#line 1 "matrix/black-box-linear-algebra.hpp"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <cstdlib>
#include <utility>
#include <vector>
using namespace std;
#line 2 "fps/berlekamp-massey.hpp"
#line 4 "fps/berlekamp-massey.hpp"
#include <iterator>
#line 6 "fps/berlekamp-massey.hpp"
using namespace std;
template <typename mint>
vector<mint> BerlekampMassey(const vector<mint> &s) {
const int N = (int)s.size();
vector<mint> b, c;
b.reserve(N + 1);
c.reserve(N + 1);
b.push_back(mint(1));
c.push_back(mint(1));
mint y = mint(1);
for (int ed = 1; ed <= N; ed++) {
int l = int(c.size()), m = int(b.size());
mint x = 0;
for (int i = 0; i < l; i++) x += c[i] * s[ed - l + i];
b.emplace_back(mint(0));
m++;
if (x == mint(0)) continue;
mint freq = x / y;
if (l < m) {
auto tmp = c;
c.insert(begin(c), m - l, mint(0));
for (int i = 0; i < m; i++) c[m - 1 - i] -= freq * b[m - 1 - i];
b = tmp;
y = x;
} else {
for (int i = 0; i < m; i++) c[l - 1 - i] -= freq * b[m - 1 - i];
}
}
reverse(begin(c), end(c));
return c;
}
#line 2 "fps/formal-power-series.hpp"
#line 8 "fps/formal-power-series.hpp"
using namespace std;
template <typename mint>
struct FormalPowerSeries : vector<mint> {
using vector<mint>::vector;
using FPS = FormalPowerSeries;
FPS &operator+=(const FPS &r) {
if (r.size() > this->size()) this->resize(r.size());
for (int i = 0; i < (int)r.size(); i++) (*this)[i] += r[i];
return *this;
}
FPS &operator+=(const mint &r) {
if (this->empty()) this->resize(1);
(*this)[0] += r;
return *this;
}
FPS &operator-=(const FPS &r) {
if (r.size() > this->size()) this->resize(r.size());
for (int i = 0; i < (int)r.size(); i++) (*this)[i] -= r[i];
return *this;
}
FPS &operator-=(const mint &r) {
if (this->empty()) this->resize(1);
(*this)[0] -= r;
return *this;
}
FPS &operator*=(const mint &v) {
for (int k = 0; k < (int)this->size(); k++) (*this)[k] *= v;
return *this;
}
FPS &operator/=(const FPS &r) {
if (this->size() < r.size()) {
this->clear();
return *this;
}
int n = this->size() - r.size() + 1;
if ((int)r.size() <= 64) {
FPS f(*this), g(r);
g.shrink();
mint coeff = g.back().inverse();
for (auto &x : g) x *= coeff;
int deg = (int)f.size() - (int)g.size() + 1;
int gs = g.size();
FPS quo(deg);
for (int i = deg - 1; i >= 0; i--) {
quo[i] = f[i + gs - 1];
for (int j = 0; j < gs; j++) f[i + j] -= quo[i] * g[j];
}
*this = quo * coeff;
this->resize(n, mint(0));
return *this;
}
return *this = ((*this).rev().pre(n) * r.rev().inv(n)).pre(n).rev();
}
FPS &operator%=(const FPS &r) {
*this -= *this / r * r;
shrink();
return *this;
}
FPS operator+(const FPS &r) const { return FPS(*this) += r; }
FPS operator+(const mint &v) const { return FPS(*this) += v; }
FPS operator-(const FPS &r) const { return FPS(*this) -= r; }
FPS operator-(const mint &v) const { return FPS(*this) -= v; }
FPS operator*(const FPS &r) const { return FPS(*this) *= r; }
FPS operator*(const mint &v) const { return FPS(*this) *= v; }
FPS operator/(const FPS &r) const { return FPS(*this) /= r; }
FPS operator%(const FPS &r) const { return FPS(*this) %= r; }
FPS operator-() const {
FPS ret(this->size());
for (int i = 0; i < (int)this->size(); i++) ret[i] = -(*this)[i];
return ret;
}
void shrink() {
while (this->size() && this->back() == mint(0)) this->pop_back();
}
FPS rev() const {
FPS ret(*this);
reverse(begin(ret), end(ret));
return ret;
}
FPS dot(FPS r) const {
FPS ret(min(this->size(), r.size()));
for (int i = 0; i < (int)ret.size(); i++) ret[i] = (*this)[i] * r[i];
return ret;
}
// 前 sz 項を取ってくる。sz に足りない項は 0 埋めする
FPS pre(int sz) const {
FPS ret(begin(*this), begin(*this) + min((int)this->size(), sz));
if ((int)ret.size() < sz) ret.resize(sz);
return ret;
}
FPS operator>>(int sz) const {
if ((int)this->size() <= sz) return {};
FPS ret(*this);
ret.erase(ret.begin(), ret.begin() + sz);
return ret;
}
FPS operator<<(int sz) const {
FPS ret(*this);
ret.insert(ret.begin(), sz, mint(0));
return ret;
}
FPS diff() const {
const int n = (int)this->size();
FPS ret(max(0, n - 1));
mint one(1), coeff(1);
for (int i = 1; i < n; i++) {
ret[i - 1] = (*this)[i] * coeff;
coeff += one;
}
return ret;
}
FPS integral() const {
const int n = (int)this->size();
FPS ret(n + 1);
ret[0] = mint(0);
if (n > 0) ret[1] = mint(1);
auto mod = mint::get_mod();
for (int i = 2; i <= n; i++) ret[i] = (-ret[mod % i]) * (mod / i);
for (int i = 0; i < n; i++) ret[i + 1] *= (*this)[i];
return ret;
}
mint eval(mint x) const {
mint r = 0, w = 1;
for (auto &v : *this) r += w * v, w *= x;
return r;
}
FPS log(int deg = -1) const {
assert(!(*this).empty() && (*this)[0] == mint(1));
if (deg == -1) deg = (int)this->size();
return (this->diff() * this->inv(deg)).pre(deg - 1).integral();
}
FPS pow(int64_t k, int deg = -1) const {
const int n = (int)this->size();
if (deg == -1) deg = n;
if (k == 0) {
FPS ret(deg);
if (deg) ret[0] = 1;
return ret;
}
for (int i = 0; i < n; i++) {
if ((*this)[i] != mint(0)) {
mint rev = mint(1) / (*this)[i];
FPS ret = (((*this * rev) >> i).log(deg) * k).exp(deg);
ret *= (*this)[i].pow(k);
ret = (ret << (i * k)).pre(deg);
if ((int)ret.size() < deg) ret.resize(deg, mint(0));
return ret;
}
if (__int128_t(i + 1) * k >= deg) return FPS(deg, mint(0));
}
return FPS(deg, mint(0));
}
static void *ntt_ptr;
static void set_fft();
FPS &operator*=(const FPS &r);
void ntt();
void intt();
void ntt_doubling();
static int ntt_pr();
FPS inv(int deg = -1) const;
FPS exp(int deg = -1) const;
};
template <typename mint>
void *FormalPowerSeries<mint>::ntt_ptr = nullptr;
template <int N>
struct FPSBackendPriority : FPSBackendPriority<N - 1> {};
template <>
struct FPSBackendPriority<0> {};
template <typename mint>
void FormalPowerSeries<mint>::set_fft() {
fps_set_fft_impl((FormalPowerSeries<mint>*)nullptr, FPSBackendPriority<1>{});
}
template <typename mint>
FormalPowerSeries<mint>& FormalPowerSeries<mint>::operator*=(const FPS& r) {
if (this->empty() || r.empty()) {
this->clear();
return *this;
}
return fps_multiply_impl(*this, r, FPSBackendPriority<1>{});
}
template <typename mint>
void FormalPowerSeries<mint>::ntt() {
fps_ntt_impl(*this, FPSBackendPriority<1>{});
}
template <typename mint>
void FormalPowerSeries<mint>::intt() {
fps_intt_impl(*this, FPSBackendPriority<1>{});
}
template <typename mint>
void FormalPowerSeries<mint>::ntt_doubling() {
fps_ntt_doubling_impl(*this, FPSBackendPriority<1>{});
}
template <typename mint>
int FormalPowerSeries<mint>::ntt_pr() {
return fps_ntt_pr_impl((FormalPowerSeries<mint>*)nullptr,
FPSBackendPriority<1>{});
}
template <typename mint>
FormalPowerSeries<mint> FormalPowerSeries<mint>::inv(int deg) const {
return fps_inv_impl(*this, deg, FPSBackendPriority<1>{});
}
template <typename mint>
FormalPowerSeries<mint> FormalPowerSeries<mint>::exp(int deg) const {
return fps_exp_impl(*this, deg, FPSBackendPriority<1>{});
}
/**
* @brief 多項式/形式的冪級数ライブラリ
*/
#line 2 "fps/mod-pow.hpp"
#line 4 "fps/mod-pow.hpp"
#line 2 "fps/arbitrary-fps.hpp"
#line 4 "fps/arbitrary-fps.hpp"
#line 2 "ntt/arbitrary-ntt.hpp"
#line 6 "ntt/arbitrary-ntt.hpp"
using namespace std;
#line 2 "modint/montgomery-modint.hpp"
#line 4 "modint/montgomery-modint.hpp"
#include <iostream>
template <uint32_t mod>
struct LazyMontgomeryModInt {
using mint = LazyMontgomeryModInt;
using i32 = int32_t;
using u32 = uint32_t;
using u64 = uint64_t;
static constexpr u32 get_r() {
u32 ret = mod;
for (i32 i = 0; i < 4; ++i) ret *= 2 - mod * ret;
return ret;
}
static constexpr u32 r = get_r();
static constexpr u32 n2 = -u64(mod) % mod;
static_assert(mod < (1 << 30), "invalid, mod >= 2 ^ 30");
static_assert((mod & 1) == 1, "invalid, mod % 2 == 0");
static_assert(r * mod == 1, "this code has bugs.");
u32 a;
constexpr LazyMontgomeryModInt() : a(0) {}
constexpr LazyMontgomeryModInt(const int64_t &b)
: a(reduce(u64(b % mod + mod) * n2)){};
static constexpr u32 reduce(const u64 &b) {
return (b + u64(u32(b) * u32(-r)) * mod) >> 32;
}
constexpr mint &operator+=(const mint &b) {
if (i32(a += b.a - 2 * mod) < 0) a += 2 * mod;
return *this;
}
constexpr mint &operator-=(const mint &b) {
if (i32(a -= b.a) < 0) a += 2 * mod;
return *this;
}
constexpr mint &operator*=(const mint &b) {
a = reduce(u64(a) * b.a);
return *this;
}
constexpr mint &operator/=(const mint &b) {
*this *= b.inverse();
return *this;
}
constexpr mint operator+(const mint &b) const { return mint(*this) += b; }
constexpr mint operator-(const mint &b) const { return mint(*this) -= b; }
constexpr mint operator*(const mint &b) const { return mint(*this) *= b; }
constexpr mint operator/(const mint &b) const { return mint(*this) /= b; }
constexpr bool operator==(const mint &b) const {
return (a >= mod ? a - mod : a) == (b.a >= mod ? b.a - mod : b.a);
}
constexpr bool operator!=(const mint &b) const {
return (a >= mod ? a - mod : a) != (b.a >= mod ? b.a - mod : b.a);
}
constexpr mint operator-() const { return mint() - mint(*this); }
constexpr mint operator+() const { return mint(*this); }
constexpr mint pow(u64 n) const {
mint ret(1), mul(*this);
while (n > 0) {
if (n & 1) ret *= mul;
mul *= mul;
n >>= 1;
}
return ret;
}
constexpr mint inverse() const {
int x = get(), y = mod, u = 1, v = 0, t = 0, tmp = 0;
while (y > 0) {
t = x / y;
x -= t * y, u -= t * v;
tmp = x, x = y, y = tmp;
tmp = u, u = v, v = tmp;
}
return mint{u};
}
friend std::ostream &operator<<(std::ostream &os, const mint &b) {
return os << b.get();
}
friend std::istream &operator>>(std::istream &is, mint &b) {
int64_t t;
is >> t;
b = LazyMontgomeryModInt<mod>(t);
return (is);
}
constexpr u32 get() const {
u32 ret = reduce(a);
return ret >= mod ? ret - mod : ret;
}
static constexpr u32 get_mod() { return mod; }
};
#line 2 "ntt/ntt.hpp"
#line 7 "ntt/ntt.hpp"
using namespace std;
template <typename mint>
struct NTT {
static constexpr uint32_t get_pr() {
uint32_t _mod = mint::get_mod();
using u64 = uint64_t;
u64 ds[32] = {};
int idx = 0;
u64 m = _mod - 1;
for (u64 i = 2; i * i <= m; ++i) {
if (m % i == 0) {
ds[idx++] = i;
while (m % i == 0) m /= i;
}
}
if (m != 1) ds[idx++] = m;
uint32_t _pr = 2;
while (1) {
int flg = 1;
for (int i = 0; i < idx; ++i) {
u64 a = _pr, b = (_mod - 1) / ds[i], r = 1;
while (b) {
if (b & 1) r = r * a % _mod;
a = a * a % _mod;
b >>= 1;
}
if (r == 1) {
flg = 0;
break;
}
}
if (flg == 1) break;
++_pr;
}
return _pr;
};
static constexpr uint32_t mod = mint::get_mod();
static constexpr uint32_t pr = get_pr();
static constexpr int level = __builtin_ctzll(mod - 1);
mint dw[level], dy[level];
void setwy(int k) {
mint w[level], y[level];
w[k - 1] = mint(pr).pow((mod - 1) / (1 << k));
y[k - 1] = w[k - 1].inverse();
for (int i = k - 2; i > 0; --i)
w[i] = w[i + 1] * w[i + 1], y[i] = y[i + 1] * y[i + 1];
dw[1] = w[1], dy[1] = y[1], dw[2] = w[2], dy[2] = y[2];
for (int i = 3; i < k; ++i) {
dw[i] = dw[i - 1] * y[i - 2] * w[i];
dy[i] = dy[i - 1] * w[i - 2] * y[i];
}
}
NTT() { setwy(level); }
void fft4(vector<mint> &a, int k) {
if ((int)a.size() <= 1) return;
if (k == 1) {
mint a1 = a[1];
a[1] = a[0] - a[1];
a[0] = a[0] + a1;
return;
}
if (k & 1) {
int v = 1 << (k - 1);
for (int j = 0; j < v; ++j) {
mint ajv = a[j + v];
a[j + v] = a[j] - ajv;
a[j] += ajv;
}
}
int u = 1 << (2 + (k & 1));
int v = 1 << (k - 2 - (k & 1));
mint one = mint(1);
mint imag = dw[1];
while (v) {
// jh = 0
{
int j0 = 0;
int j1 = v;
int j2 = j1 + v;
int j3 = j2 + v;
for (; j0 < v; ++j0, ++j1, ++j2, ++j3) {
mint t0 = a[j0], t1 = a[j1], t2 = a[j2], t3 = a[j3];
mint t0p2 = t0 + t2, t1p3 = t1 + t3;
mint t0m2 = t0 - t2, t1m3 = (t1 - t3) * imag;
a[j0] = t0p2 + t1p3, a[j1] = t0p2 - t1p3;
a[j2] = t0m2 + t1m3, a[j3] = t0m2 - t1m3;
}
}
// jh >= 1
mint ww = one, xx = one * dw[2], wx = one;
for (int jh = 4; jh < u;) {
ww = xx * xx, wx = ww * xx;
int j0 = jh * v;
int je = j0 + v;
int j2 = je + v;
for (; j0 < je; ++j0, ++j2) {
mint t0 = a[j0], t1 = a[j0 + v] * xx, t2 = a[j2] * ww,
t3 = a[j2 + v] * wx;
mint t0p2 = t0 + t2, t1p3 = t1 + t3;
mint t0m2 = t0 - t2, t1m3 = (t1 - t3) * imag;
a[j0] = t0p2 + t1p3, a[j0 + v] = t0p2 - t1p3;
a[j2] = t0m2 + t1m3, a[j2 + v] = t0m2 - t1m3;
}
xx *= dw[__builtin_ctzll((jh += 4))];
}
u <<= 2;
v >>= 2;
}
}
void ifft4(vector<mint> &a, int k) {
if ((int)a.size() <= 1) return;
if (k == 1) {
mint a1 = a[1];
a[1] = a[0] - a[1];
a[0] = a[0] + a1;
return;
}
int u = 1 << (k - 2);
int v = 1;
mint one = mint(1);
mint imag = dy[1];
while (u) {
// jh = 0
{
int j0 = 0;
int j1 = v;
int j2 = v + v;
int j3 = j2 + v;
for (; j0 < v; ++j0, ++j1, ++j2, ++j3) {
mint t0 = a[j0], t1 = a[j1], t2 = a[j2], t3 = a[j3];
mint t0p1 = t0 + t1, t2p3 = t2 + t3;
mint t0m1 = t0 - t1, t2m3 = (t2 - t3) * imag;
a[j0] = t0p1 + t2p3, a[j2] = t0p1 - t2p3;
a[j1] = t0m1 + t2m3, a[j3] = t0m1 - t2m3;
}
}
// jh >= 1
mint ww = one, xx = one * dy[2], yy = one;
u <<= 2;
for (int jh = 4; jh < u;) {
ww = xx * xx, yy = xx * imag;
int j0 = jh * v;
int je = j0 + v;
int j2 = je + v;
for (; j0 < je; ++j0, ++j2) {
mint t0 = a[j0], t1 = a[j0 + v], t2 = a[j2], t3 = a[j2 + v];
mint t0p1 = t0 + t1, t2p3 = t2 + t3;
mint t0m1 = (t0 - t1) * xx, t2m3 = (t2 - t3) * yy;
a[j0] = t0p1 + t2p3, a[j2] = (t0p1 - t2p3) * ww;
a[j0 + v] = t0m1 + t2m3, a[j2 + v] = (t0m1 - t2m3) * ww;
}
xx *= dy[__builtin_ctzll(jh += 4)];
}
u >>= 4;
v <<= 2;
}
if (k & 1) {
u = 1 << (k - 1);
for (int j = 0; j < u; ++j) {
mint ajv = a[j] - a[j + u];
a[j] += a[j + u];
a[j + u] = ajv;
}
}
}
void ntt(vector<mint> &a) {
if ((int)a.size() <= 1) return;
fft4(a, __builtin_ctz(a.size()));
}
void intt(vector<mint> &a) {
if ((int)a.size() <= 1) return;
ifft4(a, __builtin_ctz(a.size()));
mint iv = mint(a.size()).inverse();
for (auto &x : a) x *= iv;
}
vector<mint> multiply(const vector<mint> &a, const vector<mint> &b) {
int l = a.size() + b.size() - 1;
if (min<int>(a.size(), b.size()) <= 40) {
vector<mint> s(l);
for (int i = 0; i < (int)a.size(); ++i)
for (int j = 0; j < (int)b.size(); ++j) s[i + j] += a[i] * b[j];
return s;
}
int k = 2, M = 4;
while (M < l) M <<= 1, ++k;
setwy(k);
vector<mint> s(M);
for (int i = 0; i < (int)a.size(); ++i) s[i] = a[i];
fft4(s, k);
if (a.size() == b.size() && a == b) {
for (int i = 0; i < M; ++i) s[i] *= s[i];
} else {
vector<mint> t(M);
for (int i = 0; i < (int)b.size(); ++i) t[i] = b[i];
fft4(t, k);
for (int i = 0; i < M; ++i) s[i] *= t[i];
}
ifft4(s, k);
s.resize(l);
mint invm = mint(M).inverse();
for (int i = 0; i < l; ++i) s[i] *= invm;
return s;
}
void ntt_doubling(vector<mint> &a) {
int M = (int)a.size();
auto b = a;
intt(b);
mint r = 1, zeta = mint(pr).pow((mint::get_mod() - 1) / (M << 1));
for (int i = 0; i < M; i++) b[i] *= r, r *= zeta;
ntt(b);
copy(begin(b), end(b), back_inserter(a));
}
};
#line 10 "ntt/arbitrary-ntt.hpp"
namespace ArbitraryNTT {
using i64 = int64_t;
using u128 = __uint128_t;
constexpr int32_t m0 = 167772161;
constexpr int32_t m1 = 469762049;
constexpr int32_t m2 = 754974721;
using mint0 = LazyMontgomeryModInt<m0>;
using mint1 = LazyMontgomeryModInt<m1>;
using mint2 = LazyMontgomeryModInt<m2>;
constexpr int r01 = mint1(m0).inverse().get();
constexpr int r02 = mint2(m0).inverse().get();
constexpr int r12 = mint2(m1).inverse().get();
constexpr int r02r12 = i64(r02) * r12 % m2;
constexpr i64 w1 = m0;
constexpr i64 w2 = i64(m0) * m1;
template <typename T, typename submint>
vector<submint> mul(const vector<T> &a, const vector<T> &b) {
static NTT<submint> ntt;
vector<submint> s(a.size()), t(b.size());
for (int i = 0; i < (int)a.size(); ++i) s[i] = i64(a[i] % submint::get_mod());
for (int i = 0; i < (int)b.size(); ++i) t[i] = i64(b[i] % submint::get_mod());
return ntt.multiply(s, t);
}
template <typename T>
vector<int> multiply(const vector<T> &s, const vector<T> &t, int mod) {
auto d0 = mul<T, mint0>(s, t);
auto d1 = mul<T, mint1>(s, t);
auto d2 = mul<T, mint2>(s, t);
int n = d0.size();
vector<int> ret(n);
const int W1 = w1 % mod;
const int W2 = w2 % mod;
for (int i = 0; i < n; i++) {
int n1 = d1[i].get(), n2 = d2[i].get(), a = d0[i].get();
int b = i64(n1 + m1 - a) * r01 % m1;
int c = (i64(n2 + m2 - a) * r02r12 + i64(m2 - b) * r12) % m2;
ret[i] = (i64(a) + i64(b) * W1 + i64(c) * W2) % mod;
}
return ret;
}
template <typename mint>
vector<mint> multiply(const vector<mint> &a, const vector<mint> &b) {
if (a.size() == 0 && b.size() == 0) return {};
if (min<int>(a.size(), b.size()) < 128) {
vector<mint> ret(a.size() + b.size() - 1);
for (int i = 0; i < (int)a.size(); ++i)
for (int j = 0; j < (int)b.size(); ++j) ret[i + j] += a[i] * b[j];
return ret;
}
vector<int> s(a.size()), t(b.size());
for (int i = 0; i < (int)a.size(); ++i) s[i] = a[i].get();
for (int i = 0; i < (int)b.size(); ++i) t[i] = b[i].get();
vector<int> u = multiply<int>(s, t, mint::get_mod());
vector<mint> ret(u.size());
for (int i = 0; i < (int)u.size(); ++i) ret[i] = mint(u[i]);
return ret;
}
template <typename T>
vector<u128> multiply_u128(const vector<T> &s, const vector<T> &t) {
if (s.size() == 0 && t.size() == 0) return {};
if (min<int>(s.size(), t.size()) < 128) {
vector<u128> ret(s.size() + t.size() - 1);
for (int i = 0; i < (int)s.size(); ++i)
for (int j = 0; j < (int)t.size(); ++j) ret[i + j] += i64(s[i]) * t[j];
return ret;
}
auto d0 = mul<T, mint0>(s, t);
auto d1 = mul<T, mint1>(s, t);
auto d2 = mul<T, mint2>(s, t);
int n = d0.size();
vector<u128> ret(n);
for (int i = 0; i < n; i++) {
i64 n1 = d1[i].get(), n2 = d2[i].get();
i64 a = d0[i].get();
i64 b = (n1 + m1 - a) * r01 % m1;
i64 c = ((n2 + m2 - a) * r02r12 + (m2 - b) * r12) % m2;
ret[i] = a + b * w1 + u128(c) * w2;
}
return ret;
}
} // namespace ArbitraryNTT
#line 7 "fps/arbitrary-fps.hpp"
template <typename mint>
void fps_set_fft_impl(FormalPowerSeries<mint>*, FPSBackendPriority<0>) {
FormalPowerSeries<mint>::ntt_ptr = nullptr;
}
template <typename mint>
void fps_ntt_impl(FormalPowerSeries<mint>&, FPSBackendPriority<0>) {
exit(1);
}
template <typename mint>
void fps_intt_impl(FormalPowerSeries<mint>&, FPSBackendPriority<0>) {
exit(1);
}
template <typename mint>
void fps_ntt_doubling_impl(FormalPowerSeries<mint>&, FPSBackendPriority<0>) {
exit(1);
}
template <typename mint>
int fps_ntt_pr_impl(FormalPowerSeries<mint>*, FPSBackendPriority<0>) {
exit(1);
}
template <typename mint>
FormalPowerSeries<mint>& fps_multiply_impl(FormalPowerSeries<mint>& f,
const FormalPowerSeries<mint>& r,
FPSBackendPriority<0>) {
auto ret = ArbitraryNTT::multiply(f, r);
return f = FormalPowerSeries<mint>(ret.begin(), ret.end());
}
template <typename mint>
FormalPowerSeries<mint> fps_inv_impl(const FormalPowerSeries<mint>& f, int deg,
FPSBackendPriority<0>) {
assert(f[0] != mint(0));
if (deg == -1) deg = f.size();
FormalPowerSeries<mint> ret({mint(1) / f[0]});
for (int i = 1; i < deg; i <<= 1)
ret = (ret + ret - ret * ret * f.pre(i << 1)).pre(i << 1);
return ret.pre(deg);
}
template <typename mint>
FormalPowerSeries<mint> fps_exp_impl(const FormalPowerSeries<mint>& f, int deg,
FPSBackendPriority<0>) {
assert(f.size() == 0 || f[0] == mint(0));
if (deg == -1) deg = (int)f.size();
FormalPowerSeries<mint> ret({mint(1)});
for (int i = 1; i < deg; i <<= 1) {
ret = (ret * (f.pre(i << 1) + mint(1) - ret.log(i << 1))).pre(i << 1);
}
return ret.pre(deg);
}
#line 6 "fps/mod-pow.hpp"
template <typename mint>
FormalPowerSeries<mint> mod_pow(int64_t k, const FormalPowerSeries<mint>& base,
const FormalPowerSeries<mint>& d) {
using fps = FormalPowerSeries<mint>;
assert(!d.empty());
auto inv = d.rev().inv();
auto quo = [&](const fps& poly) {
if (poly.size() < d.size()) return fps{};
int n = poly.size() - d.size() + 1;
return (poly.rev().pre(n) * inv.pre(n)).pre(n).rev();
};
fps res{1}, b(base);
while (k) {
if (k & 1) {
res *= b;
res -= quo(res) * d;
res.shrink();
}
b *= b;
b -= quo(b) * d;
b.shrink();
k >>= 1;
assert(b.size() + 1 <= d.size());
assert(res.size() + 1 <= d.size());
}
return res;
}
/**
* @brief Mod-Pow ($f(x)^k \mod g(x)$)
*/
#line 2 "misc/rng.hpp"
#line 5 "misc/rng.hpp"
#include <unordered_set>
#line 7 "misc/rng.hpp"
using namespace std;
#line 2 "internal/internal-seed.hpp"
#include <chrono>
using namespace std;
namespace nyaan_internal {
unsigned long long non_deterministic_seed() {
unsigned long long m =
chrono::duration_cast<chrono::nanoseconds>(
chrono::high_resolution_clock::now().time_since_epoch())
.count();
m ^= 9845834732710364265uLL;
m ^= m << 24, m ^= m >> 31, m ^= m << 35;
return m;
}
unsigned long long deterministic_seed() { return 88172645463325252UL; }
// 64 bit の seed 値を生成 (手元では seed 固定)
// 連続で呼び出すと同じ値が何度も返ってくるので注意
// #define RANDOMIZED_SEED するとシードがランダムになる
unsigned long long seed() {
#if defined(NyaanLocal) && !defined(RANDOMIZED_SEED)
return deterministic_seed();
#else
return non_deterministic_seed();
#endif
}
} // namespace nyaan_internal
#line 10 "misc/rng.hpp"
namespace my_rand {
using i64 = long long;
using u64 = unsigned long long;
// [0, 2^64 - 1)
u64 rng() {
static u64 _x = nyaan_internal::seed();
return _x ^= _x << 7, _x ^= _x >> 9;
}
// [l, r]
i64 rng(i64 l, i64 r) {
assert(l <= r);
return l + rng() % u64(r - l + 1);
}
// [l, r)
i64 randint(i64 l, i64 r) {
assert(l < r);
return l + rng() % u64(r - l);
}
// choose n numbers from [l, r) without overlapping
vector<i64> randset(i64 l, i64 r, i64 n) {
assert(l <= r && n <= r - l);
unordered_set<i64> s;
for (i64 i = n; i; --i) {
i64 m = randint(l, r + 1 - i);
if (s.find(m) != s.end()) m = r - i;
s.insert(m);
}
vector<i64> ret;
for (auto& x : s) ret.push_back(x);
sort(begin(ret), end(ret));
return ret;
}
// [0.0, 1.0)
double rnd() { return rng() * 5.42101086242752217004e-20; }
// [l, r)
double rnd(double l, double r) {
assert(l < r);
return l + rnd() * (r - l);
}
template <typename T>
void randshf(vector<T>& v) {
int n = v.size();
for (int i = 1; i < n; i++) swap(v[i], v[randint(0, i + 1)]);
}
} // namespace my_rand
using my_rand::randint;
using my_rand::randset;
using my_rand::randshf;
using my_rand::rnd;
using my_rand::rng;
#line 13 "matrix/black-box-linear-algebra.hpp"
//
namespace BBLAImpl {
template <typename mint>
mint inner_product(const FormalPowerSeries<mint>& a,
const FormalPowerSeries<mint>& b) {
mint res = 0;
int n = a.size();
assert(n == (int)b.size());
for (int i = 0; i < n; i++) res += a[i] * b[i];
return res;
}
template <typename mint>
FormalPowerSeries<mint> random_poly(int n) {
FormalPowerSeries<mint> res(n);
for (auto& x : res) x = randint(0, mint::get_mod());
return res;
}
template <typename mint>
struct ModMatrix : vector<FormalPowerSeries<mint>> {
using fps = FormalPowerSeries<mint>;
ModMatrix(int n) : vector<fps>(n, fps(n)) {}
inline void add(int i, int j, mint x) { (*this)[i][j] += x; }
friend fps operator*(const ModMatrix& m, const fps& r) {
int n = m.size();
assert(n == (int)r.size());
fps res(n);
for (int i = 0; i < n; i++)
for (int j = 0; j < n; j++) res[i] += m[i][j] * r[j];
return res;
}
void apply(int i, mint r) {
int n = (*this).size();
for (int j = 0; j < n; j++) (*this)[i][j] *= r;
}
};
template <typename mint>
struct SparseMatrix : vector<vector<pair<int, mint>>> {
using fps = FormalPowerSeries<mint>;
template <typename... Args>
SparseMatrix(Args... args) : vector<vector<pair<int, mint>>>(args...) {}
inline void add(int i, int j, mint x) { (*this)[i].emplace_back(j, x); }
friend fps operator*(const SparseMatrix& m, const fps& r) {
int n = m.size();
assert(n == (int)r.size());
fps res(n);
for (int i = 0; i < n; i++)
for (auto&& [j, x] : m[i]) res[i] += x * r[j];
return res;
}
void apply(int i, mint r) {
for (auto&& [_, x] : (*this)[i]) x *= r;
}
};
template <typename mint>
FormalPowerSeries<mint> vector_minpoly(
const vector<FormalPowerSeries<mint>>& b) {
assert(!b.empty());
int n = b.size(), m = b[0].size();
FormalPowerSeries<mint> u = random_poly<mint>(m), a(n);
for (int i = 0; i < n; i++) a[i] = inner_product(b[i], u);
auto mp = BerlekampMassey<mint>(a);
return {mp.begin(), mp.end()};
}
template <typename mint, typename Mat>
FormalPowerSeries<mint> mat_minpoly(const Mat& A) {
int n = A.size();
FormalPowerSeries<mint> u = random_poly<mint>(n);
vector<FormalPowerSeries<mint>> b(n * 2 + 1);
for (int i = 0; i < (int)b.size(); i++) b[i] = u, u = A * u;
FormalPowerSeries<mint> mp = vector_minpoly(b);
return mp;
}
// calculate A^k b
template <typename mint, typename Mat>
FormalPowerSeries<mint> fast_pow(const Mat& A, FormalPowerSeries<mint> b,
int64_t k) {
using fps = FormalPowerSeries<mint>;
int n = b.size();
fps mp = mat_minpoly<mint, Mat>(A);
fps c = mod_pow<mint>(k, fps{0, 1}, mp.rev());
fps res(n);
for (int i = 0; i < (int)c.size(); i++) res += b * c[i], b = A * b;
return res;
}
template <typename mint, typename Mat>
mint fast_det(const Mat& A) {
using fps = FormalPowerSeries<mint>;
int n = A.size();
fps D;
while (true) {
do {
D = random_poly<mint>(n);
} while (any_of(begin(D), end(D), [](mint x) { return x == mint(0); }));
Mat AD = A;
for (int i = 0; i < n; i++) AD.apply(i, D[i]);
fps mp = mat_minpoly<mint, Mat>(AD);
if (mp.back() == 0) return 0;
if ((int)mp.size() != n + 1) continue;
mint det = n & 1 ? -mp.back() : mp.back();
mint Ddet = 1;
for (auto& d : D) Ddet *= d;
return det / Ddet;
}
exit(1);
}
template <typename mint, typename Mat>
FormalPowerSeries<mint> fast_linear_equation(const Mat& A, const FormalPowerSeries<mint>& b) {
using fps = FormalPowerSeries<mint>;
int n = A.size();
fps mp = mat_minpoly<mint, Mat>(A).rev();
fps buf = b, res(n);
for (int i = 1; i < (int)mp.size(); i++) {
res = buf * mp[i];
buf = A * buf;
}
return buf * mp[0].inverse();
}
} // namespace BBLAImpl
using BBLAImpl::fast_det;
using BBLAImpl::fast_pow;
using BBLAImpl::ModMatrix;
using BBLAImpl::SparseMatrix;
/**
* @brief Black Box Linear Algebra
*/
Black Box Linear Algebra