This documentation is automatically generated by online-judge-tools/verification-helper
#include "math/power_tower.hpp"
Solves ${ a_0 }^{ { a_1 }^{ { \cdots }^{ a_{n-1} } } }\pmod m$ for positive integers $a_i$ and $m$.
Algorithm implementation is based on maspy’s code.
template <typename T, typename Promote = unsigned long long>
T power_tower(const std::vector<T> &a, T m);
It returns ${ a_0 }^{ { a_1 }^{ { \cdots }^{ a_{ n-1 } } } }\pmod m$.
Providing first $\min(2\lfloor\lg{m}\rfloor, n)$ elements of $a$ is sufficient to calculate ${ a_0 }^{ { a_1 }^{ {\cdots }^{ a_{n-1} } } }\pmod m$. In other words ${a_0}^{ {a_1}^{ {\cdots}^{a_{n-1} } } } \equiv {a_0}^{ {a_1}^{ {\cdots}^{a_{k-1} } } } \pmod m$ where $k = \min(2\lfloor\lg{m}\rfloor, n)$. Notice that ${a_0}^{ {a_1}^{ {\cdots}^{a_{n-1} } } } \equiv {a_0}^{ {a_1}^{ {\cdots}^{a_{x-1} } } } \pmod m$ is true for any integer $x$ such that $\varphi^x(m) = 1$.
T
is an integral type such as int
, unsigned int
, long long
, and unsigned long long
. Promote
is an integral type that can express values between $0$ and $(2m-1)^2$, inclusive.
There are various methods to calculate Euler’s phi function. Miller-Rabin and Pollard’s rho and precalculated array are possible alternatives. It affects the complexity of power_tower
.
#ifndef POWER_TOWER_HPP
#define POWER_TOWER_HPP
#include <cassert>
#include <limits>
#include <type_traits>
#include <vector>
template <typename T> T euler_phi(T n) {
T phi = n;
for (T i = 2; i * i <= n; i++) {
if (n % i == 0) {
phi -= phi / i;
while (n % i == 0) {
n /= i;
}
}
}
if (n != 1) {
phi -= phi / n;
}
return phi;
}
template <typename T, typename Promote = unsigned long long>
T power_tower(const std::vector<T> &a, T m) {
static_assert(std::is_integral_v<T>);
assert(m > 0);
std::vector<unsigned long long> mod_chain{
static_cast<unsigned long long>(m)};
while (mod_chain.back() > 1) {
const auto phi = euler_phi(mod_chain.back());
mod_chain.push_back(phi);
}
const auto f = [](Promote x, Promote n, Promote mod) {
Promote r = 1;
if (x > mod) {
x = x % mod + mod;
}
while (n) {
if (n & 1) {
r *= x;
if (r > mod) {
r = r % mod + mod;
}
}
x *= x;
if (x > mod) {
x = x % mod + mod;
}
n >>= 1;
}
return r;
};
Promote r = 1;
const auto k = static_cast<int>(std::min(a.size(), mod_chain.size()));
for (auto i = k - 1; i >= 0; --i) {
assert(a[i] > 0);
r = f(static_cast<Promote>(a[i]), r, mod_chain[i]);
}
return static_cast<T>(r % static_cast<Promote>(m));
}
#endif // POWER_TOWER_HPP
#line 1 "math/power_tower.hpp"
#include <cassert>
#include <limits>
#include <type_traits>
#include <vector>
template <typename T> T euler_phi(T n) {
T phi = n;
for (T i = 2; i * i <= n; i++) {
if (n % i == 0) {
phi -= phi / i;
while (n % i == 0) {
n /= i;
}
}
}
if (n != 1) {
phi -= phi / n;
}
return phi;
}
template <typename T, typename Promote = unsigned long long>
T power_tower(const std::vector<T> &a, T m) {
static_assert(std::is_integral_v<T>);
assert(m > 0);
std::vector<unsigned long long> mod_chain{
static_cast<unsigned long long>(m)};
while (mod_chain.back() > 1) {
const auto phi = euler_phi(mod_chain.back());
mod_chain.push_back(phi);
}
const auto f = [](Promote x, Promote n, Promote mod) {
Promote r = 1;
if (x > mod) {
x = x % mod + mod;
}
while (n) {
if (n & 1) {
r *= x;
if (r > mod) {
r = r % mod + mod;
}
}
x *= x;
if (x > mod) {
x = x % mod + mod;
}
n >>= 1;
}
return r;
};
Promote r = 1;
const auto k = static_cast<int>(std::min(a.size(), mod_chain.size()));
for (auto i = k - 1; i >= 0; --i) {
assert(a[i] > 0);
r = f(static_cast<Promote>(a[i]), r, mod_chain[i]);
}
return static_cast<T>(r % static_cast<Promote>(m));
}