Tifa's CP Library

:heavy_check_mark: polymtt (src/code/poly/polymtt.hpp)

Depends on

Verified with

Code

#ifndef TIFALIBS_POLY_POLYMTT
#define TIFALIBS_POLY_POLYMTT

#include "../conv/conv_mtt.hpp"
#include "poly.hpp"

namespace tifa_libs::math {
namespace polymtt_impl_ {
template <class FP = f64>
struct cconv_mtt : public FFT<FP> {
  static constexpr auto ct_cat = ct_FFT;
  template <class mint>
  constexpr void conv(vec<mint>& l, vec<mint> const& r, u32 sz = 0) { l = conv_mtt(*this, l, r, sz); }
};
}  // namespace polymtt_impl_

template <class mint, class FP = f64>
using polymtt = poly<mint, polymtt_impl_::cconv_mtt<FP>>;

}  // namespace tifa_libs::math

#endif
#line 1 "src/code/poly/polymtt.hpp"



#line 1 "src/code/conv/conv_mtt.hpp"



#line 1 "src/code/conv/conv_naive.hpp"



#line 1 "src/code/util/util.hpp"



#include <bits/stdc++.h>

template <class T>
constexpr T abs(T x) { return x < 0 ? -x : x; }

using i8 = int8_t;
using i16 = int16_t;
using i32 = int32_t;
using i64 = int64_t;
using i128 = __int128_t;
using isz = ptrdiff_t;

using u8 = uint8_t;
using u16 = uint16_t;
using u32 = uint32_t;
using u64 = uint64_t;
using u128 = __uint128_t;
using usz = size_t;

using f32 = float;
using f64 = double;
using f128 = long double;

template <class T>
using ptt = std::pair<T, T>;
template <class T>
using pt3 = std::tuple<T, T, T>;
template <class T>
using pt4 = std::tuple<T, T, T, T>;

template <class T, usz N>
using arr = std::array<T, N>;
template <class T>
using vec = std::vector<T>;
template <class T>
using vvec = vec<vec<T>>;
template <class T>
using v3ec = vec<vvec<T>>;
template <class U, class T>
using vecp = vec<std::pair<U, T>>;
template <class U, class T>
using vvecp = vvec<std::pair<U, T>>;
template <class T>
using vecpt = vec<ptt<T>>;
template <class T>
using vvecpt = vvec<ptt<T>>;

template <class T, class C = std::less<T>>
using pq = std::priority_queue<T, vec<T>, C>;
template <class T>
using pqg = std::priority_queue<T, vec<T>, std::greater<T>>;

using strn = std::string;
using strnv = std::string_view;

using vecu = vec<u32>;
using vvecu = vvec<u32>;
using v3ecu = v3ec<u32>;
using vecu64 = vec<u64>;
using vecb = vec<bool>;
using vvecb = vvec<bool>;

#ifdef ONLINE_JUDGE
#undef assert
#define assert(x) 42
#endif

using namespace std::literals;

constexpr i8 operator""_i8(unsigned long long x) { return (i8)x; }
constexpr i16 operator""_i16(unsigned long long x) { return (i16)x; }
constexpr i32 operator""_i32(unsigned long long x) { return (i32)x; }
constexpr i64 operator""_i64(unsigned long long x) { return (i64)x; }
constexpr isz operator""_iz(unsigned long long x) { return (isz)x; }

constexpr u8 operator""_u8(unsigned long long x) { return (u8)x; }
constexpr u16 operator""_u16(unsigned long long x) { return (u16)x; }
constexpr u32 operator""_u32(unsigned long long x) { return (u32)x; }
constexpr u64 operator""_u64(unsigned long long x) { return (u64)x; }
constexpr usz operator""_uz(unsigned long long x) { return (usz)x; }

inline const auto fn_0 = [](auto&&...) {};


#line 5 "src/code/conv/conv_naive.hpp"

namespace tifa_libs::math {

template <class U, class T = U>
requires(sizeof(U) <= sizeof(T))
constexpr vec<T> conv_naive(vec<U> const &l, vec<U> const &r, u32 ans_size = 0) {
  if (l.empty() || r.empty()) return {};
  if (!ans_size) ans_size = u32(l.size() + r.size() - 1);
  u32 n = (u32)l.size(), m = (u32)r.size();
  vec<T> ans(ans_size);
  if (n < m)
    for (u32 j = 0; j < m; ++j)
      for (u32 i = 0; i < n; ++i) {
        if (i + j >= ans_size) break;
        ans[i + j] += (T)l[i] * (T)r[j];
      }
  else
    for (u32 i = 0; i < n; ++i)
      for (u32 j = 0; j < m; ++j) {
        if (i + j >= ans_size) break;
        ans[i + j] += (T)l[i] * (T)r[j];
      }
  return ans;
}

}  // namespace tifa_libs::math


#line 1 "src/code/conv/fft.hpp"



#line 5 "src/code/conv/fft.hpp"

namespace tifa_libs::math {

template <std::floating_point FP>
struct FFT {
  using C = std::complex<FP>;
  using data_t = C;

  explicit constexpr FFT() : rev(), w() {}

  constexpr u32 size() const { return (u32)rev.size(); }
  constexpr void bzr(u32 len) {
    u32 n = std::max<u32>(std::bit_ceil(len), 2);
    if (n == size()) return;
    rev.resize(n, 0);
    u32 k = (u32)(std::bit_width(n) - 1);
    for (u32 i = 0; i < n; ++i) rev[i] = (rev[i / 2] / 2) | ((i & 1) << (k - 1));
    w.resize(n);
    w[0].real(1);
    for (u32 i = 1; i < n; ++i) w[i] = {std::cos(TAU * (FP)i / (FP)n), std::sin(TAU * (FP)i / (FP)n)};
  }

  constexpr void dif(vec<C> &f, u32 n = 0) const {
    if (!n) n = size();
    if (f.size() < n) f.resize(n);
    assert(n <= size());
    for (u32 i = 0; i < n; ++i)
      if (i < rev[i]) std::swap(f[rev[i]], f[i]);
#pragma GCC diagnostic ignored "-Wsign-conversion"
    for (u32 i = 2, d = n / 2; i <= n; i *= 2, d /= 2)
      for (u32 j = 0; j < n; j += i) {
        auto l = f.begin() + j, r = f.begin() + j + i / 2;
        auto p = w.begin();
        for (u32 k = 0; k < i / 2; ++k, ++l, ++r, p += d) {
          C tmp = *r * *p;
          *r = *l - tmp;
          *l = *l + tmp;
        }
      }
#pragma GCC diagnostic warning "-Wsign-conversion"
  }
  constexpr void dit(vec<C> &f, u32 n = 0) const {
    if (!n) n = size();
    dif(f, n);
    for (u32 i = 0; i < n; ++i) f[i] /= (FP)n;
  }

 private:
  const FP TAU = std::acos((FP)-1.) * 2;

  vecu rev;
  vec<C> w;
};

}  // namespace tifa_libs::math


#line 6 "src/code/conv/conv_mtt.hpp"

namespace tifa_libs::math {

template <class mint, class FP>
constexpr vec<mint> conv_mtt(FFT<FP> &fft, vec<mint> const &l, vec<mint> const &r, u32 ans_size = 0) {
  if (!ans_size) ans_size = u32(l.size() + r.size() - 1);
  if (ans_size < 32) return conv_naive(l, r, ans_size);
  using C = typename FFT<FP>::C;
  if (l.size() == 1) {
    vec<mint> ans = r;
    ans.resize(ans_size);
    for (auto &i : ans) i *= l[0];
    return ans;
  }
  if (r.size() == 1) {
    vec<mint> ans = l;
    ans.resize(ans_size);
    for (auto &i : ans) i *= r[0];
    return ans;
  }
  fft.bzr(std::max({(u32)l.size(), (u32)r.size(), std::min(u32(l.size() + r.size() - 1), ans_size)}));
  u32 n = fft.size();
  const int OFS = ((int)sizeof(decltype(mint::mod())) * 8 - std::countl_zero(mint::mod() - 1) + 1) / 2;
  const u32 MSK = ((1u << OFS) - 1);
  vec<mint> ans(ans_size);
  vec<C> a(n), b(n);
  for (u32 i = 0; i < l.size(); ++i) a[i] = {(FP)(l[i].val() & MSK), (FP)(l[i].val() >> OFS)};
  for (u32 i = 0; i < r.size(); ++i) b[i] = {(FP)(r[i].val() & MSK), (FP)(r[i].val() >> OFS)};
  fft.dif(a);
  fft.dif(b);
  {
    vec<C> p(n), q(n);
    for (u32 i = 0, j; i < n; ++i) {
      j = (n - i) & (n - 1);
      C da = (a[i] + std::conj(a[j])) * C(.5, 0), db = (a[i] - std::conj(a[j])) * C(0, -.5), dc = (b[i] + std::conj(b[j])) * C(.5, 0), dd = (b[i] - std::conj(b[j])) * C(.5, 0);
      p[j] = da * dc + da * dd;
      q[j] = db * dc + db * dd;
    }
    a = p;
    b = q;
  }
  fft.dif(a);
  fft.dif(b);
  for (u32 i = 0; i < ans_size; ++i) {
    i64 da = (i64)(a[i].real() / (FP)n + .5) % mint::smod(), db = (i64)(a[i].imag() / (FP)n + .5) % mint::smod(), dc = (i64)(b[i].real() / (FP)n + .5) % mint::smod(), dd = (i64)(b[i].imag() / (FP)n + .5) % mint::smod();
    ans[i] = da + ((db + dc) << OFS) % mint::smod() + (dd << (OFS * 2)) % mint::smod();
  }
  return ans;
}

}  // namespace tifa_libs::math


#line 1 "src/code/poly/poly.hpp"



#line 5 "src/code/poly/poly.hpp"

namespace tifa_libs::math {

// clang-format off
enum ccore_t { ct_FFT, ct_3NTT, ct_NTT, ct_CNTT };
// clang-format on

template <class mint, class ccore>
requires requires(ccore cc, vec<mint> l, vec<mint> const &r, u32 sz) {
  { ccore::ct_cat } -> std::same_as<ccore_t const &>;
  cc.conv(l, r);
  cc.conv(l, r, sz);
}
class poly {
  vec<mint> d;

 public:
  using value_type = mint;
  using data_type = vec<value_type>;
  using ccore_type = ccore;
  static inline ccore_type conv_core;

  explicit constexpr poly(u32 sz = 1, value_type const &val = value_type{}) : d(sz, val) {}
  constexpr poly(typename data_type::const_iterator begin, typename data_type::const_iterator end) : d(begin, end) {}
  constexpr poly(std::initializer_list<value_type> v) : d(v) {}
  template <class T>
  explicit constexpr poly(vec<T> const &v) : d(v) {}

  friend constexpr std::istream &operator>>(std::istream &is, poly &poly) {
    for (auto &val : poly.d) is >> val;
    return is;
  }
  friend constexpr std::ostream &operator<<(std::ostream &os, poly const &poly) {
    if (!poly.size()) return os;
    for (u32 i = 1; i < poly.size(); ++i) os << poly[i - 1] << ' ';
    return os << poly.d.back();
  }

  constexpr u32 size() const { return (u32)d.size(); }
  constexpr bool empty() const {
    for (auto &&i : d)
      if (i != 0) return 0;
    return 1;
  }
  constexpr data_type &data() { return d; }
  constexpr data_type const &data() const { return d; }

  constexpr value_type &operator[](u32 x) { return d[x]; }
  constexpr value_type const &operator[](u32 x) const { return d[x]; }
  constexpr value_type operator()(value_type x) const {
    value_type ans = 0;
    for (u32 i = size() - 1; ~i; --i) ans = ans * x + d[i];
    return ans;
  }

  template <class F>
  requires requires(F f, u32 idx, mint &val) {
    f(idx, val);
  }
  constexpr void apply_range(u32 l, u32 r, F &&f) {
    assert(l < r && r <= size());
    for (u32 i = l; i < r; ++i) f(i, d[i]);
  }
  template <class F>
  constexpr void apply(F &&f) { apply_range(0, size(), std::forward<F>(f)); }
  constexpr void resize(u32 size) { d.resize(size); }
  constexpr poly pre(u32 size) const {
    poly _ = *this;
    _.resize(size);
    return _;
  }
  constexpr void strip() {
    auto it = std::find_if(d.rbegin(), d.rend(), [](auto const &x) { return x != 0; });
    d.resize(usz(d.rend() - it));
    if (d.empty()) d.push_back(value_type(0));
  }
  friend poly stripped(poly p) {
    p.strip();
    return p;
  }
  constexpr void reverse(u32 n = 0) { std::reverse(d.begin(), d.begin() + (n ? n : size())); }
  constexpr void conv(poly const &r, u32 ans_size = 0) { conv_core.conv(d, r.d, ans_size); }

  constexpr poly operator-() const {
    poly ret = *this;
    ret.apply([](u32, auto &v) { v = -v; });
    return ret;
  }

  friend constexpr poly operator+(poly p, value_type c) {
    p[0] += c;
    return p;
  }
  friend constexpr poly operator+(value_type c, poly const &p) { return p + c; }
  friend constexpr poly operator-(poly p, value_type c) {
    p[0] -= c;
    return p;
  }
  friend constexpr poly operator-(value_type c, poly const &p) { return p - c; }

  constexpr poly &operator*=(value_type c) {
    apply([&c](u32, auto &v) { v *= c; });
    return *this;
  }
  friend constexpr poly operator*(poly p, value_type c) { return p *= c; }
  friend constexpr poly operator*(value_type c, poly p) { return p *= c; }

  constexpr poly &operator+=(poly const &r) {
    if (!r.size()) return *this;
    resize(std::max(size(), r.size()));
    apply_range(0, r.size(), [&r](u32 i, auto &v) { v += r[i]; });
    return *this;
  }
  friend constexpr poly operator+(poly l, poly const &r) { return l += r; }

  constexpr poly &operator-=(poly const &r) {
    if (!r.size()) return *this;
    resize(std::max(size(), r.size()));
    apply_range(0, r.size(), [&r](u32 i, auto &v) { v -= r[i]; });
    return *this;
  }
  friend constexpr poly operator-(poly l, poly const &r) { return l -= r; }

  constexpr poly &operator*=(poly const &r) {
    if (!r.size()) {
      resize(1);
      d[0] = 0;
      return *this;
    }
    conv(r);
    return *this;
  }
  friend constexpr poly operator*(poly l, poly const &r) { return l *= r; }

  constexpr auto operator<=>(poly const &r) const { return stripped(*this).d <=> stripped(r).d; }
  constexpr bool operator==(poly const &r) const { return stripped(*this).d == stripped(r).d; }
};

}  // namespace tifa_libs::math


#line 6 "src/code/poly/polymtt.hpp"

namespace tifa_libs::math {
namespace polymtt_impl_ {
template <class FP = f64>
struct cconv_mtt : public FFT<FP> {
  static constexpr auto ct_cat = ct_FFT;
  template <class mint>
  constexpr void conv(vec<mint>& l, vec<mint> const& r, u32 sz = 0) { l = conv_mtt(*this, l, r, sz); }
};
}  // namespace polymtt_impl_

template <class mint, class FP = f64>
using polymtt = poly<mint, polymtt_impl_::cconv_mtt<FP>>;

}  // namespace tifa_libs::math


Back to top page