[CTSC2010]性能优化 [混合基FFT]

传送门

先来复习一下普通的FFT

对于混合基FFT, 有

然后就可以分治了，rev 数组的预处理有一些奇葩

#include<bits/stdc++.h>
#define N 2000050
using namespace std;
typedef long long ll;
int read(){
	int x = 0, f = 1; char ch = 0;
	while(!isdigit(ch)){ ch = getchar(); if(ch == '-') f = -1;}
	while(isdigit(ch)) x = x*10 + (ch-'0'), ch = getchar();
	return x * f;
}
int Mod, G;
ll add(ll a, ll b){ return (a + b) % Mod;}
ll mul(ll a, ll b){ return (a * b) % Mod;}
ll power(ll a, ll b){ ll ans = 1;
	for(;b;b>>=1){ if(b&1) ans = mul(ans, a); a = mul(a, a);}
	return ans;
}
int n, c;
ll a[N], b[N], wn[N];
int rev[N], pcnt, prim[N];
ll tp[N];
void factor(int x){
	for(int i = 2; i*i <= x; i++){
		while(x % i == 0) prim[++pcnt] = i, x /= i;
	} if(x > 1) prim[++pcnt] = x; 
}
int getG(){
	factor(n);
	for(int i = 2;; i++){
		bool flag = true;
		for(int j = 1; j <= pcnt; j++) if(power(i, n/prim[j]) == 1){
			flag = false; break;
		} if(flag) return i;
	}
}
void Reverse(int *A){
	for(int i = pcnt, block = n; i >= 1; i--){
		for(int idx = 0, j = 0; j < n; j += block)
			for(int k = 0; k < prim[i]; k++)
				for(int l = 0; l < block; l += prim[i])
					tp[idx++] = A[j+k+l];
		for(int j = 0; j < n; j++) A[j] = tp[j];
		block /= prim[i];
	}
}
void DFT(ll *a, int typ){
	for(int i = 0; i < n; i++) tp[i] = a[rev[i]];
	for(int i = 0; i < n; i++) a[i] = tp[i];
	for(int i = 1, block = 1; i <= pcnt; i++){
		int mid = block; block *= prim[i]; 
		ll wi = wn[n/block];
		for(int j = 0; j < n; j++) tp[j] = 0;
		for(int j = 0; j < n; j += block){
			ll wk = 1;
			for(int k = 0; k < block; k++){
				for(int l = k%mid, w=1; l < block; l+=mid, w = mul(w,wk)){
					tp[j+k] = add(tp[j+k], mul(w, a[j+l]));
				} wk = mul(wk, wi);
			}
		}
		for(int j = 0; j < n; j++) a[j] = tp[j];
	}
	if(typ == -1){
		reverse(a+1, a+n);
		for(int i = 0, inv = power(n, Mod-2); i < n; i++) a[i] = mul(a[i], inv);
	}
}
int main(){
	n = read(), c = read(); Mod = n+1; G = getG();
	wn[0] = 1; wn[1] = G;
	for(int i = 2; i < n; i++) wn[i] = mul(wn[i-1], G);
	for(int i = 0; i < n; i++) a[i] = read();
	for(int i = 0; i < n; i++) b[i] = read();
	for(int i = 0; i < n; i++) rev[i] = i; Reverse(rev);
	DFT(a, 1); DFT(b, 1); 
	for(int i = 0; i < n; i++) a[i] = mul(a[i], power(b[i], c));
	DFT(a, -1);
	for(int i = 0; i < n; i++) cout << a[i] << '\n';
	return 0;
}