#include <complex>
#include <vector>
#include <iostream>
using namespace std;
#define ll long long
#pragma GCC optimize("O3")
#pragma GCC optimize("Ofast")
#pragma GCC target("avx,avx2,fma")
#pragma GCC target("sse,sse2,sse3,ssse3,sse4,popcnt,abm,mmx,avx,tune=native")
#define usecppio ios::sync_with_stdio(0); cin.tie(0); cout.tie(0);
#define all(x)((x).begin()), ((x).end())
using pii = pair < int, int >;
#define int ll
#define INF 0x7f7f7f7f7f7f7f7f
const bool debug = false;
int N, M;
/* FFT Library : Originally Written by Myungwoo *
* https://blog.myungwoo.kr/54 *
* Nonrecursive, Bit-Flipping Trick *
* Several Modifications */
#define sz(v)((int)(v).size())
typedef complex<double> base;
typedef vector<int> vi;
typedef vector<base> vb;
const double PI = acos(-1);
void fft(vector < base >& a, bool inv = false)
{
int n = (int)a.size();
for (int i = 1, j = 0; i < n; i++)
{
int bit = n >> 1;
for (; j >= bit; bit >>= 1) j -= bit;
j += bit;
if (i < j)
swap(a[i], a[j]);
}
vector <base> root(n / 2);
double ang = 2 * acos(-1) / n * (inv ? -1 : 1);
for (int i = 0; i < n / 2; i++)
root[i] = base(cos(ang * i), sin(ang * i));
for (int idx = 2; idx <= n; idx <<= 1)
{
int step = n / idx;
for (int i = 0; i < n; i += idx)
{
for (int j = 0; j < idx / 2; j++)
{
base u = a[i + j], v = a[i + j + idx / 2] * root[step * j];
a[i + j] = u + v;
a[i + j + idx / 2] = u - v;
}
}
}
if (inv)
for (auto& i : a)
i /= n;
}
const int LARGE = 1e6;
/* FFT polynomial Multiplication with higher precision */
void multiply(const vi& a, const vi& b, vi& res)
{
vector < base > fa_big, fb_big;
vector < base > fa_small, fb_small;
int cut_val = sqrt(LARGE);
int n = 1;
while (n < max(sz(a), sz(b)))
n <<= 1;
fa_big.resize(n);
fa_small.resize(n);
fb_big.resize(n);
fb_small.resize(n);
for (int i = 0; i < sz(a); i++)
{
fa_big[i] = a[i] / cut_val;
fa_small[i] = a[i] % cut_val;
}
for (int i = 0; i < sz(b); i++)
{
fb_big[i] = b[i] / cut_val;
fb_small[i] = b[i] % cut_val;
}
fft(fa_big, 0), fft(fb_big, 0);
fft(fa_small, 0), fft(fb_small, 0);
vector<base> fa_big_2(all(fa_big));
vector<base> fa_small_2(all(fa_small));
for (int i = 0; i < n; i++)
{
fa_big_2[i] *= fb_big[i];
fa_big[i] *= fb_small[i];
fa_small[i] *= fb_small[i];
fa_small_2[i] *= fb_big[i];
}
fft(fa_small, 1);
fft(fa_small_2, 1);
fft(fa_big, 1);
fft(fa_big_2, 1);
res.resize(n);
for (int i = 0; i < n; i++)
{
int ss = (int64_t)round(fa_small[i].real());
int sb = (int64_t)round(fa_small_2[i].real());
int bs = (int64_t)round(fa_big[i].real());
int bb = (int64_t)round(fa_big_2[i].real());
res[i] = ss;
res[i] += (sb + bs) * cut_val;
res[i] += bb * cut_val * cut_val;
}
}
vector < int > f, g, res;
int32_t main()
{
usecppio cin >> N >> M;
f.resize(2 * (N + 1) + 5);
g.resize(2 * (M + 1) + 5);
for (int i = 0; i <= N; i++)
cin >> f[i];
for (int i = 0; i <= M; i++)
cin >> g[i];
multiply(f, g, res);
int t = 0;
for (int i = 0; i < res.size(); i++)
t ^= res[i];
cout << t << '\n';
}
Compilation message
tt.cpp: In function 'int32_t main()':
tt.cpp:147:23: warning: comparison of integer expressions of different signedness: 'long long int' and 'std::vector<long long int>::size_type' {aka 'long unsigned int'} [-Wsign-compare]
147 | for (int i = 0; i < res.size(); i++)
| ~~^~~~~~~~~~~~
tt.cpp:147:5: warning: this 'for' clause does not guard... [-Wmisleading-indentation]
147 | for (int i = 0; i < res.size(); i++)
| ^~~
tt.cpp:150:2: note: ...this statement, but the latter is misleadingly indented as if it were guarded by the 'for'
150 | cout << t << '\n';
| ^~~~
# |
결과 |
실행 시간 |
메모리 |
Grader output |
1 |
Correct |
0 ms |
212 KB |
Output is correct |
2 |
Correct |
1 ms |
212 KB |
Output is correct |
3 |
Correct |
1 ms |
324 KB |
Output is correct |
4 |
Correct |
1 ms |
468 KB |
Output is correct |
5 |
Correct |
2 ms |
468 KB |
Output is correct |
6 |
Correct |
2 ms |
468 KB |
Output is correct |
7 |
Correct |
2 ms |
468 KB |
Output is correct |
8 |
Correct |
2 ms |
452 KB |
Output is correct |
# |
결과 |
실행 시간 |
메모리 |
Grader output |
1 |
Correct |
21 ms |
4160 KB |
Output is correct |
2 |
Correct |
177 ms |
28996 KB |
Output is correct |
3 |
Correct |
189 ms |
29260 KB |
Output is correct |
4 |
Correct |
187 ms |
30144 KB |
Output is correct |
5 |
Correct |
188 ms |
30668 KB |
Output is correct |
6 |
Correct |
220 ms |
30308 KB |
Output is correct |
7 |
Correct |
183 ms |
30720 KB |
Output is correct |
8 |
Correct |
193 ms |
30928 KB |
Output is correct |
# |
결과 |
실행 시간 |
메모리 |
Grader output |
1 |
Correct |
458 ms |
60660 KB |
Output is correct |
2 |
Correct |
1244 ms |
119648 KB |
Output is correct |
3 |
Correct |
1267 ms |
121976 KB |
Output is correct |
4 |
Correct |
2939 ms |
251416 KB |
Output is correct |
5 |
Correct |
3100 ms |
246920 KB |
Output is correct |
6 |
Correct |
2914 ms |
255436 KB |
Output is correct |
7 |
Correct |
2954 ms |
260848 KB |
Output is correct |
8 |
Correct |
3014 ms |
259224 KB |
Output is correct |