec_clear(pow, 2 * n);
flint_free(a);
}
}
void frob(const mpoly_t P, const ctx_t ctxFracQt,
const qadic_t t1, const qadic_ctx_t Qq,
prec_t *prec, const prec_t *prec_in,
int verbose)
{
const padic_ctx_struct *Qp = &Qq->pctx;
const fmpz *p = Qp->p;
const long a = qadic_ctx_degree(Qq);
const long n = P->n - 1;
const long d = mpoly_degree(P, -1, ctxFracQt);
const long b = gmc_basis_size(n, d);
long i, j, k;
/* Diagonal fibre */
padic_mat_t F0;
/* Gauss--Manin Connection */
mat_t M;
mon_t *bR, *bC;
fmpz_poly_t r;
/* Local solution */
fmpz_poly_mat_t C, Cinv;
long vC, vCinv;
/* Frobenius */
fmpz_poly_mat_t F;
long vF;
fmpz_poly_mat_t F1;
long vF1;
fmpz_poly_t cp;
clock_t c0, c1;
double c;
if (verbose)
{
printf("Input:\n");
printf(" P = "), mpoly_print(P, ctxFracQt), printf("\n");
printf(" p = "), fmpz_print(p), printf("\n");
printf(" t1 = "), qadic_print_pretty(t1, Qq), printf("\n");
printf("\n");
fflush(stdout);
}
/* Step 1 {M, r} *********************************************************/
c0 = clock();
mat_init(M, b, b, ctxFracQt);
fmpz_poly_init(r);
gmc_compute(M, &bR, &bC, P, ctxFracQt);
{
fmpz_poly_t t;
fmpz_poly_init(t);
fmpz_poly_set_ui(r, 1);
for (i = 0; i < M->m; i++)
for (j = 0; j < M->n; j++)
{
fmpz_poly_lcm(t, r, fmpz_poly_q_denref(
(fmpz_poly_q_struct *) mat_entry(M, i, j, ctxFracQt)));
fmpz_poly_swap(r, t);
}
fmpz_poly_clear(t);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Gauss-Manin connection:\n");
printf(" r(t) = "), fmpz_poly_print_pretty(r, "t"), printf("\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
{
qadic_t t;
qadic_init2(t, 1);
fmpz_poly_evaluate_qadic(t, r, t1, Qq);
if (qadic_is_zero(t))
{
printf("Exception (deformation_frob).\n");
printf("The resultant r evaluates to zero (mod p) at t1.\n");
abort();
}
qadic_clear(t);
}
/* Precisions ************************************************************/
if (prec_in != NULL)
{
*prec = *prec_in;
}
else
{
deformation_precisions(prec, p, a, n, d, fmpz_poly_degree(r));
}
if (verbose)
{
printf("Precisions:\n");
printf(" N0 = %ld\n", prec->N0);
printf(" N1 = %ld\n", prec->N1);
printf(" N2 = %ld\n", prec->N2);
printf(" N3 = %ld\n", prec->N3);
printf(" N3i = %ld\n", prec->N3i);
printf(" N3w = %ld\n", prec->N3w);
printf(" N3iw = %ld\n", prec->N3iw);
printf(" N4 = %ld\n", prec->N4);
printf(" m = %ld\n", prec->m);
printf(" K = %ld\n", prec->K);
printf(" r = %ld\n", prec->r);
printf(" s = %ld\n", prec->s);
printf("\n");
fflush(stdout);
}
/* Initialisation ********************************************************/
padic_mat_init2(F0, b, b, prec->N4);
fmpz_poly_mat_init(C, b, b);
fmpz_poly_mat_init(Cinv, b, b);
fmpz_poly_mat_init(F, b, b);
vF = 0;
fmpz_poly_mat_init(F1, b, b);
vF1 = 0;
fmpz_poly_init(cp);
/* Step 2 {F0} ***********************************************************/
{
padic_ctx_t pctx_F0;
fmpz *t;
padic_ctx_init(pctx_F0, p, FLINT_MIN(prec->N4 - 10, 0), prec->N4, PADIC_VAL_UNIT);
t = _fmpz_vec_init(n + 1);
c0 = clock();
mpoly_diagonal_fibre(t, P, ctxFracQt);
diagfrob(F0, t, n, d, prec->N4, pctx_F0, 0);
padic_mat_transpose(F0, F0);
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Diagonal fibre:\n");
printf(" P(0) = {"), _fmpz_vec_print(t, n + 1), printf("}\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
_fmpz_vec_clear(t, n + 1);
padic_ctx_clear(pctx_F0);
}
/* Step 3 {C, Cinv} ******************************************************/
/*
Compute C as a matrix over Z_p[[t]]. A is the same but as a series
of matrices over Z_p. Mt is the matrix -M^t, and Cinv is C^{-1}^t,
the local solution of the differential equation replacing M by Mt.
*/
c0 = clock();
{
const long K = prec->K;
padic_mat_struct *A;
gmde_solve(&A, K, p, prec->N3, prec->N3w, M, ctxFracQt);
gmde_convert_soln(C, &vC, A, K, p);
for(i = 0; i < K; i++)
padic_mat_clear(A + i);
free(A);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Local solution:\n");
printf(" Time for C = %f\n", c);
fflush(stdout);
}
c0 = clock();
{
const long K = (prec->K + (*p) - 1) / (*p);
mat_t Mt;
padic_mat_struct *Ainv;
mat_init(Mt, b, b, ctxFracQt);
mat_transpose(Mt, M, ctxFracQt);
mat_neg(Mt, Mt, ctxFracQt);
gmde_solve(&Ainv, K, p, prec->N3i, prec->N3iw, Mt, ctxFracQt);
gmde_convert_soln(Cinv, &vCinv, Ainv, K, p);
fmpz_poly_mat_transpose(Cinv, Cinv);
fmpz_poly_mat_compose_pow(Cinv, Cinv, *p);
for(i = 0; i < K; i++)
padic_mat_clear(Ainv + i);
free(Ainv);
mat_clear(Mt, ctxFracQt);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf(" Time for C^{-1} = %f\n", c);
printf("\n");
fflush(stdout);
}
/* Step 4 {F(t) := C(t) F(0) C(t^p)^{-1}} ********************************/
/*
Computes the product C(t) F(0) C(t^p)^{-1} modulo (p^{N_2}, t^K).
This is done by first computing the unit part of the product
exactly over the integers modulo t^K.
*/
c0 = clock();
{
fmpz_t pN;
fmpz_poly_mat_t T;
fmpz_init(pN);
fmpz_poly_mat_init(T, b, b);
for (i = 0; i < b; i++)
{
/* Find the unique k s.t. F0(i,k) is non-zero */
for (k = 0; k < b; k++)
if (!fmpz_is_zero(padic_mat_entry(F0, i, k)))
break;
if (k == b)
{
printf("Exception (frob). F0 is singular.\n\n");
abort();
}
for (j = 0; j < b; j++)
{
fmpz_poly_scalar_mul_fmpz(fmpz_poly_mat_entry(T, i, j),
fmpz_poly_mat_entry(Cinv, k, j),
padic_mat_entry(F0, i, k));
}
}
fmpz_poly_mat_mul(F, C, T);
fmpz_poly_mat_truncate(F, prec->K);
vF = vC + padic_mat_val(F0) + vCinv;
/* Canonicalise (F, vF) */
{
long v = fmpz_poly_mat_ord_p(F, p);
if (v == LONG_MAX)
{
printf("ERROR (deformation_frob). F(t) == 0.\n");
abort();
}
else if (v > 0)
{
fmpz_pow_ui(pN, p, v);
fmpz_poly_mat_scalar_divexact_fmpz(F, F, pN);
vF = vF + v;
}
}
/* Reduce (F, vF) modulo p^{N2} */
fmpz_pow_ui(pN, p, prec->N2 - vF);
fmpz_poly_mat_scalar_mod_fmpz(F, F, pN);
fmpz_clear(pN);
fmpz_poly_mat_clear(T);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Matrix for F(t):\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
/* Step 5 {G = r(t)^m F(t)} **********************************************/
c0 = clock();
{
fmpz_t pN;
fmpz_poly_t t;
fmpz_init(pN);
fmpz_poly_init(t);
fmpz_pow_ui(pN, p, prec->N2 - vF);
/* Compute r(t)^m mod p^{N2-vF} */
if (prec->denR == NULL)
{
fmpz_mod_poly_t _t;
fmpz_mod_poly_init(_t, pN);
fmpz_mod_poly_set_fmpz_poly(_t, r);
fmpz_mod_poly_pow(_t, _t, prec->m);
fmpz_mod_poly_get_fmpz_poly(t, _t);
fmpz_mod_poly_clear(_t);
}
else
{
/* TODO: We don't really need a copy */
fmpz_poly_set(t, prec->denR);
}
fmpz_poly_mat_scalar_mul_fmpz_poly(F, F, t);
fmpz_poly_mat_scalar_mod_fmpz(F, F, pN);
/* TODO: This should not be necessary? */
fmpz_poly_mat_truncate(F, prec->K);
fmpz_clear(pN);
fmpz_poly_clear(t);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Analytic continuation:\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
/* Steps 6 and 7 *********************************************************/
if (a == 1)
{
/* Step 6 {F(1) = r(t_1)^{-m} G(t_1)} ********************************/
c0 = clock();
{
const long N = prec->N2 - vF;
fmpz_t f, g, t, pN;
fmpz_init(f);
fmpz_init(g);
fmpz_init(t);
fmpz_init(pN);
fmpz_pow_ui(pN, p, N);
/* f := \hat{t_1}, g := r(\hat{t_1})^{-m} */
_padic_teichmuller(f, t1->coeffs + 0, p, N);
if (prec->denR == NULL)
{
_fmpz_mod_poly_evaluate_fmpz(g, r->coeffs, r->length, f, pN);
fmpz_powm_ui(t, g, prec->m, pN);
}
else
{
_fmpz_mod_poly_evaluate_fmpz(t, prec->denR->coeffs, prec->denR->length, f, pN);
}
_padic_inv(g, t, p, N);
/* F1 := g G(\hat{t_1}) */
for (i = 0; i < b; i++)
for (j = 0; j < b; j++)
{
const fmpz_poly_struct *poly = fmpz_poly_mat_entry(F, i, j);
const long len = poly->length;
if (len == 0)
{
fmpz_poly_zero(fmpz_poly_mat_entry(F1, i, j));
}
else
{
fmpz_poly_fit_length(fmpz_poly_mat_entry(F1, i, j), 1);
_fmpz_mod_poly_evaluate_fmpz(t, poly->coeffs, len, f, pN);
fmpz_mul(fmpz_poly_mat_entry(F1, i, j)->coeffs + 0, g, t);
fmpz_mod(fmpz_poly_mat_entry(F1, i, j)->coeffs + 0,
fmpz_poly_mat_entry(F1, i, j)->coeffs + 0, pN);
_fmpz_poly_set_length(fmpz_poly_mat_entry(F1, i, j), 1);
_fmpz_poly_normalise(fmpz_poly_mat_entry(F1, i, j));
}
}
vF1 = vF;
fmpz_poly_mat_canonicalise(F1, &vF1, p);
fmpz_clear(f);
fmpz_clear(g);
fmpz_clear(t);
fmpz_clear(pN);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Evaluation:\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
}
else
{
/* Step 6 {F(1) = r(t_1)^{-m} G(t_1)} ********************************/
c0 = clock();
{
const long N = prec->N2 - vF;
fmpz_t pN;
fmpz *f, *g, *t;
fmpz_init(pN);
f = _fmpz_vec_init(a);
g = _fmpz_vec_init(2 * a - 1);
t = _fmpz_vec_init(2 * a - 1);
fmpz_pow_ui(pN, p, N);
/* f := \hat{t_1}, g := r(\hat{t_1})^{-m} */
_qadic_teichmuller(f, t1->coeffs, t1->length, Qq->a, Qq->j, Qq->len, p, N);
if (prec->denR == NULL)
{
fmpz_t e;
fmpz_init_set_ui(e, prec->m);
_fmpz_mod_poly_compose_smod(g, r->coeffs, r->length, f, a,
Qq->a, Qq->j, Qq->len, pN);
_qadic_pow(t, g, a, e, Qq->a, Qq->j, Qq->len, pN);
fmpz_clear(e);
}
else
{
_fmpz_mod_poly_reduce(prec->denR->coeffs, prec->denR->length, Qq->a, Qq->j, Qq->len, pN);
_fmpz_poly_normalise(prec->denR);
_fmpz_mod_poly_compose_smod(t, prec->denR->coeffs, prec->denR->length, f, a,
Qq->a, Qq->j, Qq->len, pN);
}
_qadic_inv(g, t, a, Qq->a, Qq->j, Qq->len, p, N);
/* F1 := g G(\hat{t_1}) */
for (i = 0; i < b; i++)
for (j = 0; j < b; j++)
{
const fmpz_poly_struct *poly = fmpz_poly_mat_entry(F, i, j);
const long len = poly->length;
fmpz_poly_struct *poly2 = fmpz_poly_mat_entry(F1, i, j);
if (len == 0)
{
fmpz_poly_zero(poly2);
}
else
{
_fmpz_mod_poly_compose_smod(t, poly->coeffs, len, f, a,
Qq->a, Qq->j, Qq->len, pN);
fmpz_poly_fit_length(poly2, 2 * a - 1);
_fmpz_poly_mul(poly2->coeffs, g, a, t, a);
_fmpz_mod_poly_reduce(poly2->coeffs, 2 * a - 1, Qq->a, Qq->j, Qq->len, pN);
_fmpz_poly_set_length(poly2, a);
_fmpz_poly_normalise(poly2);
}
}
/* Now the matrix for p^{-1} F_p at t=t_1 is (F1, vF1). */
vF1 = vF;
fmpz_poly_mat_canonicalise(F1, &vF1, p);
fmpz_clear(pN);
_fmpz_vec_clear(f, a);
_fmpz_vec_clear(g, 2 * a - 1);
_fmpz_vec_clear(t, 2 * a - 1);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Evaluation:\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
/* Step 7 {Norm} *****************************************************/
/*
Computes the matrix for $q^{-1} F_q$ at $t = t_1$ as the
product $F \sigma(F) \dotsm \sigma^{a-1}(F)$ up appropriate
transpositions because our convention of columns vs rows is
the opposite of that used by Gerkmann.
Note that, in any case, transpositions do not affect
the characteristic polynomial.
*/
c0 = clock();
{
const long N = prec->N1 - a * vF1;
fmpz_t pN;
fmpz_poly_mat_t T;
fmpz_init(pN);
fmpz_poly_mat_init(T, b, b);
fmpz_pow_ui(pN, p, N);
fmpz_poly_mat_frobenius(T, F1, 1, p, N, Qq);
_qadic_mat_mul(F1, F1, T, pN, Qq);
for (i = 2; i < a; i++)
{
fmpz_poly_mat_frobenius(T, T, 1, p, N, Qq);
_qadic_mat_mul(F1, F1, T, pN, Qq);
}
vF1 = a * vF1;
fmpz_poly_mat_canonicalise(F1, &vF1, p);
fmpz_clear(pN);
fmpz_poly_mat_clear(T);
}
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Norm:\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
}
/* Step 8 {Reverse characteristic polynomial} ****************************/
c0 = clock();
deformation_revcharpoly(cp, F1, vF1, n, d, prec->N0, prec->r, prec->s, Qq);
c1 = clock();
c = (double) (c1 - c0) / CLOCKS_PER_SEC;
if (verbose)
{
printf("Reverse characteristic polynomial:\n");
printf(" p(T) = "), fmpz_poly_print_pretty(cp, "T"), printf("\n");
printf(" Time = %f\n", c);
printf("\n");
fflush(stdout);
}
/* Clean up **************************************************************/
padic_mat_clear(F0);
mat_clear(M, ctxFracQt);
free(bR);
free(bC);
fmpz_poly_clear(r);
fmpz_poly_mat_clear(C);
fmpz_poly_mat_clear(Cinv);
fmpz_poly_mat_clear(F);
fmpz_poly_mat_clear(F1);
fmpz_poly_clear(cp);
}
int main(void)
{
int i, result;
FLINT_TEST_INIT(state);
flint_printf("sqrtmod....");
fflush(stdout);
for (i = 0; i < 100 * flint_test_multiplier(); i++) /* Test random integers */
{
int ans;
fmpz_t a, b, c, p;
mp_limb_t prime;
prime = n_randint(state, UWORD(1) << (FLINT_BITS - 1));
prime = n_nextprime(prime, 1);
fmpz_init(a);
fmpz_init(b);
fmpz_init(c);
fmpz_init(p);
fmpz_set_ui(p, prime);
fmpz_randm(a, state, p);
ans = fmpz_sqrtmod(b, a, p);
fmpz_mul(c, b, b);
fmpz_mod(c, c, p);
result = (ans == 0 || fmpz_equal(a, c));
if (!result)
{
flint_printf("FAIL (random):\n");
flint_printf("p = "), fmpz_print(p), flint_printf("\n");
flint_printf("a = "), fmpz_print(a), flint_printf("\n");
flint_printf("b = "), fmpz_print(b), flint_printf("\n");
flint_printf("c = "), fmpz_print(c), flint_printf("\n");
abort();
}
fmpz_clear(a);
fmpz_clear(b);
fmpz_clear(c);
fmpz_clear(p);
}
for (i = 0; i < 100 * flint_test_multiplier(); i++) /* Test random squares */
{
int ans;
fmpz_t a, b, c, d, p;
mp_limb_t prime;
prime = n_randint(state, UWORD(1) << (FLINT_BITS - 1));
prime = n_nextprime(prime, 1);
fmpz_init(a);
fmpz_init(b);
fmpz_init(c);
fmpz_init(d);
fmpz_init(p);
fmpz_set_ui(p, prime);
do
fmpz_randm(b, state, p);
while (fmpz_is_zero(b));
fmpz_mul(a, b, b);
fmpz_mod(a, a, p);
/* check a special case */
if (i == 0)
{
fmpz_set_str(p, "15951355998396157", 10);
fmpz_set_str(a, "7009303413761286", 10);
}
ans = fmpz_sqrtmod(c, a, p);
fmpz_mul(d, c, c);
fmpz_mod(d, d, p);
result = (ans && fmpz_equal(a, d));
if (!result)
{
flint_printf("FAIL (squares):\n");
flint_printf("p = "), fmpz_print(p), flint_printf("\n");
flint_printf("a (= b^2) = "), fmpz_print(a), flint_printf("\n");
flint_printf("b = "), fmpz_print(b), flint_printf("\n");
flint_printf("c (= sqrt(a) = "), fmpz_print(c), flint_printf("\n");
flint_printf("d (= c^2) = "), fmpz_print(d), flint_printf("\n");
flint_printf("ans = %d\n", ans);
abort();
}
fmpz_clear(a);
fmpz_clear(b);
fmpz_clear(c);
fmpz_clear(d);
fmpz_clear(p);
}
FLINT_TEST_CLEANUP(state);
flint_printf("PASS\n");
return 0;
}
