#line 1 "verify/graph/graph_counting.test.cpp"
#define PROBLEM "https://judge.yosupo.jp/problem/aplusb"
#include <algorithm>
#include <cassert>
#line 1 "utilities/fast_io.hpp"
#line 5 "utilities/fast_io.hpp"
#include <array>
#include <cerrno>
#include <charconv>
#include <cstddef>
#include <cstdio>
#include <cstdlib>
#include <cstdint>
#include <cstring>
#include <iterator>
#include <string>
#include <sys/stat.h>
#include <type_traits>
#include <utility>
#include <unistd.h>
#include <vector>
namespace m1une {
namespace utilities {
struct FastOutput;
namespace internal {
// Shared with the convenience helpers in template.hpp.
inline FastOutput* standard_output_instance = nullptr;
// Detect std::begin(x), std::end(x).
template <class T, class = void>
struct is_range : std::false_type {};
template <class T>
struct is_range<T, std::void_t<
decltype(std::begin(std::declval<T&>())),
decltype(std::end(std::declval<T&>()))
>> : std::true_type {};
template <class T>
inline constexpr bool is_range_v = is_range<T>::value;
template <class T>
using range_reference_t = decltype(*std::begin(std::declval<T&>()));
template <class T>
using range_value_t = std::remove_cv_t<std::remove_reference_t<range_reference_t<T>>>;
template <class T, class = void>
struct range_stored_value {
using type = range_value_t<T>;
};
template <class T>
struct range_stored_value<T, std::void_t<typename std::remove_cv_t<std::remove_reference_t<T>>::value_type>> {
using type = typename std::remove_cv_t<std::remove_reference_t<T>>::value_type;
};
template <class T>
using range_stored_value_t = typename range_stored_value<T>::type;
// Treat strings and C strings as scalar output objects, not as ranges.
template <class T>
struct is_char_array : std::false_type {};
template <class T, std::size_t N>
struct is_char_array<T[N]>
: std::bool_constant<std::is_same_v<std::remove_cv_t<T>, char>> {};
template <class T>
struct is_string_like
: std::bool_constant<
std::is_same_v<std::decay_t<T>, std::string>
|| std::is_same_v<std::decay_t<T>, const char*>
|| std::is_same_v<std::decay_t<T>, char*>
|| is_char_array<std::remove_reference_t<T>>::value
> {};
template <class T>
inline constexpr bool is_string_like_v = is_string_like<T>::value;
// ModInt-like type: x.val() is printable, and x can be assigned from long long.
template <class T, class = void>
struct has_val_method : std::false_type {};
template <class T>
struct has_val_method<T, std::void_t<decltype(std::declval<const T&>().val())>>
: std::true_type {};
template <class T>
inline constexpr bool has_val_method_v = has_val_method<T>::value;
template <class T, class = void>
struct has_static_mod_raw : std::false_type {};
template <class T>
struct has_static_mod_raw<
T, std::void_t<decltype(T::mod()), decltype(T::raw(std::declval<uint32_t>()))>>
: std::true_type {};
template <class T>
inline constexpr bool has_static_mod_raw_v = has_static_mod_raw<T>::value;
// libstdc++ before GCC 16 does not classify __int128 as an integral type in
// strict ISO modes such as -std=c++23. Keep the fast-I/O interface independent
// of that implementation detail.
template <class T>
inline constexpr bool is_integral_v =
std::is_integral_v<T>
|| std::is_same_v<std::remove_cv_t<T>, __int128_t>
|| std::is_same_v<std::remove_cv_t<T>, __uint128_t>;
template <class T>
inline constexpr bool is_signed_v =
std::is_signed_v<T>
|| std::is_same_v<std::remove_cv_t<T>, __int128_t>;
template <class T>
struct make_unsigned {
using type = std::make_unsigned_t<T>;
};
template <>
struct make_unsigned<__int128_t> {
using type = __uint128_t;
};
template <>
struct make_unsigned<__uint128_t> {
using type = __uint128_t;
};
template <class T>
using make_unsigned_t = typename make_unsigned<std::remove_cv_t<T>>::type;
} // namespace internal
struct FastInput {
static constexpr int buffer_size = 1 << 20;
private:
std::FILE* _stream;
char _buffer[buffer_size];
int _position;
int _length;
int _file_descriptor;
bool _streaming;
bool refill() {
_position = 0;
if (_streaming) {
ssize_t length;
do {
length = ::read(_file_descriptor, _buffer, buffer_size);
} while (length < 0 && errno == EINTR);
if (length <= 0) {
_length = 0;
return false;
}
_length = int(length);
} else {
_length = int(std::fread(_buffer, 1, buffer_size, _stream));
}
return _length != 0;
}
template <class T>
bool read_integer_from_stream(T& value) {
if (!skip_spaces()) return false;
int c = read_char_raw();
bool negative = false;
if (c == '-') {
negative = true;
c = read_char_raw();
}
if constexpr (internal::is_signed_v<T>) {
T result = 0;
while ('0' <= c && c <= '9') {
result = negative ? result * 10 - (c - '0')
: result * 10 + (c - '0');
c = read_char_raw();
}
value = result;
} else {
T result = 0;
while ('0' <= c && c <= '9') {
result = result * 10 + T(c - '0');
c = read_char_raw();
}
value = negative ? T(0) - result : result;
}
return true;
}
bool prepare_number() {
if (_length - _position >= 64) return true;
const int remaining = _length - _position;
if (remaining > 0) std::memmove(_buffer, _buffer + _position, remaining);
const int added = int(std::fread(_buffer + remaining, 1, buffer_size - remaining, _stream));
_position = 0;
_length = remaining + added;
if (_length < buffer_size) _buffer[_length] = '\0';
return _length != 0;
}
public:
explicit FastInput(std::FILE* stream = stdin)
: _stream(stream),
_position(0),
_length(0),
_file_descriptor(::fileno(stream)),
_streaming([&] {
struct stat status;
return _file_descriptor >= 0
&& ::fstat(_file_descriptor, &status) == 0
&& !S_ISREG(status.st_mode);
}()) {}
FastInput(const FastInput&) = delete;
FastInput& operator=(const FastInput&) = delete;
int read_char_raw() {
if (_position == _length && !refill()) return EOF;
return _buffer[_position++];
}
bool skip_spaces() {
int c = read_char_raw();
while (c != EOF && c <= ' ') c = read_char_raw();
if (c == EOF) return false;
--_position;
return true;
}
bool read(char& value) {
if (!skip_spaces()) return false;
value = char(read_char_raw());
return true;
}
bool read(std::string& value) {
if (!skip_spaces()) return false;
value.clear();
while (true) {
const int begin = _position;
while (_position < _length &&
static_cast<unsigned char>(_buffer[_position]) > ' ') {
++_position;
}
value.append(_buffer + begin, _position - begin);
if (_position < _length) {
++_position;
return true;
}
if (!refill()) return true;
}
}
bool read(bool& value) {
int x;
if (!read(x)) return false;
value = x != 0;
return true;
}
template <class T>
std::enable_if_t<
internal::is_integral_v<T>
&& !std::is_same_v<std::remove_cv_t<T>, bool>
&& !std::is_same_v<std::remove_cv_t<T>, char>,
bool
>
read(T& value) {
if (_streaming) return read_integer_from_stream(value);
if (!prepare_number()) return false;
int c = static_cast<unsigned char>(_buffer[_position++]);
while (c <= ' ') c = static_cast<unsigned char>(_buffer[_position++]);
bool negative = false;
if (c == '-') {
negative = true;
c = static_cast<unsigned char>(_buffer[_position++]);
}
if constexpr (internal::is_signed_v<T>) {
T result = 0;
while ('0' <= c && c <= '9') {
const int first = c - '0';
const int second = static_cast<unsigned char>(_buffer[_position]) - '0';
if (0 <= second && second <= 9) {
result = negative ? result * 100 - (first * 10 + second)
: result * 100 + (first * 10 + second);
++_position;
} else {
result = negative ? result * 10 - first : result * 10 + first;
}
c = static_cast<unsigned char>(_buffer[_position++]);
}
value = result;
} else {
T result = 0;
while ('0' <= c && c <= '9') {
const unsigned first = unsigned(c - '0');
const int second = static_cast<unsigned char>(_buffer[_position]) - '0';
if (0 <= second && second <= 9) {
result = result * 100 + T(first * 10 + unsigned(second));
++_position;
} else {
result = result * 10 + T(first);
}
c = static_cast<unsigned char>(_buffer[_position++]);
}
value = negative ? T(0) - result : result;
}
if (_position > _length) _position = _length;
return true;
}
template <class T>
std::enable_if_t<std::is_floating_point_v<T>, bool>
read(T& value) {
if (!skip_spaces()) return false;
int c = read_char_raw();
bool negative = false;
if (c == '-' || c == '+') {
negative = c == '-';
c = read_char_raw();
}
long double result = 0;
while ('0' <= c && c <= '9') {
result = result * 10 + (c - '0');
c = read_char_raw();
}
if (c == '.') {
long double place = 0.1L;
c = read_char_raw();
while ('0' <= c && c <= '9') {
result += (c - '0') * place;
place *= 0.1L;
c = read_char_raw();
}
}
if (c == 'e' || c == 'E') {
c = read_char_raw();
bool exponent_negative = false;
if (c == '-' || c == '+') {
exponent_negative = c == '-';
c = read_char_raw();
}
int exponent = 0;
while ('0' <= c && c <= '9') {
exponent = exponent * 10 + (c - '0');
c = read_char_raw();
}
long double scale = 1;
long double power = 10;
while (exponent > 0) {
if (exponent & 1) scale *= power;
power *= power;
exponent >>= 1;
}
result = exponent_negative ? result / scale : result * scale;
}
value = static_cast<T>(negative ? -result : result);
return true;
}
template <class T>
std::enable_if_t<
internal::has_val_method_v<T>
&& !internal::is_integral_v<T>
&& !internal::is_range_v<T>,
bool
>
read(T& value) {
long long x;
if (!read(x)) return false;
if constexpr (internal::has_static_mod_raw_v<T>) {
if (x >= 0 && uint64_t(x) < uint64_t(T::mod())) {
value = T::raw(uint32_t(x));
} else {
value = T(x);
}
} else {
value = T(x);
}
return true;
}
template <class First, class Second>
bool read(std::pair<First, Second>& value) {
if (!read(value.first)) return false;
return read(value.second);
}
template <class Range>
std::enable_if_t<
internal::is_range_v<Range>
&& !internal::is_string_like_v<Range>,
bool
>
read(Range& range) {
using StoredValue = internal::range_stored_value_t<Range>;
constexpr bool nested = internal::is_range_v<StoredValue>
&& !internal::is_string_like_v<StoredValue>;
for (auto&& value : range) {
if constexpr (std::is_same_v<StoredValue, bool> && !nested) {
bool x;
if (!read(x)) return false;
value = x;
} else {
if (!read(value)) return false;
}
}
return true;
}
template <class First, class Second, class... Rest>
bool read(First& first, Second& second, Rest&... rest) {
if (!read(first)) return false;
return read(second, rest...);
}
template <class T>
FastInput& operator>>(T& value) {
if (!read(value)) std::abort();
return *this;
}
};
struct FastOutput {
static constexpr int buffer_size = 1 << 20;
private:
inline static const auto digit_quads = [] {
std::array<char, 40000> result{};
for (int i = 0; i < 10000; i++) {
int value = i;
for (int j = 3; j >= 0; j--) {
result[4 * i + j] = char('0' + value % 10);
value /= 10;
}
}
return result;
}();
std::FILE* _stream;
char _buffer[buffer_size];
int _position;
int _precision;
std::chars_format _float_format;
char _range_separator;
std::string* _capture = nullptr;
template <class T>
std::string format_cell(const T& value) {
std::string result;
struct CaptureGuard {
std::string*& target;
std::string* previous;
~CaptureGuard() { target = previous; }
} guard{_capture, _capture};
_capture = &result;
write(value);
return result;
}
template <class Matrix>
void write_aligned_matrix(const Matrix& matrix) {
std::vector<std::vector<std::string>> rows;
std::vector<std::size_t> widths;
for (const auto& row : matrix) {
auto& cells = rows.emplace_back();
std::size_t column = 0;
for (const auto& value : row) {
cells.push_back(format_cell(value));
if (column == widths.size()) widths.push_back(0);
widths[column] = std::max(widths[column], cells.back().size());
++column;
}
}
bool first = true;
for (const auto& row : rows) {
if (!first) write_char('\n');
first = false;
for (std::size_t column = 0; column < row.size(); ++column) {
if (column != 0) write_char(_range_separator);
for (std::size_t padding = row[column].size();
padding < widths[column]; ++padding) {
write_char(' ');
}
write(row[column]);
}
}
}
public:
explicit FastOutput(std::FILE* stream = stdout)
: _stream(stream),
_position(0),
_precision(6),
_float_format(std::chars_format::general),
_range_separator(' ') {
if (_stream == stdout
&& internal::standard_output_instance == nullptr) {
internal::standard_output_instance = this;
}
}
FastOutput(const FastOutput&) = delete;
FastOutput& operator=(const FastOutput&) = delete;
~FastOutput() {
flush();
if (internal::standard_output_instance == this) {
internal::standard_output_instance = nullptr;
}
}
void flush() {
if (_position != 0) {
std::fwrite(_buffer, 1, _position, _stream);
_position = 0;
}
std::fflush(_stream);
}
void write_char(char c) {
if (_capture != nullptr) {
_capture->push_back(c);
return;
}
if (_position == buffer_size) flush();
_buffer[_position++] = c;
}
void write(const char* s) {
while (*s != '\0') write_char(*s++);
}
void write(const std::string& s) {
if (_capture != nullptr) {
_capture->append(s);
return;
}
std::size_t position = 0;
while (position < s.size()) {
if (_position == buffer_size) flush();
const std::size_t copied =
std::min<std::size_t>(buffer_size - _position, s.size() - position);
std::memcpy(_buffer + _position, s.data() + position, copied);
_position += int(copied);
position += copied;
}
}
void write(char c) {
write_char(c);
}
void write(bool value) {
write_char(value ? '1' : '0');
}
template <class T>
std::enable_if_t<std::is_floating_point_v<T>>
write(T value) {
char digits[128];
auto [end, error] = std::to_chars(
digits,
digits + sizeof(digits),
value,
_float_format,
_precision
);
if (error != std::errc()) std::abort();
for (const char* pointer = digits; pointer != end; pointer++) {
write_char(*pointer);
}
}
template <class T>
std::enable_if_t<
internal::is_integral_v<T>
&& !std::is_same_v<std::remove_cv_t<T>, bool>
&& !std::is_same_v<std::remove_cv_t<T>, char>
>
write(T value) {
using Raw = std::remove_cv_t<T>;
using Unsigned = internal::make_unsigned_t<Raw>;
Unsigned magnitude;
if constexpr (internal::is_signed_v<Raw>) {
if (value < 0) {
write_char('-');
magnitude = Unsigned(0) - Unsigned(value);
} else {
magnitude = Unsigned(value);
}
} else {
magnitude = value;
}
if (magnitude == 0) {
write_char('0');
return;
}
unsigned chunks[16];
int count = 0;
while (magnitude >= 10000) {
const Unsigned quotient = magnitude / 10000;
chunks[count++] = unsigned(magnitude - quotient * 10000);
magnitude = quotient;
}
if (_capture == nullptr && _position > buffer_size - 64) flush();
char captured[64];
char* const begin = _capture != nullptr ? captured : _buffer + _position;
char* destination = begin;
const unsigned leading = unsigned(magnitude);
const char* first = digit_quads.data() + 4 * leading;
int skip = leading < 10 ? 3 : leading < 100 ? 2 : leading < 1000 ? 1 : 0;
for (; skip < 4; skip++) *destination++ = first[skip];
while (count--) {
const char* digits = digit_quads.data() + 4 * chunks[count];
std::memcpy(destination, digits, 4);
destination += 4;
}
if (_capture != nullptr) {
_capture->append(begin, destination - begin);
} else {
_position += int(destination - begin);
}
}
template <class T>
std::enable_if_t<
internal::has_val_method_v<T>
&& !internal::is_integral_v<T>
&& !internal::is_range_v<T>
>
write(const T& value) {
write(value.val());
}
template <class First, class Second>
void write(const std::pair<First, Second>& value) {
write(value.first);
write_char(' ');
write(value.second);
}
template <class Range>
std::enable_if_t<
internal::is_range_v<Range>
&& !internal::is_string_like_v<Range>
>
write(const Range& range) {
using StoredValue = internal::range_stored_value_t<const Range>;
constexpr bool nested = internal::is_range_v<StoredValue>
&& !internal::is_string_like_v<StoredValue>;
bool first = true;
for (const auto& value : range) {
if (!first) write_char(nested ? '\n' : _range_separator);
first = false;
if constexpr (std::is_same_v<StoredValue, bool> && !nested) {
write(static_cast<bool>(value));
} else {
write(value);
}
}
}
template <class First, class... Rest>
void print(const First& first, const Rest&... rest) {
write(first);
((write_char(' '), write(rest)), ...);
}
void println() {
write_char('\n');
}
void set_precision(int precision) {
_precision = precision;
}
void set_fixed(int precision = 6) {
_float_format = std::chars_format::fixed;
_precision = precision;
}
void set_general(int precision = 6) {
_float_format = std::chars_format::general;
_precision = precision;
}
void set_range_separator(char separator) {
_range_separator = separator;
}
template <class Matrix>
void write_aligned(const Matrix& matrix) {
using Row = internal::range_stored_value_t<const Matrix>;
using Cell = internal::range_stored_value_t<const Row>;
static_assert(internal::is_range_v<Row> && !internal::is_string_like_v<Row>,
"write_aligned requires a two-dimensional range");
static_assert(!internal::is_range_v<Cell> || internal::is_string_like_v<Cell>,
"write_aligned requires scalar cells");
write_aligned_matrix(matrix);
}
template <class Matrix>
void println_aligned(const Matrix& matrix) {
write_aligned(matrix);
write_char('\n');
}
template <class... Args>
void println(const Args&... args) {
print(args...);
write_char('\n');
}
template <class T>
FastOutput& operator<<(const T& value) {
write(value);
return *this;
}
};
} // namespace utilities
} // namespace m1une
#line 6 "verify/graph/graph_counting.test.cpp"
#include <numeric>
#line 8 "verify/graph/graph_counting.test.cpp"
#line 1 "graph/counting.hpp"
#line 6 "graph/counting.hpp"
#include <optional>
#line 9 "graph/counting.hpp"
#line 1 "math/fps/convolution.hpp"
#line 9 "math/fps/convolution.hpp"
#include <new>
#line 13 "math/fps/convolution.hpp"
#if defined(__GNUC__) && !defined(__clang__) && \
(defined(__x86_64__) || defined(__i386__)) && \
!defined(M1UNE_FPS_DISABLE_X86_SIMD)
#include <immintrin.h>
#define M1UNE_FPS_HAS_X86_SIMD 1
#pragma GCC push_options
#pragma GCC target("avx2,bmi")
#endif
#line 1 "math/fps/internal/ntt998_faster.hpp"
#ifdef M1UNE_FPS_HAS_X86_SIMD
#line 9 "math/fps/internal/ntt998_faster.hpp"
#include <immintrin.h>
namespace m1une {
namespace fps {
namespace internal {
namespace fast998_v2 {
// Fixed-modulus AVX2 transform with an in-register degree-8 residue product.
using u32=unsigned;
using u64=unsigned long long;
using idt=std::size_t;
using I256=__m256i;
inline void store256(void*p,I256 x){
_mm256_store_si256((I256*)p,x);
}
inline I256 load256(const void*p){
return _mm256_load_si256((const I256*)p);
}
constexpr u32 shrk(u32 x,u32 M){
return std::min(x,x-M);
}
constexpr u32 dilt(u32 x,u32 M){
return std::min(x,x+M);
}
constexpr u32 reduce(u64 x,u32 niv,u32 M){
return (x+u64(u32(x)*niv)*M)>>32;
}
constexpr u32 mul(u32 x,u32 y,u32 niv,u32 M){
return reduce(u64(x)*y,niv,M);
}
constexpr u32 mul_s(u32 x,u32 y,u32 niv,u32 M){
return shrk(reduce(u64(x)*y,niv,M),M);
}
constexpr u32 qpw(u32 a,u32 b,u32 niv,u32 M,u32 r){
for(;b;b>>=1,a=mul(a,a,niv,M)){
if(b&1){
r=mul(r,a,niv,M);
}
}
return r;
}
constexpr u32 qpw_s(u32 a,u32 b,u32 niv,u32 M,u32 r){
return shrk(qpw(a,b,niv,M,r),M);
}
inline I256 shrk32(I256 x,I256 M){
return _mm256_min_epu32(x,_mm256_sub_epi32(x,M));
}
inline I256 dilt32(I256 x,I256 M){
return _mm256_min_epu32(x,_mm256_add_epi32(x,M));
}
inline I256 Ladd32(I256 x,I256 y,I256){
return _mm256_add_epi32(x,y);
}
inline I256 Lsub32(I256 x,I256 y,I256 M){
return _mm256_add_epi32(_mm256_sub_epi32(x,y),M);
}
inline I256 add32(I256 x,I256 y,I256 M){
return shrk32(_mm256_add_epi32(x,y),M);
}
inline I256 sub32(I256 x,I256 y,I256 M){
return dilt32(_mm256_sub_epi32(x,y),M);
}
template<int msk>inline I256 neg32_m(I256 x,I256 M){
return _mm256_blend_epi32(x,_mm256_sub_epi32(M,x),msk);
}
inline I256 reduce(I256 a,I256 b,I256 niv,I256 M){
I256 c=_mm256_mul_epu32(a,niv),d=_mm256_mul_epu32(b,niv);
c=_mm256_mul_epu32(c,M),d=_mm256_mul_epu32(d,M);
return _mm256_blend_epi32(_mm256_srli_epi64(_mm256_add_epi64(a,c),32),_mm256_add_epi64(b,d),0xaa);
}
inline I256 mul(I256 a,I256 b,I256 niv,I256 M){
return reduce(_mm256_mul_epu32(a,b),_mm256_mul_epu32(_mm256_srli_epi64(a,32),_mm256_srli_epi64(b,32)),niv,M);
}
inline I256 mul_s(I256 a,I256 b,I256 niv,I256 M){
return shrk32(mul(a,b,niv,M),M);
}
inline I256 mul_bsm(I256 a,I256 b,I256 niv,I256 M){
return reduce(_mm256_mul_epu32(a,b),_mm256_mul_epu32(_mm256_srli_epi64(a,32),b),niv,M);
}
inline I256 mul_bsmfxd(I256 a,I256 b,I256 bniv,I256 M){
I256 cc=_mm256_mul_epu32(a,bniv),dd=_mm256_mul_epu32(_mm256_srli_epi64(a,32),bniv);
I256 c=_mm256_mul_epu32(a,b),d=_mm256_mul_epu32(_mm256_srli_epi64(a,32),b);
cc=_mm256_mul_epu32(cc,M),dd=_mm256_mul_epu32(dd,M);
return _mm256_blend_epi32(_mm256_srli_epi64(_mm256_add_epi64(c,cc),32),_mm256_add_epi64(d,dd),0xaa);
}
inline I256 mul_bfxd(I256 a,I256 b,I256 bniv,I256 M){
I256 cc=_mm256_mul_epu32(a,bniv),dd=_mm256_mul_epu32(_mm256_srli_epi64(a,32),_mm256_srli_epi64(bniv,32));
I256 c=_mm256_mul_epu32(a,b),d=_mm256_mul_epu32(_mm256_srli_epi64(a,32),_mm256_srli_epi64(b,32));
cc=_mm256_mul_epu32(cc,M),dd=_mm256_mul_epu32(dd,M);
return _mm256_blend_epi32(_mm256_srli_epi64(_mm256_add_epi64(c,cc),32),_mm256_add_epi64(d,dd),0xaa);
}
inline I256 mul_upd_rt(I256 a,I256 bu,I256 M){
I256 cc=_mm256_mul_epu32(a,bu),c=_mm256_mul_epu32(a,_mm256_srli_epi64(bu,32));
cc=_mm256_mul_epu32(cc,M);
return shrk32(_mm256_srli_epi64(_mm256_add_epi64(c,cc),32),M);
}
constexpr auto _mxlg=26,_lg_itth=6;
constexpr auto _itth=idt(1)<<_lg_itth;
static_assert(_lg_itth%2==0);
struct FNTT32_info{
u32 mod,mod2,niv,one,r2,r3,img,imgniv,RT1[_mxlg];
alignas(32) std::array<u32,8> rt3[_mxlg-2],rt3i[_mxlg-2],bwbr,bwb,bwbi,rt4[_mxlg-3],rt4niv[_mxlg-3],rt4i[_mxlg-3],rt4iniv[_mxlg-3],pr2,pr4,pr2niv,pr4niv,pr2i,pr2iniv,pr4i,pr4iniv;
constexpr FNTT32_info(const u32 m):mod(m),mod2(m*2),niv([&]{u32 n=2+m;for(int i=0;i<4;++i){n*=2+m*n;}return n;}()),one((-m)%m),r2((-u64(m))%m),r3(mul_s(r2,r2,niv,m)),img{},imgniv{},RT1{},rt3{},rt3i{},bwbr{},bwb{},bwbi{},rt4{},rt4niv{},rt4i{},rt4iniv{},pr2{},pr4{},pr2niv{},pr4niv{},pr2i{},pr2iniv{},pr4i{},pr4iniv{}{
const int k=__builtin_ctz(m-1);
u32 _g=mul(3,r2,niv,mod);
for(;;++_g){
if(qpw_s(_g,mod>>1,niv,mod,one)!=one){
break;
}
}
_g=qpw(_g,mod>>k,niv,mod,one);
u32 rt1[_mxlg-1],rt1i[_mxlg-1];
rt1[k-2]=_g,rt1i[k-2]=qpw(_g,mod-2,niv,mod,one);
for(int i=k-2;i>0;--i){
rt1[i-1]=mul(rt1[i],rt1[i],niv,mod);
rt1i[i-1]=mul(rt1i[i],rt1i[i],niv,mod);
}
RT1[k-1]=qpw_s(_g,3,niv,mod,one);
for(int i=k-1;i>0;--i){
RT1[i-1]=mul_s(RT1[i],RT1[i],niv,mod);
}
img=rt1[0],imgniv=img*niv;
bwbr={one,0,one,0,one};
bwb={rt1[1],0,rt1[0],0,mod-mul_s(rt1[0],rt1[1],niv,mod)};
bwbi={rt1i[1],0,rt1i[0],0,mul_s(rt1i[0],rt1i[1],niv,mod)};
u32 pr=one,pri=one;
for(int i=0;i<k-2;++i){
const u32 r=mul_s(pr,rt1[i+1],niv,mod),ri=mul_s(pri,rt1i[i+1],niv,mod);
const u32 r2=mul_s(r,r,niv,mod),r2i=mul_s(ri,ri,niv,mod);
const u32 r3=mul_s(r,r2,niv,mod),r3i=mul_s(ri,r2i,niv,mod);
rt3[i]={r*niv,r,r2*niv,r2,r3*niv,r3};
rt3i[i]={ri*niv,ri,r2i*niv,r2i,r3i*niv,r3i};
pr=mul(pr,rt1i[i+1],niv,mod),pri=mul(pri,rt1[i+1],niv,mod);
}
pr=one,pri=one;
for(int i=0;i<k-3;++i){
const u32 r=mul_s(pr,rt1[i+2],niv,mod),ri=mul_s(pri,rt1i[i+2],niv,mod);
rt4[i][0]=rt4i[i][0]=one;
for(int j=1;j<8;++j){
rt4[i][j]=mul_s(rt4[i][j-1],r,niv,mod);
rt4i[i][j]=mul_s(rt4i[i][j-1],ri,niv,mod);
}
for(int j=0;j<8;++j){
rt4niv[i][j]=rt4[i][j]*niv;
rt4iniv[i][j]=rt4i[i][j]*niv;
}
pr=mul(pr,rt1i[i+2],niv,mod),pri=mul(pri,rt1[i+2],niv,mod);
}
pr2={one,one,one,img,one,one,one,img};
pr4={one,one,one,one,one,rt1[1],img,mul_s(img,rt1[1],niv,mod)};
const u32 nr2=mod-r2,imgr2=mul_s(img,r2,niv,mod);
pr2i={nr2,nr2,nr2,imgr2,nr2,nr2,nr2,imgr2};
pr4i={one,one,one,one,one,rt1i[1],rt1i[0],mul_s(rt1i[0],rt1i[1],niv,mod)};
for(int j=0;j<8;++j){
pr2niv[j]=pr2[j]*niv,pr4niv[j]=pr4[j]*niv;
pr2iniv[j]=pr2i[j]*niv,pr4iniv[j]=pr4i[j]*niv;
}
}
};
inline void vector_dif(I256*const f,const idt n,const FNTT32_info*info){
alignas(32) std::array<u32,8> st_1[_mxlg>>1];
const I256 Mod=_mm256_set1_epi32(info->mod),Mod2=_mm256_set1_epi32(info->mod2),Niv=_mm256_set1_epi32(info->niv);
const I256 Img=_mm256_set1_epi32(info->img),ImgNiv=_mm256_set1_epi32(info->imgniv),id=_mm256_setr_epi32(0,2,0,4,0,2,0,4);
const int lgn=__builtin_ctzll(n);
std::fill(st_1,st_1+(lgn>>1),info->bwb);
const idt nn=n>>(lgn&1),m=std::min(n,_itth),mm=std::min(nn,_itth);
// I256 rr=_mm256_set1_epi32(info->one);
if(nn!=n){
for(idt i=0;i<nn;++i){
auto const p0=f+i,p1=f+nn+i;
const auto f0=load256(p0),f1=load256(p1);
const auto g0=add32(f0,f1,Mod2),g1=Lsub32(f0,f1,Mod2);
store256(p0,g0),store256(p1,g1);
}
}
for(idt L=nn>>2;L>0;L>>=2){
for(idt i=0;i<L;++i){
auto const p0=f+i,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f1=load256(p1),f3=load256(p3),f2=load256(p2),f0=load256(p0);
const auto g3=mul_bsmfxd(Lsub32(f1,f3,Mod2),Img,ImgNiv,Mod),g1=add32(f1,f3,Mod2);
const auto g0=add32(f0,f2,Mod2),g2=sub32(f0,f2,Mod2);
const auto h0=add32(g0,g1,Mod2),h1=Lsub32(g0,g1,Mod2);
const auto h2=Ladd32(g2,g3,Mod2),h3=Lsub32(g2,g3,Mod2);
store256(p0,h0),store256(p1,h1),store256(p2,h2),store256(p3,h3);
}
}
for(idt j=0;j<n;j+=m){
int t=((j==0)?std::min(_lg_itth,lgn):__builtin_ctzll(j))&-2,p=(t-2)>>1;
for(idt L=(idt(1)<<t)>>2;L>=_itth;L>>=2,t-=2,--p){
auto rt=load256(st_1+p);
const auto r1=_mm256_permutevar8x32_epi32(rt,id);
const auto r1Niv=_mm256_permutevar8x32_epi32(_mm256_mul_epu32(rt,Niv),id);
rt=mul_upd_rt(rt,load256(info->rt3+__builtin_ctzll(~j>>t)),Mod);
const auto r2=_mm256_shuffle_epi32(r1,_MM_PERM_BBBB),nr3=_mm256_shuffle_epi32(r1,_MM_PERM_DDDD);
const auto r2Niv=_mm256_shuffle_epi32(r1Niv,_MM_PERM_BBBB),nr3Niv=_mm256_shuffle_epi32(r1Niv,_MM_PERM_DDDD);
store256(st_1+p,rt);
for(idt i=0;i<L;++i){
auto const p0=f+i+j,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f1=load256(p1),f3=load256(p3),f2=load256(p2),f0=load256(p0);
const auto g1=mul_bsmfxd(f1,r1,r1Niv,Mod),ng3=mul_bsmfxd(f3,nr3,nr3Niv,Mod);
const auto g2=mul_bsmfxd(f2,r2,r2Niv,Mod),g0=shrk32(f0,Mod2);
const auto h3=mul_bsmfxd(Ladd32(g1,ng3,Mod2),Img,ImgNiv,Mod),h1=sub32(g1,ng3,Mod2);
const auto h0=add32(g0,g2,Mod2),h2=sub32(g0,g2,Mod2);
const auto u0=Ladd32(h0,h1,Mod2),u1=Lsub32(h0,h1,Mod2);
const auto u2=Ladd32(h2,h3,Mod2),u3=Lsub32(h2,h3,Mod2);
store256(p0,u0),store256(p1,u1),store256(p2,u2),store256(p3,u3);
}
}
I256*const g=f+j;
for(idt l=mm,L=mm>>2;L;l=L,L>>=2,t-=2,--p){
auto rt=load256(st_1+p);
for(idt i=(j==0?l:0),k=(j+i)>>t;i<m;i+=l,++k){
const auto r1=_mm256_permutevar8x32_epi32(rt,id);
const auto r2=_mm256_shuffle_epi32(r1,_MM_PERM_BBBB);
const auto nr3=_mm256_shuffle_epi32(r1,_MM_PERM_DDDD);
for(idt j=0;j<L;++j){
auto const p0=g+i+j,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f1=load256(p1),f3=load256(p3),f2=load256(p2),f0=load256(p0);
const auto g1=mul_bsm(f1,r1,Niv,Mod),ng3=mul_bsm(f3,nr3,Niv,Mod);
const auto g2=mul_bsm(f2,r2,Niv,Mod),g0=shrk32(f0,Mod2);
const auto h3=mul_bsmfxd(Ladd32(g1,ng3,Mod2),Img,ImgNiv,Mod),h1=sub32(g1,ng3,Mod2);
const auto h0=add32(g0,g2,Mod2),h2=sub32(g0,g2,Mod2);
const auto u0=Ladd32(h0,h1,Mod2),u1=Lsub32(h0,h1,Mod2);
const auto u2=Ladd32(h2,h3,Mod2),u3=Lsub32(h2,h3,Mod2);
store256(p0,u0),store256(p1,u1),store256(p2,u2),store256(p3,u3);
}
rt=mul_upd_rt(rt,load256(info->rt3+__builtin_ctzll(~k)),Mod);
}
store256(st_1+p,rt);
}
// const auto pr2=load256(&info->pr2),pr4=load256(&info->pr4);
// const auto pr2Niv=load256(&info->pr2niv),pr4Niv=load256(&info->pr4niv);
// for(idt i=j;i<j+m;++i){
// auto fi=load256(f+i);
// fi=mul(fi,rr,Niv,Mod);
// rr=shrk32(mul_bfxd(rr,load256(info->rt4+__builtin_ctzll(~i)),load256(info->rt4niv+__builtin_ctzll(~i)),Mod),Mod);
// fi=mul_bfxd(Ladd32(neg32_m<0xf0>(fi,Mod2),_mm256_permute2x128_si256(fi,fi,1),Mod2),pr4,pr4Niv,Mod);
// fi=mul_bfxd(Ladd32(neg32_m<0xcc>(fi,Mod2),_mm256_shuffle_epi32(fi,0x4e),Mod2),pr2,pr2Niv,Mod);
// fi=sub32(_mm256_shuffle_epi32(fi,0xb1),neg32_m<0x55>(fi,Mod2),Mod2);
// store256(f+i,fi);
// }
}
}
template<bool shrk=false>inline void vector_dit(I256*const f,idt n,const FNTT32_info*const info){
alignas(32) std::array<u32,8> st_1[_mxlg>>1];
const I256 Mod=_mm256_set1_epi32(info->mod),Mod2=_mm256_set1_epi32(info->mod2),Niv=_mm256_set1_epi32(info->niv);
const I256 Img=_mm256_set1_epi32(info->img),ImgNiv=_mm256_set1_epi32(info->imgniv),id=_mm256_setr_epi32(0,2,0,4,0,2,0,4);
const int lgn=__builtin_ctzll(n);
std::fill(st_1,st_1+(_lg_itth>>1),info->bwbr);
std::fill(st_1+(_lg_itth>>1),st_1+(_mxlg>>1),info->bwbi);
const idt nn=n>>(lgn&1),mm=std::min(nn,_itth);
// I256 rr=_mm256_set1_epi32((info->mod-1)>>(lgn+3));
for(idt j=0;j<n;j+=mm){
// const auto pr2=load256(&info->pr2i),pr4=load256(&info->pr4i);
// const auto pr2Niv=load256(&info->pr2iniv),pr4Niv=load256(&info->pr4iniv);
// for(idt i=j;i<j+mm;++i){
// auto fi=load256(f+i);
// const auto rt=rr;
// rr=shrk32(mul_bfxd(rr,load256(info->rt4i+__builtin_ctzll(~i)),load256(info->rt4iniv+__builtin_ctzll(~i)),Mod),Mod);
// fi=mul_bfxd(Ladd32(neg32_m<0xaa>(fi,Mod2),_mm256_shuffle_epi32(fi,0xb1),Mod2),pr2,pr2Niv,Mod);
// fi=mul_bfxd(Ladd32(neg32_m<0xcc>(fi,Mod2),_mm256_shuffle_epi32(fi,0x4e),Mod2),pr4,pr4Niv,Mod);
// fi=mul(Ladd32(neg32_m<0xf0>(fi,Mod2),_mm256_permute2x128_si256(fi,fi,1),Mod2),rt,Niv,Mod);
// store256(f+i,fi);
// }
I256*const g=f+j;
int t=2,p=0;
for(idt l=4,L=1;l<=mm;L=l,l<<=2,t+=2,++p){
auto rt=load256(st_1+p);
for(idt i=0,k=j>>t;i<mm;i+=l,++k){
const auto r1=_mm256_permutevar8x32_epi32(rt,id);
const auto r2=_mm256_shuffle_epi32(r1,_MM_PERM_BBBB);
const auto r3=_mm256_shuffle_epi32(r1,_MM_PERM_DDDD);
for(idt j=0;j<L;++j){
auto const p0=g+i+j,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f0=load256(p0),f1=load256(p1),f2=load256(p2),f3=load256(p3);
const auto g0=add32(f0,f1,Mod2),g1=sub32(f0,f1,Mod2);
const auto g2=add32(f2,f3,Mod2),g3=mul_bsmfxd(Lsub32(f3,f2,Mod2),Img,ImgNiv,Mod);
const auto h0=Ladd32(g0,g2,Mod2),h1=Ladd32(g1,g3,Mod2);
const auto h2=Lsub32(g0,g2,Mod2),h3=Lsub32(g1,g3,Mod2);
const auto u0=shrk32(h0,Mod2),u1=mul_bsm(h1,r1,Niv,Mod);
const auto u2=mul_bsm(h2,r2,Niv,Mod),u3=mul_bsm(h3,r3,Niv,Mod);
store256(p0,u0),store256(p1,u1),store256(p2,u2),store256(p3,u3);
}
rt=mul_upd_rt(rt,load256(info->rt3i+__builtin_ctzll(~k)),Mod);
}
store256(st_1+p,rt);
}
int tt=std::min(__builtin_ctzll(~(j>>_lg_itth))+_lg_itth,lgn);
for(idt L=_itth,l=L<<2;t<=tt;L=l,l<<=2,t+=2,++p){
if((j+_itth)==l){
if(shrk && l==n){
for(idt i=0;i<L;++i){
auto const p0=f+i,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f2=load256(p2),f3=load256(p3),f0=load256(p0),f1=load256(p1);
const auto g3=mul_bsmfxd(Lsub32(f3,f2,Mod2),Img,ImgNiv,Mod),g2=add32(f2,f3,Mod2);
const auto g0=add32(f0,f1,Mod2),g1=sub32(f0,f1,Mod2);
const auto h0=add32(g0,g2,Mod2),h1=add32(g1,g3,Mod2);
const auto h2=sub32(g0,g2,Mod2),h3=sub32(g1,g3,Mod2);
const auto u0=shrk32(h0,Mod),u1=shrk32(h1,Mod);
const auto u2=shrk32(h2,Mod),u3=shrk32(h3,Mod);
store256(p0,u0),store256(p1,u1),store256(p2,u2),store256(p3,u3);
}
}
else{
for(idt i=0;i<L;++i){
auto const p0=f+i,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f2=load256(p2),f3=load256(p3),f0=load256(p0),f1=load256(p1);
const auto g3=mul_bsmfxd(Lsub32(f3,f2,Mod2),Img,ImgNiv,Mod),g2=add32(f2,f3,Mod2);
const auto g0=add32(f0,f1,Mod2),g1=sub32(f0,f1,Mod2);
const auto h0=add32(g0,g2,Mod2),h1=add32(g1,g3,Mod2);
const auto h2=sub32(g0,g2,Mod2),h3=sub32(g1,g3,Mod2);
store256(p0,h0),store256(p1,h1),store256(p2,h2),store256(p3,h3);
}
}
}
else{
auto rt=load256(st_1+p);
const auto r1=_mm256_permutevar8x32_epi32(rt,id);
const auto r1Niv=_mm256_permutevar8x32_epi32(_mm256_mul_epu32(rt,Niv),id);
rt=mul_upd_rt(rt,load256(info->rt3i+__builtin_ctzll(~j>>t)),Mod);
const auto r2=_mm256_shuffle_epi32(r1,_MM_PERM_BBBB),r3=_mm256_shuffle_epi32(r1,_MM_PERM_DDDD);
const auto r2Niv=_mm256_shuffle_epi32(r1Niv,_MM_PERM_BBBB),r3Niv=_mm256_shuffle_epi32(r1Niv,_MM_PERM_DDDD);
store256(st_1+p,rt);
for(idt i=0;i<L;++i){
auto const p0=f+j+_itth-l+i,p1=p0+L,p2=p1+L,p3=p2+L;
const auto f0=load256(p0),f1=load256(p1),f2=load256(p2),f3=load256(p3);
const auto g0=add32(f0,f1,Mod2),g1=sub32(f0,f1,Mod2);
const auto g2=add32(f2,f3,Mod2),g3=mul_bsmfxd(Lsub32(f3,f2,Mod2),Img,ImgNiv,Mod);
const auto h0=Ladd32(g0,g2,Mod2),h1=Ladd32(g1,g3,Mod2);
const auto h2=Lsub32(g0,g2,Mod2),h3=Lsub32(g1,g3,Mod2);
const auto u0=shrk32(h0,Mod2),u1=mul_bsmfxd(h1,r1,r1Niv,Mod);
const auto u2=mul_bsmfxd(h2,r2,r2Niv,Mod),u3=mul_bsmfxd(h3,r3,r3Niv,Mod);
store256(p0,u0),store256(p1,u1),store256(p2,u2),store256(p3,u3);
}
}
}
}
if(shrk && nn==n && n<=_itth){
for(idt i=0;i<n;++i){
const auto f0=load256(f+i);
store256(f+i,shrk32(f0,Mod));
}
}
if(nn!=n){
for(idt i=0;i<nn;++i){
auto const p0=f+i,p1=f+nn+i;
const auto f0=load256(p0),f1=load256(p1);
const auto g0=add32(f0,f1,Mod2),g1=sub32(f0,f1,Mod2);
if constexpr(shrk){
const auto h0=shrk32(g0,Mod),h1=shrk32(g1,Mod);
store256(p0,h0),store256(p1,h1);
}
else{
store256(p0,g0),store256(p1,g1);
}
}
}
}
// Returns fx * f[0,8) * g[0,8) (mod x^8 - ww).
[[gnu::always_inline]] inline I256 convolve8(const I256*f,const I256*g,I256 ww,I256 fx,I256 Niv,I256 Mod,I256 Mod2){
const auto raa=load256(f),rbb=load256(g);
const auto taa=shrk32(raa,Mod2),bb=shrk32(mul_bsm(rbb,fx,Niv,Mod),Mod);
const auto aw=shrk32(mul_bsm(taa,ww,Niv,Mod),Mod);
const auto aa=shrk32(taa,Mod);
const auto awa=_mm256_permute2x128_si256(aa,aw,3);
const auto b0=_mm256_permute4x64_epi64(bb,0x00),b1=_mm256_shuffle_epi32(b0,_MM_PERM_CDAB);
const auto a0=aa,a1=_mm256_srli_epi64(a0,32);
const auto aw7=_mm256_alignr_epi8(aa,awa,12);
auto res00=_mm256_mul_epu32(a0,b0);
auto res01=_mm256_mul_epu32(a1,b0);
auto res10=_mm256_mul_epu32(aw7,b1);
auto res11=_mm256_mul_epu32(a0,b1);
const auto b2=_mm256_permute4x64_epi64(bb,0x55),b3=_mm256_shuffle_epi32(b2,_MM_PERM_CDAB);
const auto aw6=_mm256_alignr_epi8(aa,awa,8);
const auto aw5=_mm256_alignr_epi8(aa,awa,4);
res00=_mm256_add_epi64(res00,_mm256_mul_epu32(aw6,b2));
res01=_mm256_add_epi64(res01,_mm256_mul_epu32(aw7,b2));
res10=_mm256_add_epi64(res10,_mm256_mul_epu32(aw5,b3));
res11=_mm256_add_epi64(res11,_mm256_mul_epu32(aw6,b3));
const auto b4=_mm256_permute4x64_epi64(bb,0xaa),b5=_mm256_shuffle_epi32(b4,_MM_PERM_CDAB);
const auto aw3=_mm256_alignr_epi8(awa,aw,12);
res00=_mm256_add_epi64(res00,_mm256_mul_epu32(awa,b4));
res01=_mm256_add_epi64(res01,_mm256_mul_epu32(aw5,b4));
res10=_mm256_add_epi64(res10,_mm256_mul_epu32(aw3,b5));
res11=_mm256_add_epi64(res11,_mm256_mul_epu32(awa,b5));
const auto b6=_mm256_permute4x64_epi64(bb,0xff),b7=_mm256_shuffle_epi32(b6,_MM_PERM_CDAB);
const auto aw2=_mm256_alignr_epi8(awa,aw,8);
const auto aw1=_mm256_alignr_epi8(awa,aw,4);
res00=_mm256_add_epi64(res00,_mm256_mul_epu32(aw2,b6));
res01=_mm256_add_epi64(res01,_mm256_mul_epu32(aw3,b6));
res10=_mm256_add_epi64(res10,_mm256_mul_epu32(aw1,b7));
res11=_mm256_add_epi64(res11,_mm256_mul_epu32(aw2,b7));
res00=_mm256_add_epi64(res00,res10);
res01=_mm256_add_epi64(res01,res11);
return shrk32(reduce(res00,res01,Niv,Mod),Mod2);
}
inline void vector_convolution_direct(I256*f,const I256*g,idt lm,const FNTT32_info*const info){
u32 RR=info->one;
const auto mod=info->mod,niv=info->niv;
const auto Fx=_mm256_set1_epi32(mul_s((mod-((mod-1)>>(__builtin_ctzll(lm)))),info->r3,niv,mod));
const auto Niv=_mm256_set1_epi32(niv),Mod=_mm256_set1_epi32(mod),Mod2=_mm256_set1_epi32(info->mod2);
for(idt i=0;i<lm;++i){
store256(f+i,convolve8(f+i,g+i,_mm256_set1_epi32(RR),Fx,Niv,Mod,Mod2));
RR=mul(RR,info->RT1[__builtin_ctzll(~i)],niv,mod);
}
}
inline void vector_convolution_accumulate(I256*const result,const I256*const f,
const I256*const g,idt lm,
const FNTT32_info*const info){
u32 RR=info->one;
const auto mod=info->mod,niv=info->niv;
const auto Fx=_mm256_set1_epi32(mul_s((mod-((mod-1)>>(__builtin_ctzll(lm)))),info->r3,niv,mod));
const auto Niv=_mm256_set1_epi32(niv),Mod=_mm256_set1_epi32(mod),Mod2=_mm256_set1_epi32(info->mod2);
for(idt i=0;i<lm;++i){
const auto product=convolve8(f+i,g+i,_mm256_set1_epi32(RR),Fx,Niv,Mod,Mod2);
store256(result+i,add32(load256(result+i),product,Mod2));
RR=mul(RR,info->RT1[__builtin_ctzll(~i)],niv,mod);
}
}
} // namespace fast998_v2
} // namespace internal
} // namespace fps
} // namespace m1une
#endif // M1UNE_FPS_HAS_X86_SIMD
#line 24 "math/fps/convolution.hpp"
#ifdef M1UNE_FPS_HAS_X86_SIMD
#pragma GCC pop_options
#endif
#line 1 "math/modint.hpp"
#line 6 "math/modint.hpp"
#include <iostream>
#line 9 "math/modint.hpp"
namespace m1une {
namespace math {
template <uint32_t Modulus>
struct ModInt {
static_assert(0 < Modulus, "Modulus must be positive");
private:
uint32_t _v;
public:
static constexpr uint32_t mod() {
return Modulus;
}
static constexpr ModInt raw(uint32_t v) noexcept {
ModInt x;
x._v = v;
return x;
}
constexpr ModInt() noexcept : _v(0) {}
template <class Integer, std::enable_if_t<std::is_integral_v<Integer>, int> = 0>
constexpr ModInt(Integer v) noexcept {
if constexpr (std::is_signed_v<Integer>) {
int64_t x = static_cast<int64_t>(v) % static_cast<int64_t>(Modulus);
if (x < 0) x += Modulus;
_v = static_cast<uint32_t>(x);
} else {
_v = static_cast<uint32_t>(static_cast<uint64_t>(v) % Modulus);
}
}
constexpr uint32_t val() const noexcept {
return _v;
}
constexpr ModInt& operator++() noexcept {
_v++;
if (_v == Modulus) _v = 0;
return *this;
}
constexpr ModInt& operator--() noexcept {
if (_v == 0) _v = Modulus;
_v--;
return *this;
}
constexpr ModInt operator++(int) noexcept {
ModInt res = *this;
++*this;
return res;
}
constexpr ModInt operator--(int) noexcept {
ModInt res = *this;
--*this;
return res;
}
constexpr ModInt& operator+=(const ModInt& rhs) noexcept {
_v += rhs._v;
if (_v >= Modulus) _v -= Modulus;
return *this;
}
constexpr ModInt& operator-=(const ModInt& rhs) noexcept {
_v -= rhs._v;
if (_v >= Modulus) _v += Modulus;
return *this;
}
constexpr ModInt& operator*=(const ModInt& rhs) noexcept {
uint64_t z = _v;
z *= rhs._v;
_v = static_cast<uint32_t>(z % Modulus);
return *this;
}
constexpr ModInt& operator/=(const ModInt& rhs) noexcept {
return *this *= rhs.inv();
}
constexpr ModInt operator+(const ModInt& rhs) const noexcept {
return ModInt(*this) += rhs;
}
constexpr ModInt operator-(const ModInt& rhs) const noexcept {
return ModInt(*this) -= rhs;
}
constexpr ModInt operator*(const ModInt& rhs) const noexcept {
return ModInt(*this) *= rhs;
}
constexpr ModInt operator/(const ModInt& rhs) const noexcept {
return ModInt(*this) /= rhs;
}
constexpr bool operator==(const ModInt& rhs) const noexcept {
return _v == rhs._v;
}
constexpr bool operator!=(const ModInt& rhs) const noexcept {
return _v != rhs._v;
}
constexpr ModInt pow(long long n) const noexcept {
ModInt res = raw(1 % Modulus);
ModInt x = n < 0 ? inv() : *this;
uint64_t exponent = n < 0 ? uint64_t(-(n + 1)) + 1 : uint64_t(n);
while (exponent > 0) {
if (exponent & 1) res *= x;
x *= x;
exponent >>= 1;
}
return res;
}
constexpr ModInt inv() const noexcept {
int64_t a = _v, b = Modulus, u = 1, v = 0;
while (b) {
int64_t t = a / b;
a -= t * b;
std::swap(a, b);
u -= t * v;
std::swap(u, v);
}
assert(a == 1);
u %= Modulus;
if (u < 0) u += Modulus;
return raw(static_cast<uint32_t>(u));
}
friend std::ostream& operator<<(std::ostream& os, const ModInt& rhs) {
return os << rhs._v;
}
friend std::istream& operator>>(std::istream& is, ModInt& rhs) {
long long v;
is >> v;
rhs = ModInt(v);
return is;
}
};
using modint998244353 = ModInt<998244353>;
using modint1000000007 = ModInt<1000000007>;
template <int Id = 0>
struct DynamicModInt {
private:
uint32_t _v;
inline static uint32_t _mod = 1;
public:
static uint32_t mod() noexcept {
return _mod;
}
static void set_mod(uint32_t modulus) noexcept {
assert(modulus > 0);
assert(modulus <= uint32_t(1) << 31);
_mod = modulus;
}
static DynamicModInt raw(uint32_t v) noexcept {
assert(v < _mod);
DynamicModInt x;
x._v = v;
return x;
}
DynamicModInt() noexcept : _v(0) {}
template <class Integer, std::enable_if_t<std::is_integral_v<Integer>, int> = 0>
DynamicModInt(Integer v) noexcept {
if constexpr (std::is_signed_v<Integer>) {
int64_t x = static_cast<int64_t>(v) % static_cast<int64_t>(_mod);
if (x < 0) x += _mod;
_v = static_cast<uint32_t>(x);
} else {
_v = static_cast<uint32_t>(static_cast<uint64_t>(v) % _mod);
}
}
uint32_t val() const noexcept {
return _v;
}
DynamicModInt& operator++() noexcept {
_v++;
if (_v == _mod) _v = 0;
return *this;
}
DynamicModInt& operator--() noexcept {
if (_v == 0) _v = _mod;
_v--;
return *this;
}
DynamicModInt operator++(int) noexcept {
DynamicModInt result = *this;
++*this;
return result;
}
DynamicModInt operator--(int) noexcept {
DynamicModInt result = *this;
--*this;
return result;
}
DynamicModInt& operator+=(const DynamicModInt& rhs) noexcept {
_v += rhs._v;
if (_v >= _mod) _v -= _mod;
return *this;
}
DynamicModInt& operator-=(const DynamicModInt& rhs) noexcept {
_v -= rhs._v;
if (_v >= _mod) _v += _mod;
return *this;
}
DynamicModInt& operator*=(const DynamicModInt& rhs) noexcept {
_v = static_cast<uint32_t>(uint64_t(_v) * rhs._v % _mod);
return *this;
}
DynamicModInt& operator/=(const DynamicModInt& rhs) noexcept {
return *this *= rhs.inv();
}
DynamicModInt operator+(const DynamicModInt& rhs) const noexcept {
return DynamicModInt(*this) += rhs;
}
DynamicModInt operator-(const DynamicModInt& rhs) const noexcept {
return DynamicModInt(*this) -= rhs;
}
DynamicModInt operator*(const DynamicModInt& rhs) const noexcept {
return DynamicModInt(*this) *= rhs;
}
DynamicModInt operator/(const DynamicModInt& rhs) const noexcept {
return DynamicModInt(*this) /= rhs;
}
bool operator==(const DynamicModInt& rhs) const noexcept {
return _v == rhs._v;
}
bool operator!=(const DynamicModInt& rhs) const noexcept {
return _v != rhs._v;
}
DynamicModInt pow(long long exponent) const noexcept {
DynamicModInt result = raw(1 % _mod);
DynamicModInt base = exponent < 0 ? inv() : *this;
uint64_t magnitude =
exponent < 0 ? uint64_t(-(exponent + 1)) + 1 : uint64_t(exponent);
while (magnitude > 0) {
if (magnitude & 1) result *= base;
base *= base;
magnitude >>= 1;
}
return result;
}
DynamicModInt inv() const noexcept {
int64_t a = _v, b = _mod, u = 1, v = 0;
while (b) {
int64_t quotient = a / b;
a -= quotient * b;
std::swap(a, b);
u -= quotient * v;
std::swap(u, v);
}
assert(a == 1);
u %= _mod;
if (u < 0) u += _mod;
return raw(static_cast<uint32_t>(u));
}
friend std::ostream& operator<<(std::ostream& os, const DynamicModInt& rhs) {
return os << rhs._v;
}
friend std::istream& operator>>(std::istream& is, DynamicModInt& rhs) {
long long value;
is >> value;
rhs = DynamicModInt(value);
return is;
}
};
} // namespace math
} // namespace m1une
#line 29 "math/fps/convolution.hpp"
namespace m1une {
namespace fps {
namespace internal {
template <class Mint, class = void>
struct has_static_modulus : std::false_type {};
template <class Mint>
struct has_static_modulus<
Mint, std::void_t<decltype(std::integral_constant<uint32_t, Mint::mod()>{})>>
: std::true_type {};
constexpr uint32_t primitive_root_constexpr(uint32_t mod) {
if (mod == 2) return 1;
if (mod == 167772161) return 3;
if (mod == 469762049) return 3;
if (mod == 754974721) return 11;
if (mod == 998244353) return 3;
if (mod == 1224736769) return 3;
uint32_t divisors[32] = {};
int count = 0;
uint32_t x = mod - 1;
for (uint32_t p = 2; uint64_t(p) * p <= x; p++) {
if (x % p != 0) continue;
divisors[count++] = p;
while (x % p == 0) x /= p;
}
if (x > 1) divisors[count++] = x;
for (uint32_t g = 2;; g++) {
bool ok = true;
for (int i = 0; i < count; i++) {
uint64_t value = 1;
uint64_t base = g;
uint32_t exponent = (mod - 1) / divisors[i];
while (exponent > 0) {
if (exponent & 1) value = value * base % mod;
base = base * base % mod;
exponent >>= 1;
}
if (value == 1) {
ok = false;
break;
}
}
if (ok) return g;
}
}
constexpr int two_adic_order(uint32_t x) {
int result = 0;
while ((x & 1) == 0) {
x >>= 1;
result++;
}
return result;
}
template <class Mint>
struct NttRoots {
static constexpr int max_base = two_adic_order(Mint::mod() - 1);
std::array<Mint, max_base + 1> root;
std::array<Mint, max_base + 1> inverse_root;
std::array<Mint, max_base> rate;
std::array<Mint, max_base> inverse_rate;
std::array<Mint, max_base> rate_radix4;
std::array<Mint, max_base> inverse_rate_radix4;
NttRoots() {
constexpr uint32_t primitive_root = primitive_root_constexpr(Mint::mod());
for (int level = 1; level <= max_base; level++) {
root[level] = Mint(primitive_root).pow((Mint::mod() - 1) >> level);
inverse_root[level] = root[level].inv();
}
Mint product = 1;
Mint inverse_product = 1;
for (int i = 0; i + 1 < max_base; i++) {
rate[i] = root[i + 2] * product;
inverse_rate[i] = inverse_root[i + 2] * inverse_product;
product *= inverse_root[i + 2];
inverse_product *= root[i + 2];
}
product = 1;
inverse_product = 1;
for (int i = 0; i + 2 < max_base; i++) {
rate_radix4[i] = root[i + 3] * product;
inverse_rate_radix4[i] = inverse_root[i + 3] * inverse_product;
product *= inverse_root[i + 3];
inverse_product *= root[i + 3];
}
}
};
template <class Mint>
const NttRoots<Mint>& ntt_roots() {
static const NttRoots<Mint> roots;
return roots;
}
template <class Mint>
void ntt(std::vector<Mint>& a, bool inverse, bool normalize = true) {
const int n = int(a.size());
assert(n > 0 && (n & (n - 1)) == 0);
assert((Mint::mod() - 1) % uint32_t(n) == 0);
const auto& roots = ntt_roots<Mint>();
const int height = two_adic_order(uint32_t(n));
if (!inverse) {
int phase = 0;
while (phase < height) {
if (height - phase == 1) {
const int width = 1 << (height - phase - 1);
Mint twiddle = 1;
for (int block = 0; block < (1 << phase); block++) {
const int offset = block << (height - phase);
for (int i = 0; i < width; i++) {
const Mint left = a[offset + i];
const Mint right = a[offset + i + width] * twiddle;
a[offset + i] = left + right;
a[offset + i + width] = left - right;
}
if (block + 1 != (1 << phase))
twiddle *= roots.rate[__builtin_ctz(~uint32_t(block))];
}
phase++;
continue;
}
const int width = 1 << (height - phase - 2);
Mint twiddle = 1;
const Mint imaginary = roots.root[2];
for (int block = 0; block < (1 << phase); block++) {
const Mint twiddle2 = twiddle * twiddle;
const Mint twiddle3 = twiddle2 * twiddle;
const int offset = block << (height - phase);
for (int i = 0; i < width; i++) {
const uint64_t mod2 = uint64_t(Mint::mod()) * Mint::mod();
const uint64_t a0 = a[offset + i].val();
const uint64_t a1 = uint64_t(a[offset + i + width].val()) * twiddle.val();
const uint64_t a2 =
uint64_t(a[offset + i + 2 * width].val()) * twiddle2.val();
const uint64_t a3 =
uint64_t(a[offset + i + 3 * width].val()) * twiddle3.val();
const uint64_t a1na3i =
uint64_t(Mint(a1 + mod2 - a3).val()) * imaginary.val();
const uint64_t negative_a2 = mod2 - a2;
a[offset + i] = Mint(a0 + a2 + a1 + a3);
a[offset + i + width] = Mint(a0 + a2 + 2 * mod2 - a1 - a3);
a[offset + i + 2 * width] = Mint(a0 + negative_a2 + a1na3i);
a[offset + i + 3 * width] = Mint(a0 + negative_a2 + mod2 - a1na3i);
}
if (block + 1 != (1 << phase))
twiddle *= roots.rate_radix4[__builtin_ctz(~uint32_t(block))];
}
phase += 2;
}
} else {
int phase = height;
while (phase > 0) {
if (phase == 1) {
const int width = 1 << (height - phase);
Mint twiddle = 1;
for (int block = 0; block < (1 << (phase - 1)); block++) {
const int offset = block << (height - phase + 1);
for (int i = 0; i < width; i++) {
const Mint left = a[offset + i];
const Mint right = a[offset + i + width];
a[offset + i] = left + right;
a[offset + i + width] = (left - right) * twiddle;
}
if (block + 1 != (1 << (phase - 1)))
twiddle *= roots.inverse_rate[__builtin_ctz(~uint32_t(block))];
}
phase--;
continue;
}
const int width = 1 << (height - phase);
Mint twiddle = 1;
const Mint inverse_imaginary = roots.inverse_root[2];
for (int block = 0; block < (1 << (phase - 2)); block++) {
const Mint twiddle2 = twiddle * twiddle;
const Mint twiddle3 = twiddle2 * twiddle;
const int offset = block << (height - phase + 2);
for (int i = 0; i < width; i++) {
const uint64_t a0 = a[offset + i].val();
const uint64_t a1 = a[offset + i + width].val();
const uint64_t a2 = a[offset + i + 2 * width].val();
const uint64_t a3 = a[offset + i + 3 * width].val();
const uint64_t a2na3i =
uint64_t(Mint((Mint::mod() + a2 - a3) * inverse_imaginary.val()).val());
a[offset + i] = Mint(a0 + a1 + a2 + a3);
a[offset + i + width] =
Mint((a0 + Mint::mod() - a1 + a2na3i) * twiddle.val());
a[offset + i + 2 * width] = Mint(
(a0 + a1 + 2ULL * Mint::mod() - a2 - a3) * twiddle2.val());
a[offset + i + 3 * width] = Mint(
(a0 + Mint::mod() - a1 + Mint::mod() - a2na3i) * twiddle3.val());
}
if (block + 1 != (1 << (phase - 2)))
twiddle *= roots.inverse_rate_radix4[__builtin_ctz(~uint32_t(block))];
}
phase -= 2;
}
if (normalize) {
const Mint inverse_n = Mint(n).inv();
for (Mint& value : a) value *= inverse_n;
}
}
}
#ifdef M1UNE_FPS_HAS_X86_SIMD
#pragma GCC push_options
#pragma GCC target("avx2,bmi")
template <class Mint>
__attribute__((target("avx2,bmi"), hot))
std::vector<Mint> convolution_998244353_simd(const std::vector<Mint>& a,
const std::vector<Mint>& b) {
const int result_size = int(a.size() + b.size() - 1);
int n = 1;
while (n < result_size) n <<= 1;
const bool squaring = &a == &b;
auto* transformed_a = static_cast<uint32_t*>(
::operator new[](sizeof(uint32_t) * n, std::align_val_t(32)));
auto* transformed_b = squaring
? transformed_a
: static_cast<uint32_t*>(::operator new[](
sizeof(uint32_t) * n, std::align_val_t(32)));
if constexpr (std::is_same_v<Mint, math::ModInt<998244353>>) {
static_assert(sizeof(Mint) == sizeof(uint32_t) && std::is_trivially_copyable_v<Mint>);
std::memcpy(transformed_a, a.data(), sizeof(uint32_t) * a.size());
if (!squaring)
std::memcpy(transformed_b, b.data(), sizeof(uint32_t) * b.size());
} else {
for (int i = 0; i < int(a.size()); i++) transformed_a[i] = a[i].val();
if (!squaring)
for (int i = 0; i < int(b.size()); i++) transformed_b[i] = b[i].val();
}
std::memset(transformed_a + a.size(), 0, sizeof(uint32_t) * (n - a.size()));
if (!squaring)
std::memset(transformed_b + b.size(), 0, sizeof(uint32_t) * (n - b.size()));
static constexpr fast998_v2::FNTT32_info transform(998244353);
const std::size_t vector_size = std::size_t(n) >> 3;
fast998_v2::vector_dif(reinterpret_cast<__m256i*>(transformed_a), vector_size, &transform);
if (!squaring)
fast998_v2::vector_dif(reinterpret_cast<__m256i*>(transformed_b), vector_size,
&transform);
fast998_v2::vector_convolution_direct(
reinterpret_cast<__m256i*>(transformed_a),
reinterpret_cast<const __m256i*>(transformed_b), vector_size, &transform);
fast998_v2::vector_dit<true>(reinterpret_cast<__m256i*>(transformed_a), vector_size,
&transform);
std::vector<Mint> result(result_size);
for (int j = 0; j < result_size; j++) result[j] = Mint::raw(transformed_a[j]);
::operator delete[](transformed_a, std::align_val_t(32));
if (!squaring) ::operator delete[](transformed_b, std::align_val_t(32));
return result;
}
#pragma GCC pop_options
#endif
} // namespace internal
template <class Mint>
std::vector<Mint> convolution_naive(const std::vector<Mint>& a, const std::vector<Mint>& b) {
if (a.empty() || b.empty()) return {};
std::vector<Mint> result(a.size() + b.size() - 1);
if (a.size() < b.size()) {
for (int i = 0; i < int(a.size()); i++) {
for (int j = 0; j < int(b.size()); j++) result[i + j] += a[i] * b[j];
}
} else {
for (int j = 0; j < int(b.size()); j++) {
for (int i = 0; i < int(a.size()); i++) result[i + j] += a[i] * b[j];
}
}
return result;
}
template <class Mint>
std::vector<Mint> convolution_ntt(const std::vector<Mint>& a, const std::vector<Mint>& b) {
const int result_size = int(a.size() + b.size() - 1);
int n = 1;
while (n < result_size) n <<= 1;
assert((Mint::mod() - 1) % uint32_t(n) == 0);
#ifdef M1UNE_FPS_HAS_X86_SIMD
if constexpr (Mint::mod() == 998244353) {
if (n >= 64 && __builtin_cpu_supports("avx2"))
return internal::convolution_998244353_simd(a, b);
}
#endif
// Allocate the padded buffers directly. Constructing from the inputs and
// then resizing used to allocate and copy both large operands twice.
const bool squaring = &a == &b;
std::vector<Mint> fa(n);
std::copy(a.begin(), a.end(), fa.begin());
internal::ntt(fa, false);
const Mint inverse_n = Mint(n).inv();
if (squaring) {
for (int i = 0; i < n; i++) fa[i] *= fa[i] * inverse_n;
} else {
std::vector<Mint> fb(n);
std::copy(b.begin(), b.end(), fb.begin());
internal::ntt(fb, false);
for (int i = 0; i < n; i++) fa[i] *= fb[i] * inverse_n;
}
internal::ntt(fa, true, false);
fa.resize(result_size);
return fa;
}
namespace internal {
template <class Mint>
std::vector<Mint> convolution_998244353_blocked_scalar(const std::vector<Mint>& a,
const std::vector<Mint>& b,
int transform_size) {
assert(Mint::mod() == 998244353);
assert(transform_size >= 2 && (transform_size & (transform_size - 1)) == 0);
assert((Mint::mod() - 1) % uint32_t(transform_size) == 0);
const int block_size = transform_size / 2;
const int a_blocks = int((a.size() + block_size - 1) / block_size);
const int b_blocks = int((b.size() + block_size - 1) / block_size);
auto transform_blocks = [&](const std::vector<Mint>& values, int block_count) {
std::vector<std::vector<Mint>> blocks;
blocks.reserve(block_count);
for (int block = 0; block < block_count; block++) {
const int begin = block * block_size;
const int count = std::min(block_size, int(values.size()) - begin);
std::vector<Mint> transformed(transform_size);
std::copy_n(values.begin() + begin, count, transformed.begin());
ntt(transformed, false);
blocks.emplace_back(std::move(transformed));
}
return blocks;
};
std::vector<std::vector<Mint>> transformed_a = transform_blocks(a, a_blocks);
std::vector<std::vector<Mint>> transformed_b = transform_blocks(b, b_blocks);
const int result_size = int(a.size() + b.size() - 1);
std::vector<Mint> result(result_size);
std::vector<Mint> transformed_result(transform_size);
for (int diagonal = 0; diagonal < a_blocks + b_blocks - 1; diagonal++) {
std::fill(transformed_result.begin(), transformed_result.end(), Mint(0));
const int first_a = std::max(0, diagonal - (b_blocks - 1));
const int last_a = std::min(a_blocks - 1, diagonal);
for (int a_block = first_a; a_block <= last_a; a_block++) {
const int b_block = diagonal - a_block;
for (int i = 0; i < transform_size; i++)
transformed_result[i] +=
transformed_a[a_block][i] * transformed_b[b_block][i];
}
ntt(transformed_result, true);
const int output_offset = diagonal * block_size;
const int output_count = std::min(transform_size, result_size - output_offset);
for (int i = 0; i < output_count; i++)
result[output_offset + i] += transformed_result[i];
}
return result;
}
#ifdef M1UNE_FPS_HAS_X86_SIMD
class AlignedUint32Buffer {
private:
uint32_t* data_;
public:
explicit AlignedUint32Buffer(std::size_t size)
: data_(static_cast<uint32_t*>(
::operator new[](sizeof(uint32_t) * size, std::align_val_t(32)))) {}
AlignedUint32Buffer(const AlignedUint32Buffer&) = delete;
AlignedUint32Buffer& operator=(const AlignedUint32Buffer&) = delete;
AlignedUint32Buffer(AlignedUint32Buffer&& other) noexcept : data_(other.data_) {
other.data_ = nullptr;
}
AlignedUint32Buffer& operator=(AlignedUint32Buffer&& other) noexcept {
if (this == &other) return *this;
::operator delete[](data_, std::align_val_t(32));
data_ = other.data_;
other.data_ = nullptr;
return *this;
}
~AlignedUint32Buffer() {
::operator delete[](data_, std::align_val_t(32));
}
uint32_t* data() {
return data_;
}
const uint32_t* data() const {
return data_;
}
};
template <class Mint>
__attribute__((target("avx2,bmi"), hot))
std::vector<Mint> convolution_998244353_blocked_simd(const std::vector<Mint>& a,
const std::vector<Mint>& b,
int transform_size) {
assert(Mint::mod() == 998244353);
assert(transform_size >= 64 && (transform_size & (transform_size - 1)) == 0);
assert((Mint::mod() - 1) % uint32_t(transform_size) == 0);
const int block_size = transform_size / 2;
const int a_blocks = int((a.size() + block_size - 1) / block_size);
const int b_blocks = int((b.size() + block_size - 1) / block_size);
static constexpr fast998_v2::FNTT32_info transform(998244353);
const std::size_t vector_size = std::size_t(transform_size) / 8;
auto transform_blocks = [&](const std::vector<Mint>& values, int block_count) {
std::vector<AlignedUint32Buffer> blocks;
blocks.reserve(block_count);
for (int block = 0; block < block_count; block++) {
const int begin = block * block_size;
const int count = std::min(block_size, int(values.size()) - begin);
AlignedUint32Buffer transformed(transform_size);
if constexpr (std::is_same_v<Mint, math::ModInt<998244353>>) {
static_assert(sizeof(Mint) == sizeof(uint32_t) &&
std::is_trivially_copyable_v<Mint>);
std::memcpy(transformed.data(), values.data() + begin,
sizeof(uint32_t) * count);
} else {
for (int i = 0; i < count; i++)
transformed.data()[i] = values[begin + i].val();
}
std::memset(transformed.data() + count, 0,
sizeof(uint32_t) * (transform_size - count));
fast998_v2::vector_dif(reinterpret_cast<__m256i*>(transformed.data()),
vector_size, &transform);
blocks.emplace_back(std::move(transformed));
}
return blocks;
};
std::vector<AlignedUint32Buffer> transformed_a = transform_blocks(a, a_blocks);
std::vector<AlignedUint32Buffer> transformed_b = transform_blocks(b, b_blocks);
const int result_size = int(a.size() + b.size() - 1);
std::vector<Mint> result(result_size);
AlignedUint32Buffer transformed_result(transform_size);
for (int diagonal = 0; diagonal < a_blocks + b_blocks - 1; diagonal++) {
std::memset(transformed_result.data(), 0, sizeof(uint32_t) * transform_size);
const int first_a = std::max(0, diagonal - (b_blocks - 1));
const int last_a = std::min(a_blocks - 1, diagonal);
for (int a_block = first_a; a_block <= last_a; a_block++) {
const int b_block = diagonal - a_block;
fast998_v2::vector_convolution_accumulate(
reinterpret_cast<__m256i*>(transformed_result.data()),
reinterpret_cast<const __m256i*>(transformed_a[a_block].data()),
reinterpret_cast<const __m256i*>(transformed_b[b_block].data()),
vector_size, &transform);
}
fast998_v2::vector_dit<true>(
reinterpret_cast<__m256i*>(transformed_result.data()), vector_size,
&transform);
const int output_offset = diagonal * block_size;
const int output_count = std::min(transform_size, result_size - output_offset);
for (int i = 0; i < output_count; i++) {
uint32_t value = result[output_offset + i].val() + transformed_result.data()[i];
if (value >= Mint::mod()) value -= Mint::mod();
result[output_offset + i] = Mint::raw(value);
}
}
return result;
}
#endif
template <class Mint>
std::vector<Mint> convolution_998244353_blocked(const std::vector<Mint>& a,
const std::vector<Mint>& b,
int transform_size = 1 << 23) {
#ifdef M1UNE_FPS_HAS_X86_SIMD
if (transform_size >= 64 && __builtin_cpu_supports("avx2"))
return convolution_998244353_blocked_simd(a, b, transform_size);
#endif
return convolution_998244353_blocked_scalar(a, b, transform_size);
}
} // namespace internal
template <class Mint>
std::vector<Mint> convolution(const std::vector<Mint>& a, const std::vector<Mint>& b) {
if (a.empty() || b.empty()) return {};
if (std::min(a.size(), b.size()) <= 32) return convolution_naive(a, b);
const int result_size = int(a.size() + b.size() - 1);
int n = 1;
while (n < result_size) n <<= 1;
if constexpr (internal::has_static_modulus<Mint>::value) {
if constexpr (Mint::mod() == 998244353) {
if (n > (1 << 23))
return internal::convolution_998244353_blocked(a, b);
}
if ((Mint::mod() - 1) % uint32_t(n) == 0) return convolution_ntt(a, b);
}
using Mint1 = math::ModInt<167772161>;
using Mint2 = math::ModInt<469762049>;
using Mint3 = math::ModInt<754974721>;
assert(n <= (1 << 24));
[[maybe_unused]] const unsigned __int128 coefficient_bound =
static_cast<unsigned __int128>(std::min(a.size(), b.size())) * (Mint::mod() - 1) *
(Mint::mod() - 1);
[[maybe_unused]] const unsigned __int128 crt_modulus =
static_cast<unsigned __int128>(Mint1::mod()) * Mint2::mod() * Mint3::mod();
assert(coefficient_bound < crt_modulus);
auto converted_convolution = [&]<class OtherMint>() {
std::vector<OtherMint> converted_a(a.size());
std::vector<OtherMint> converted_b(b.size());
for (int i = 0; i < int(a.size()); i++) converted_a[i] = OtherMint(a[i].val());
for (int i = 0; i < int(b.size()); i++) converted_b[i] = OtherMint(b[i].val());
return convolution_ntt(converted_a, converted_b);
};
std::vector<Mint1> c1 = converted_convolution.template operator()<Mint1>();
std::vector<Mint2> c2 = converted_convolution.template operator()<Mint2>();
std::vector<Mint3> c3 = converted_convolution.template operator()<Mint3>();
static const uint64_t inverse_mod1_mod2 = Mint2(Mint1::mod()).inv().val();
static const uint64_t mod1_mod3 = Mint1::mod() % Mint3::mod();
static const uint64_t mod1_mod2_mod3 =
mod1_mod3 * (Mint2::mod() % Mint3::mod()) % Mint3::mod();
static const uint64_t inverse_mod1_mod2_mod3 = Mint3(uint32_t(mod1_mod2_mod3)).inv().val();
const uint64_t target_mod = Mint::mod();
const uint64_t mod1_target = Mint1::mod() % target_mod;
const uint64_t mod1_mod2_target = mod1_target * (Mint2::mod() % target_mod) % target_mod;
std::vector<Mint> result(result_size);
for (int i = 0; i < result_size; i++) {
const uint64_t r1 = c1[i].val();
const uint64_t r2 = c2[i].val();
const uint64_t r3 = c3[i].val();
const uint64_t first =
(r2 + Mint2::mod() - r1 % Mint2::mod()) % Mint2::mod() * inverse_mod1_mod2 %
Mint2::mod();
const uint64_t combined_mod3 =
(r1 % Mint3::mod() + mod1_mod3 * (first % Mint3::mod())) % Mint3::mod();
const uint64_t second =
(r3 + Mint3::mod() - combined_mod3) % Mint3::mod() * inverse_mod1_mod2_mod3 %
Mint3::mod();
uint64_t value = r1 % target_mod;
value = (value + mod1_target * (first % target_mod)) % target_mod;
value = (value + mod1_mod2_target * (second % target_mod)) % target_mod;
result[i] = Mint::raw(uint32_t(value));
}
return result;
}
} // namespace fps
} // namespace m1une
#ifdef M1UNE_FPS_HAS_X86_SIMD
#undef M1UNE_FPS_HAS_X86_SIMD
#endif
#line 1 "math/fps/formal_power_series.hpp"
#line 10 "math/fps/formal_power_series.hpp"
#line 1 "math/modular_square_root.hpp"
#line 7 "math/modular_square_root.hpp"
namespace m1une {
namespace math {
namespace internal {
inline uint64_t modular_square_root_multiply(uint64_t lhs, uint64_t rhs, uint64_t mod) {
return static_cast<uint64_t>(static_cast<unsigned __int128>(lhs) * rhs % mod);
}
inline uint64_t modular_square_root_power(uint64_t base, uint64_t exponent, uint64_t mod) {
uint64_t result = 1 % mod;
while (exponent > 0) {
if (exponent & 1) result = modular_square_root_multiply(result, base, mod);
base = modular_square_root_multiply(base, base, mod);
exponent >>= 1;
}
return result;
}
} // namespace internal
// Returns x such that x * x = value (mod prime), or nullopt when no such x exists.
// The modulus must be prime.
inline std::optional<uint64_t> modular_square_root(uint64_t value, uint64_t prime) {
assert(prime >= 2);
value %= prime;
if (value == 0 || prime == 2) return value;
if (internal::modular_square_root_power(value, (prime - 1) / 2, prime) != 1) {
return std::nullopt;
}
if (prime % 4 == 3) {
return internal::modular_square_root_power(value, prime / 4 + 1, prime);
}
uint64_t odd_part = prime - 1;
int power_of_two = 0;
while ((odd_part & 1) == 0) {
odd_part >>= 1;
power_of_two++;
}
uint64_t non_residue = 2;
while (internal::modular_square_root_power(non_residue, (prime - 1) / 2, prime) == 1) {
non_residue++;
}
uint64_t c = internal::modular_square_root_power(non_residue, odd_part, prime);
uint64_t root = internal::modular_square_root_power(value, odd_part / 2 + 1, prime);
uint64_t remainder = internal::modular_square_root_power(value, odd_part, prime);
int remaining_power = power_of_two;
while (remainder != 1) {
int exponent = 1;
uint64_t squared = internal::modular_square_root_multiply(remainder, remainder, prime);
while (squared != 1) {
squared = internal::modular_square_root_multiply(squared, squared, prime);
exponent++;
}
uint64_t correction = c;
for (int i = 0; i < remaining_power - exponent - 1; i++) {
correction = internal::modular_square_root_multiply(correction, correction, prime);
}
root = internal::modular_square_root_multiply(root, correction, prime);
c = internal::modular_square_root_multiply(correction, correction, prime);
remainder = internal::modular_square_root_multiply(remainder, c, prime);
remaining_power = exponent;
}
return root;
}
template <class Mint>
std::optional<Mint> modular_square_root(Mint value) {
auto root = modular_square_root(static_cast<uint64_t>(value.val()),
static_cast<uint64_t>(Mint::mod()));
if (!root.has_value()) return std::nullopt;
return Mint(*root);
}
} // namespace math
} // namespace m1une
#line 13 "math/fps/formal_power_series.hpp"
namespace m1une {
namespace fps {
template <class Mint>
struct FormalPowerSeries : std::vector<Mint> {
using std::vector<Mint>::vector;
using Fps = FormalPowerSeries;
FormalPowerSeries() = default;
FormalPowerSeries(const std::vector<Mint>& values) : std::vector<Mint>(values) {}
FormalPowerSeries(std::vector<Mint>&& values) : std::vector<Mint>(std::move(values)) {}
Fps& shrink() {
while (!this->empty() && this->back() == Mint(0)) this->pop_back();
return *this;
}
Fps pre(int degree) const {
assert(degree >= 0);
Fps result(this->begin(), this->begin() + std::min<int>(degree, this->size()));
result.resize(degree);
return result;
}
Fps reversed(int size = -1) const {
Fps result = *this;
if (size >= 0) result.resize(size);
std::reverse(result.begin(), result.end());
return result;
}
Fps& operator+=(const Fps& rhs) {
if (this->size() < rhs.size()) this->resize(rhs.size());
for (int i = 0; i < int(rhs.size()); i++) (*this)[i] += rhs[i];
return *this;
}
Fps& operator-=(const Fps& rhs) {
if (this->size() < rhs.size()) this->resize(rhs.size());
for (int i = 0; i < int(rhs.size()); i++) (*this)[i] -= rhs[i];
return *this;
}
Fps& operator*=(const Fps& rhs) {
std::vector<Mint> lhs(this->begin(), this->end());
*this = convolution(lhs, rhs);
return *this;
}
Fps& operator*=(Mint rhs) {
for (Mint& value : *this) value *= rhs;
return *this;
}
Fps& operator/=(Mint rhs) {
return *this *= rhs.inv();
}
Fps& operator<<=(int shift) {
assert(shift >= 0);
this->insert(this->begin(), shift, Mint(0));
return *this;
}
Fps& operator>>=(int shift) {
assert(shift >= 0);
if (shift >= int(this->size())) {
this->clear();
} else {
this->erase(this->begin(), this->begin() + shift);
}
return *this;
}
Fps operator+() const {
return *this;
}
Fps operator-() const {
Fps result = *this;
for (Mint& value : result) value = Mint(0) - value;
return result;
}
friend Fps operator+(Fps lhs, const Fps& rhs) {
return lhs += rhs;
}
friend Fps operator-(Fps lhs, const Fps& rhs) {
return lhs -= rhs;
}
friend Fps operator*(Fps lhs, const Fps& rhs) {
return lhs *= rhs;
}
friend Fps operator*(Fps lhs, Mint rhs) {
return lhs *= rhs;
}
friend Fps operator*(Mint lhs, Fps rhs) {
return rhs *= lhs;
}
friend Fps operator/(Fps lhs, Mint rhs) {
return lhs /= rhs;
}
friend Fps operator<<(Fps lhs, int shift) {
return lhs <<= shift;
}
friend Fps operator>>(Fps lhs, int shift) {
return lhs >>= shift;
}
Fps derivative() const {
if (this->empty()) return {};
Fps result(this->size() - 1);
for (int i = 1; i < int(this->size()); i++) result[i - 1] = (*this)[i] * Mint(i);
return result;
}
Fps integral() const {
Fps result(this->size() + 1);
if (this->empty()) return result;
assert(this->size() < Mint::mod());
std::vector<Mint> inverse(this->size() + 1);
inverse[1] = 1;
for (int i = 2; i <= int(this->size()); i++) {
inverse[i] = Mint(0) - Mint(Mint::mod() / uint32_t(i)) * inverse[Mint::mod() % uint32_t(i)];
}
for (int i = 0; i < int(this->size()); i++) result[i + 1] = (*this)[i] * inverse[i + 1];
return result;
}
Mint evaluate(Mint x) const {
Mint result = 0;
for (auto it = this->rbegin(); it != this->rend(); ++it) result = result * x + *it;
return result;
}
Fps inv(int degree = -1) const {
if (degree < 0) degree = int(this->size());
assert(degree >= 0);
if (degree == 0) return {};
assert(!this->empty() && (*this)[0] != Mint(0));
Fps result(1, (*this)[0].inv());
for (int size = 1; size < degree; size <<= 1) {
const int next_size = std::min(size << 1, degree);
const int transform_size = size << 1;
if (size >= 32 && (Mint::mod() - 1) % uint32_t(transform_size) == 0) {
// Newton's g <- g(2-fg), restricted to the newly determined
// half. Keeping g in the frequency domain avoids two general
// convolutions and their 2x larger padding.
std::vector<Mint> transformed_f(transform_size);
std::copy_n(this->begin(), std::min<int>(this->size(), next_size),
transformed_f.begin());
std::vector<Mint> transformed_g(transform_size);
std::copy(result.begin(), result.end(), transformed_g.begin());
internal::ntt(transformed_f, false);
internal::ntt(transformed_g, false);
std::vector<Mint> error(transform_size);
for (int i = 0; i < transform_size; i++)
error[i] = transformed_f[i] * transformed_g[i];
internal::ntt(error, true);
std::fill(error.begin(), error.begin() + size, Mint(0));
internal::ntt(error, false);
for (int i = 0; i < transform_size; i++) error[i] *= transformed_g[i];
internal::ntt(error, true);
result.resize(next_size);
for (int i = size; i < next_size; i++) result[i] = Mint(0) - error[i];
continue;
}
Fps product = this->pre(next_size) * result;
product.resize(next_size);
for (Mint& value : product) value = Mint(0) - value;
product[0] += Mint(2);
result = (result * product).pre(next_size);
}
return result.pre(degree);
}
Fps log(int degree = -1) const {
if (degree < 0) degree = int(this->size());
assert(degree >= 0);
if (degree == 0) return {};
assert(!this->empty() && (*this)[0] == Mint(1));
return (derivative() * inv(degree)).pre(degree - 1).integral();
}
Fps exp(int degree = -1) const {
if (degree < 0) degree = int(this->size());
assert(degree >= 0);
if (degree == 0) return {};
assert(this->empty() || (*this)[0] == Mint(0));
Fps result(1, Mint(1));
for (int size = 1; size < degree; size <<= 1) {
const int next_size = std::min(size << 1, degree);
Fps correction = this->pre(next_size) - result.log(next_size);
correction[0] += Mint(1);
result = (result * correction).pre(next_size);
}
return result.pre(degree);
}
Fps pow(long long exponent, int degree = -1) const {
if (degree < 0) degree = int(this->size());
assert(exponent >= 0 && degree >= 0);
if (degree == 0) return {};
if (exponent == 0) {
Fps result(degree);
result[0] = 1;
return result;
}
int first = 0;
while (first < int(this->size()) && (*this)[first] == Mint(0)) first++;
if (first == int(this->size()) || first > (degree - 1) / exponent) return Fps(degree);
const int shift = int(first * exponent);
const Mint leading = (*this)[first];
Fps normalized = (*this >> first) / leading;
Fps result = (normalized.log(degree - shift) * Mint(exponent)).exp(degree - shift);
result *= leading.pow(exponent);
result <<= shift;
result.resize(degree);
return result;
}
std::optional<Fps> sqrt(int degree = -1) const {
if (degree < 0) degree = int(this->size());
assert(degree >= 0);
if (degree == 0) return Fps();
int first = 0;
while (first < int(this->size()) && (*this)[first] == Mint(0)) first++;
if (first == int(this->size())) return Fps(degree);
if (first >= degree) return Fps(degree);
if (first & 1) return std::nullopt;
const int shift = first / 2;
auto leading_root = m1une::math::modular_square_root((*this)[first]);
if (!leading_root.has_value()) return std::nullopt;
const int result_degree = degree - shift;
Fps normalized = (*this >> first) / (*this)[first];
Fps result = (normalized.log(result_degree) / Mint(2)).exp(result_degree);
result *= *leading_root;
result <<= shift;
result.resize(degree);
return result;
}
std::pair<Fps, Fps> divmod(const Fps& divisor) const {
Fps dividend = *this;
Fps normalized_divisor = divisor;
dividend.shrink();
normalized_divisor.shrink();
assert(!normalized_divisor.empty());
if (dividend.size() < normalized_divisor.size()) return std::make_pair(Fps(), dividend);
const int quotient_size = int(dividend.size() - normalized_divisor.size() + 1);
Fps quotient =
(dividend.reversed().pre(quotient_size) * normalized_divisor.reversed().inv(quotient_size))
.pre(quotient_size)
.reversed();
quotient.shrink();
Fps remainder = dividend - normalized_divisor * quotient;
remainder.resize(normalized_divisor.size() - 1);
remainder.shrink();
return std::make_pair(std::move(quotient), std::move(remainder));
}
Fps& operator/=(const Fps& rhs) {
*this = divmod(rhs).first;
return *this;
}
Fps& operator%=(const Fps& rhs) {
*this = divmod(rhs).second;
return *this;
}
friend Fps operator/(Fps lhs, const Fps& rhs) {
return lhs /= rhs;
}
friend Fps operator%(Fps lhs, const Fps& rhs) {
return lhs %= rhs;
}
Fps taylor_shift(Mint shift) const {
const int n = int(this->size());
if (n == 0) return {};
assert(uint32_t(n) < Mint::mod());
std::vector<Mint> factorial(n, Mint(1));
std::vector<Mint> inverse_factorial(n, Mint(1));
for (int i = 1; i < n; i++) factorial[i] = factorial[i - 1] * Mint(i);
inverse_factorial[n - 1] = factorial[n - 1].inv();
for (int i = n - 1; i > 0; i--) inverse_factorial[i - 1] = inverse_factorial[i] * Mint(i);
Fps left(n);
Fps right(n);
Mint power = 1;
for (int i = 0; i < n; i++) {
left[n - 1 - i] = (*this)[i] * factorial[i];
right[i] = power * inverse_factorial[i];
power *= shift;
}
Fps product = left * right;
Fps result(n);
for (int i = 0; i < n; i++) result[i] = product[n - 1 - i] * inverse_factorial[i];
return result;
}
};
} // namespace fps
} // namespace m1une
#line 1 "math/combinatorics.hpp"
#line 7 "math/combinatorics.hpp"
namespace m1une {
namespace math {
template <class Mint>
struct Combinatorics {
private:
std::vector<Mint> _factorial;
std::vector<Mint> _inverse_factorial;
public:
explicit Combinatorics(int maximum = 0) : _factorial(1, Mint(1)), _inverse_factorial(1, Mint(1)) {
ensure(maximum);
}
int maximum() const {
return int(_factorial.size()) - 1;
}
void ensure(int maximum) {
assert(maximum >= 0);
assert(static_cast<uint64_t>(maximum) < Mint::mod());
if (maximum <= this->maximum()) return;
const int old_maximum = this->maximum();
_factorial.resize(maximum + 1);
_inverse_factorial.resize(maximum + 1);
for (int i = old_maximum + 1; i <= maximum; i++) {
_factorial[i] = _factorial[i - 1] * Mint(i);
}
_inverse_factorial[maximum] = _factorial[maximum].inv();
for (int i = maximum; i > old_maximum; i--) {
_inverse_factorial[i - 1] = _inverse_factorial[i] * Mint(i);
}
}
Mint factorial(int n) const {
assert(0 <= n && n <= maximum());
return _factorial[n];
}
Mint inverse_factorial(int n) const {
assert(0 <= n && n <= maximum());
return _inverse_factorial[n];
}
Mint inverse(int n) const {
assert(1 <= n && n <= maximum());
return _factorial[n - 1] * _inverse_factorial[n];
}
Mint binom(int n, int k) const {
if (k < 0 || k > n) return Mint(0);
assert(n <= maximum());
return _factorial[n] * _inverse_factorial[k] * _inverse_factorial[n - k];
}
Mint perm(int n, int k) const {
if (k < 0 || k > n) return Mint(0);
assert(n <= maximum());
return _factorial[n] * _inverse_factorial[n - k];
}
Mint multiset(int types, int count) const {
if (types < 0 || count < 0) return Mint(0);
if (types == 0) return Mint(count == 0);
const long long total = static_cast<long long>(types) + count - 1;
assert(total <= maximum());
return binom(static_cast<int>(total), count);
}
Mint catalan(int n) const {
assert(n >= 0);
const long long doubled = 2LL * n;
assert(doubled <= maximum());
return binom(int(doubled), n) - binom(int(doubled), n + 1);
}
};
} // namespace math
} // namespace m1une
#line 13 "graph/counting.hpp"
namespace m1une {
namespace graph {
namespace graph_counting_detail {
template <class Mint>
using Fps = fps::FormalPowerSeries<Mint>;
template <class Mint>
void assert_maximum(int maximum) {
assert(maximum >= 0);
assert(static_cast<uint64_t>(maximum) < Mint::mod());
}
template <class Mint>
Mint inverse_two() {
assert(Mint::mod() != 2);
return Mint(2).inv();
}
template <class Mint>
std::vector<Mint> to_egf(
std::vector<Mint> values,
const math::Combinatorics<Mint>& combinations
) {
for (int i = 0; i < int(values.size()); i++) {
values[i] *= combinations.inverse_factorial(i);
}
return values;
}
template <class Mint>
std::vector<Mint> from_egf(
std::vector<Mint> coefficients,
const math::Combinatorics<Mint>& combinations
) {
for (int i = 0; i < int(coefficients.size()); i++) {
coefficients[i] *= combinations.factorial(i);
}
return coefficients;
}
template <class Mint>
std::vector<Mint> two_to_binom2(int maximum) {
std::vector<Mint> result(maximum + 1);
result[0] = 1;
Mint multiplier = 1;
for (int n = 1; n <= maximum; n++) {
result[n] = result[n - 1] * multiplier;
multiplier += multiplier;
}
return result;
}
template <class Mint>
Fps<Mint> colored_bipartite_egf(
int maximum,
const math::Combinatorics<Mint>& combinations
) {
const Mint half = inverse_two<Mint>();
Fps<Mint> kernel(maximum + 1);
kernel[0] = 1;
Mint multiplier = 1;
for (int i = 1; i <= maximum; i++) {
kernel[i] = kernel[i - 1] * multiplier;
multiplier *= half;
}
for (int i = 0; i <= maximum; i++) {
kernel[i] *= combinations.inverse_factorial(i);
}
Fps<Mint> result = (kernel * kernel).pre(maximum + 1);
std::vector<Mint> edge_powers = two_to_binom2<Mint>(maximum);
for (int i = 0; i <= maximum; i++) result[i] *= edge_powers[i];
return result;
}
} // namespace graph_counting_detail
template <class Mint>
std::vector<Mint> count_labeled_undirected_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
return graph_counting_detail::two_to_binom2<Mint>(maximum);
}
template <class Mint>
std::vector<Mint> count_labeled_connected_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
graph_counting_detail::Fps<Mint> egf =
graph_counting_detail::to_egf(count_labeled_undirected_graphs<Mint>(maximum), combinations);
egf = egf.log(maximum + 1);
return graph_counting_detail::from_egf(std::move(egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_trees(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
std::vector<Mint> result(maximum + 1);
for (int n = 1; n <= maximum; n++) {
result[n] = (n == 1 ? Mint(1) : Mint(n).pow(n - 2));
}
return result;
}
template <class Mint>
std::vector<Mint> count_labeled_forests(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
graph_counting_detail::Fps<Mint> egf =
graph_counting_detail::to_egf(count_labeled_trees<Mint>(maximum), combinations);
egf = egf.exp(maximum + 1);
return graph_counting_detail::from_egf(std::move(egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_unicyclic_connected_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
using Fps = graph_counting_detail::Fps<Mint>;
Fps rooted_tree_egf(maximum + 1);
for (int n = 1; n <= maximum; n++) {
rooted_tree_egf[n] =
Mint(n).pow(n - 1) * combinations.inverse_factorial(n);
}
Fps one_minus_rooted(maximum + 1);
one_minus_rooted[0] = 1;
for (int i = 1; i <= maximum; i++) {
one_minus_rooted[i] = Mint(0) - rooted_tree_egf[i];
}
Fps egf = one_minus_rooted.log(maximum + 1);
for (Mint& coefficient : egf) coefficient = Mint(0) - coefficient;
egf -= rooted_tree_egf;
egf -= ((rooted_tree_egf * rooted_tree_egf).pre(maximum + 1) *
graph_counting_detail::inverse_two<Mint>());
egf *= graph_counting_detail::inverse_two<Mint>();
return graph_counting_detail::from_egf(std::move(egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_connected_eulerian_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
std::vector<Mint> all_even(maximum + 1);
all_even[0] = 1;
if (maximum >= 1) {
std::vector<Mint> shifted = count_labeled_undirected_graphs<Mint>(maximum - 1);
for (int n = 1; n <= maximum; n++) all_even[n] = shifted[n - 1];
}
graph_counting_detail::Fps<Mint> egf =
graph_counting_detail::to_egf(std::move(all_even), combinations);
egf = egf.log(maximum + 1);
return graph_counting_detail::from_egf(std::move(egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_connected_bipartite_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
graph_counting_detail::Fps<Mint> egf =
graph_counting_detail::colored_bipartite_egf(maximum, combinations).log(maximum + 1);
egf *= graph_counting_detail::inverse_two<Mint>();
return graph_counting_detail::from_egf(std::move(egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_bipartite_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
std::optional<graph_counting_detail::Fps<Mint>> egf =
graph_counting_detail::colored_bipartite_egf(maximum, combinations).sqrt(maximum + 1);
assert(egf.has_value());
return graph_counting_detail::from_egf(std::move(*egf), combinations);
}
template <class Mint>
std::vector<Mint> count_labeled_directed_graphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
std::vector<Mint> result(maximum + 1);
result[0] = 1;
Mint multiplier = 1;
const Mint four = 4;
for (int n = 1; n <= maximum; n++) {
result[n] = result[n - 1] * multiplier;
multiplier *= four;
}
return result;
}
template <class Mint>
std::vector<Mint> count_labeled_dags(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
using Fps = graph_counting_detail::Fps<Mint>;
Fps denominator(maximum + 1);
Mint multiplier = 1;
const Mint half = graph_counting_detail::inverse_two<Mint>();
denominator[0] = 1;
for (int n = 1; n <= maximum; n++) {
denominator[n] = denominator[n - 1] * multiplier;
multiplier *= half;
}
for (int n = 0; n <= maximum; n++) {
denominator[n] *= combinations.inverse_factorial(n);
if (n & 1) denominator[n] = Mint(0) - denominator[n];
}
Fps egf = denominator.inv(maximum + 1);
std::vector<Mint> edge_powers = graph_counting_detail::two_to_binom2<Mint>(maximum);
for (int n = 0; n <= maximum; n++) {
egf[n] *= combinations.factorial(n) * edge_powers[n];
}
return egf;
}
template <class Mint>
std::vector<Mint> count_labeled_strongly_connected_digraphs(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
graph_counting_detail::Fps<Mint> egf(maximum + 1);
std::vector<Mint> edge_powers = graph_counting_detail::two_to_binom2<Mint>(maximum);
for (int n = 0; n <= maximum; n++) {
egf[n] = edge_powers[n] * combinations.inverse_factorial(n);
}
egf = egf.inv(maximum + 1);
for (int n = 0; n <= maximum; n++) egf[n] *= edge_powers[n];
egf = egf.log(maximum + 1);
for (int n = 0; n <= maximum; n++) {
egf[n] = Mint(0) - egf[n] * combinations.factorial(n);
}
return egf;
}
template <class Mint>
std::vector<Mint> count_labeled_tournaments(int maximum) {
return count_labeled_undirected_graphs<Mint>(maximum);
}
template <class Mint>
std::vector<Mint> count_labeled_strongly_connected_tournaments(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
math::Combinatorics<Mint> combinations(maximum);
graph_counting_detail::Fps<Mint> egf =
graph_counting_detail::to_egf(count_labeled_tournaments<Mint>(maximum), combinations);
egf = egf.inv(maximum + 1);
if (!egf.empty()) egf[0] = 0;
for (int n = 0; n <= maximum; n++) {
egf[n] = Mint(0) - egf[n] * combinations.factorial(n);
}
return egf;
}
template <class Mint>
std::vector<Mint> count_unlabeled_rooted_trees(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
std::vector<Mint> result(maximum + 1);
if (maximum == 0) return result;
std::vector<Mint> divisor_sum(maximum + 1);
result[1] = 1;
for (int multiple = 1; multiple <= maximum; multiple++) {
divisor_sum[multiple] += result[1];
}
for (int n = 1; n < maximum; n++) {
Mint sum = 0;
for (int i = 1; i <= n; i++) {
sum += divisor_sum[i] * result[n - i + 1];
}
result[n + 1] = sum / Mint(n);
const int size = n + 1;
const Mint contribution = Mint(size) * result[size];
for (int multiple = size; multiple <= maximum; multiple += size) {
divisor_sum[multiple] += contribution;
}
}
return result;
}
template <class Mint>
std::vector<Mint> count_unlabeled_trees(int maximum) {
graph_counting_detail::assert_maximum<Mint>(maximum);
using Fps = graph_counting_detail::Fps<Mint>;
Fps rooted = count_unlabeled_rooted_trees<Mint>(maximum);
Fps rooted_square = (rooted * rooted).pre(maximum + 1);
const Mint half = graph_counting_detail::inverse_two<Mint>();
std::vector<Mint> result(maximum + 1);
for (int n = 1; n <= maximum; n++) {
result[n] = rooted[n] - rooted_square[n] * half;
if ((n & 1) == 0) result[n] += rooted[n / 2] * half;
}
return result;
}
} // namespace graph
} // namespace m1une
#line 11 "verify/graph/graph_counting.test.cpp"
using mint = m1une::math::modint998244353;
void assert_prefix(const std::vector<mint>& actual, const std::vector<long long>& expected) {
assert(int(actual.size()) >= int(expected.size()));
for (int i = 0; i < int(expected.size()); i++) {
assert(actual[i] == mint(expected[i]));
}
}
struct Dsu {
std::vector<int> parent;
explicit Dsu(int n) : parent(n) {
std::iota(parent.begin(), parent.end(), 0);
}
int leader(int v) {
if (parent[v] == v) return v;
return parent[v] = leader(parent[v]);
}
bool merge(int a, int b) {
a = leader(a);
b = leader(b);
if (a == b) return false;
parent[b] = a;
return true;
}
};
struct UndirectedBrute {
long long all = 0;
long long connected = 0;
long long tree = 0;
long long forest = 0;
long long unicyclic_connected = 0;
long long connected_eulerian = 0;
long long bipartite = 0;
long long connected_bipartite = 0;
};
bool is_bipartite(int n, const std::vector<std::vector<int>>& adjacency) {
std::vector<int> color(n, -1);
for (int start = 0; start < n; start++) {
if (color[start] != -1) continue;
color[start] = 0;
std::vector<int> stack;
stack.push_back(start);
while (!stack.empty()) {
int v = stack.back();
stack.pop_back();
for (int to : adjacency[v]) {
if (color[to] == -1) {
color[to] = color[v] ^ 1;
stack.push_back(to);
} else if (color[to] == color[v]) {
return false;
}
}
}
}
return true;
}
UndirectedBrute brute_undirected(int n) {
std::vector<std::pair<int, int>> edges;
for (int i = 0; i < n; i++) {
for (int j = i + 1; j < n; j++) edges.emplace_back(i, j);
}
UndirectedBrute result;
const int total = 1 << int(edges.size());
for (int mask = 0; mask < total; mask++) {
result.all++;
Dsu dsu(n);
std::vector<int> degree(n);
std::vector<std::vector<int>> adjacency(n);
bool acyclic = true;
int edge_count = 0;
for (int bit = 0; bit < int(edges.size()); bit++) {
if (((mask >> bit) & 1) == 0) continue;
const auto [u, v] = edges[bit];
edge_count++;
degree[u]++;
degree[v]++;
adjacency[u].push_back(v);
adjacency[v].push_back(u);
if (!dsu.merge(u, v)) acyclic = false;
}
int components = n;
for (int i = 0; i < n; i++) {
if (dsu.leader(i) != i) components--;
}
const bool connected = n > 0 && components == 1;
const bool bipartite = is_bipartite(n, adjacency);
const bool all_even =
std::all_of(degree.begin(), degree.end(), [](int x) { return x % 2 == 0; });
if (connected) result.connected++;
if (n > 0 && connected && edge_count == n - 1) result.tree++;
if (acyclic) result.forest++;
if (connected && edge_count == n) result.unicyclic_connected++;
if (connected && all_even) result.connected_eulerian++;
if (bipartite) result.bipartite++;
if (connected && bipartite) result.connected_bipartite++;
}
return result;
}
struct DirectedBrute {
long long all = 0;
long long dag = 0;
long long strongly_connected = 0;
};
DirectedBrute brute_directed(int n) {
std::vector<std::pair<int, int>> edges;
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
if (i != j) edges.emplace_back(i, j);
}
}
DirectedBrute result;
const int total = 1 << int(edges.size());
for (int mask = 0; mask < total; mask++) {
result.all++;
std::vector<std::vector<int>> reach(n, std::vector<int>(n));
for (int bit = 0; bit < int(edges.size()); bit++) {
if (((mask >> bit) & 1) == 0) continue;
const auto [from, to] = edges[bit];
reach[from][to] = 1;
}
for (int k = 0; k < n; k++) {
for (int i = 0; i < n; i++) {
if (!reach[i][k]) continue;
for (int j = 0; j < n; j++) reach[i][j] = reach[i][j] || reach[k][j];
}
}
bool dag = true;
for (int i = 0; i < n; i++) dag = dag && !reach[i][i];
bool strongly_connected = n > 0;
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
if (i != j && !reach[i][j]) strongly_connected = false;
}
}
if (dag) result.dag++;
if (strongly_connected) result.strongly_connected++;
}
return result;
}
struct TournamentBrute {
long long all = 0;
long long strongly_connected = 0;
};
TournamentBrute brute_tournament(int n) {
std::vector<std::pair<int, int>> pairs;
for (int i = 0; i < n; i++) {
for (int j = i + 1; j < n; j++) pairs.emplace_back(i, j);
}
TournamentBrute result;
const int total = 1 << int(pairs.size());
for (int mask = 0; mask < total; mask++) {
result.all++;
std::vector<std::vector<int>> reach(n, std::vector<int>(n));
for (int bit = 0; bit < int(pairs.size()); bit++) {
const auto [u, v] = pairs[bit];
if ((mask >> bit) & 1) {
reach[u][v] = 1;
} else {
reach[v][u] = 1;
}
}
for (int k = 0; k < n; k++) {
for (int i = 0; i < n; i++) {
if (!reach[i][k]) continue;
for (int j = 0; j < n; j++) reach[i][j] = reach[i][j] || reach[k][j];
}
}
bool strongly_connected = n > 0;
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
if (i != j && !reach[i][j]) strongly_connected = false;
}
}
if (strongly_connected) result.strongly_connected++;
}
return result;
}
void test_undirected_counts() {
const int maximum = 5;
const auto all = m1une::graph::count_labeled_undirected_graphs<mint>(maximum);
const auto connected = m1une::graph::count_labeled_connected_graphs<mint>(maximum);
const auto trees = m1une::graph::count_labeled_trees<mint>(maximum);
const auto forests = m1une::graph::count_labeled_forests<mint>(maximum);
const auto unicyclic =
m1une::graph::count_labeled_unicyclic_connected_graphs<mint>(maximum);
const auto eulerian =
m1une::graph::count_labeled_connected_eulerian_graphs<mint>(maximum);
const auto bipartite = m1une::graph::count_labeled_bipartite_graphs<mint>(maximum);
const auto connected_bipartite =
m1une::graph::count_labeled_connected_bipartite_graphs<mint>(maximum);
for (int n = 0; n <= maximum; n++) {
const UndirectedBrute brute = brute_undirected(n);
assert(all[n] == mint(brute.all));
assert(connected[n] == mint(brute.connected));
assert(trees[n] == mint(brute.tree));
assert(forests[n] == mint(brute.forest));
assert(unicyclic[n] == mint(brute.unicyclic_connected));
assert(eulerian[n] == mint(brute.connected_eulerian));
assert(bipartite[n] == mint(brute.bipartite));
assert(connected_bipartite[n] == mint(brute.connected_bipartite));
}
assert_prefix(connected, std::vector<long long>{0, 1, 1, 4, 38, 728});
assert_prefix(forests, std::vector<long long>{1, 1, 2, 7, 38, 291});
assert_prefix(unicyclic, std::vector<long long>{0, 0, 0, 1, 15, 222});
assert_prefix(eulerian, std::vector<long long>{0, 1, 0, 1, 3, 38});
assert_prefix(bipartite, std::vector<long long>{1, 1, 2, 7, 41, 376});
assert_prefix(connected_bipartite, std::vector<long long>{0, 1, 1, 3, 19, 195});
}
void test_directed_counts() {
const int maximum = 4;
const auto all = m1une::graph::count_labeled_directed_graphs<mint>(maximum);
const auto dags = m1une::graph::count_labeled_dags<mint>(maximum);
const auto strongly_connected =
m1une::graph::count_labeled_strongly_connected_digraphs<mint>(maximum);
for (int n = 0; n <= maximum; n++) {
const DirectedBrute brute = brute_directed(n);
assert(all[n] == mint(brute.all));
assert(dags[n] == mint(brute.dag));
assert(strongly_connected[n] == mint(brute.strongly_connected));
}
assert_prefix(all, std::vector<long long>{1, 1, 4, 64, 4096});
assert_prefix(dags, std::vector<long long>{1, 1, 3, 25, 543});
assert_prefix(strongly_connected, std::vector<long long>{0, 1, 1, 18, 1606});
}
void test_tournament_counts() {
const int maximum = 5;
const auto all = m1une::graph::count_labeled_tournaments<mint>(maximum);
const auto strongly_connected =
m1une::graph::count_labeled_strongly_connected_tournaments<mint>(maximum);
for (int n = 0; n <= maximum; n++) {
const TournamentBrute brute = brute_tournament(n);
assert(all[n] == mint(brute.all));
assert(strongly_connected[n] == mint(brute.strongly_connected));
}
assert_prefix(all, std::vector<long long>{1, 1, 2, 8, 64, 1024});
assert_prefix(strongly_connected, std::vector<long long>{0, 1, 0, 2, 24, 544});
}
void test_unlabeled_trees() {
const auto rooted = m1une::graph::count_unlabeled_rooted_trees<mint>(12);
const auto unrooted = m1une::graph::count_unlabeled_trees<mint>(12);
assert_prefix(rooted, std::vector<long long>{0, 1, 1, 2, 4, 9, 20, 48, 115});
assert_prefix(unrooted, std::vector<long long>{0, 1, 1, 1, 2, 3, 6, 11, 23, 47, 106});
}
int main() {
m1une::utilities::FastInput fast_input;
m1une::utilities::FastOutput fast_output;
test_undirected_counts();
test_directed_counts();
test_tournament_counts();
test_unlabeled_trees();
long long a, b;
fast_input >> a >> b;
fast_output << a + b << '\n';
return 0;
}
verify/graph/graph_counting.test.cpp