static void poly_mod_mul(UV* px, UV* py, UV* res, UV r, UV mod)
{
UV degpx, degpy;
UV i, j, pxi, pyj, rindex;
/* Determine max degree of px and py */
for (degpx = r-1; degpx > 0 && !px[degpx]; degpx--) ; /* */
for (degpy = r-1; degpy > 0 && !py[degpy]; degpy--) ; /* */
/* We can sum at least j values at once */
j = (mod >= HALF_WORD) ? 0 : (UV_MAX / ((mod-1)*(mod-1)));
if (j >= degpx || j >= degpy) {
/* res will be written completely, so no need to set */
for (rindex = 0; rindex < r; rindex++) {
UV sum = 0;
j = rindex;
for (i = 0; i <= degpx; i++) {
if (j <= degpy)
sum += px[i] * py[j];
j = (j == 0) ? r-1 : j-1;
}
res[rindex] = sum % mod;
}
} else {
memset(res, 0, r * sizeof(UV)); /* Zero result accumulator */
for (i = 0; i <= degpx; i++) {
pxi = px[i];
if (pxi == 0) continue;
if (mod < HALF_WORD) {
for (j = 0; j <= degpy; j++) {
pyj = py[j];
rindex = i+j; if (rindex >= r) rindex -= r;
res[rindex] = (res[rindex] + (pxi*pyj) ) % mod;
}
} else {
for (j = 0; j <= degpy; j++) {
pyj = py[j];
rindex = i+j; if (rindex >= r) rindex -= r;
res[rindex] = muladdmod(pxi, pyj, res[rindex], mod);
}
}
}
}
memcpy(px, res, r * sizeof(UV)); /* put result in px */
}
/* Bach and Sorenson (1993) would be better */
static int is_perfect_power(UV n) {
UV b, last;
if ((n <= 3) || (n == UV_MAX)) return 0;
if ((n & (n-1)) == 0) return 1; /* powers of 2 */
last = log2floor(n-1) + 1;
#if (BITS_PER_WORD == 32) || (DBL_DIG > 19)
if (1) {
#elif DBL_DIG == 10
if (n < UVCONST(10000000000)) {
#elif DBL_DIG == 15
if (n < UVCONST(1000000000000000)) {
#else
if ( n < (UV) pow(10, DBL_DIG) ) {
#endif
/* Simple floating point method. Fast, but need enough mantissa. */
b = sqrt(n)+0.5; if (b*b == n) return 1; /* perfect square */
for (b = 3; b < last; b = _XS_next_prime(b)) {
UV root = pow(n, 1.0 / (double)b) + 0.5;
if ( ((UV)(pow(root, b)+0.5)) == n) return 1;
}
} else {
/* Dietzfelbinger, algorithm 2.3.5 (without optimized exponential) */
for (b = 2; b <= last; b++) {
UV a = 1;
UV c = n;
while (c >= HALF_WORD) c = (1+c)>>1;
while ((c-a) >= 2) {
UV m, maxm, p, i;
m = (a+c) >> 1;
maxm = UV_MAX / m;
p = m;
for (i = 2; i <= b; i++) {
if (p > maxm) { p = n+1; break; }
p *= m;
}
if (p == n) return 1;
if (p < n)
a = m;
else
c = m;
}
}
}
return 0;
}
static UV order(UV r, UV n, UV limit) {
UV j;
UV t = 1;
for (j = 1; j <= limit; j++) {
t = (t * n) % r;
if (t == 1)
break;
}
return j;
}
static void poly_print(UV* poly, UV r)
{
int i;
for (i = r-1; i >= 1; i--) {
if (poly[i] != 0)
printf("%lux^%d + ", poly[i], i);
}
if (poly[0] != 0) printf("%lu", poly[0]);
printf("\n");
}
static void poly_mod_mul(UV* px, UV* py, UV* res, UV r, UV mod)
{
UV i, j, pxi, pyj, rindex;
memset(res, 0, r * sizeof(UV));
for (i = 0; i < r; i++) {
pxi = px[i];
if (pxi == 0) continue;
for (j = 0; j < r; j++) {
pyj = py[j];
if (pyj == 0) continue;
rindex = (i+j) < r ? i+j : i+j-r; /* (i+j) % r */
if (mod < HALF_WORD) {
res[rindex] = (res[rindex] + (pxi*pyj) ) % mod;
} else {
res[rindex] = muladdmod(pxi, pyj, res[rindex], mod);
}
}
}
memcpy(px, res, r * sizeof(UV)); /* put result in px */
}
