[模板]BM优化线性递推

又水了一篇博客=w=

namespace BM{
	vector<int> h[N];
	int cnt,fail[N],d[N],an[N],mx,k,f[N],g[N],p[N],trs[30][N];

	void mult(int *a,int *b,int *c) {
		fo(i,0,2*k-2) g[i]=0;
		fo(i,0,k-1) fo(j,0,k-1) (g[i+j]+=(ll)a[i]*b[j]%Mo)%=Mo;
		fd(i,2*k-2,k)
			fd(j,k-1,0) 
				(g[i-k+j]+=Mo-(ll)g[i]*p[j]%Mo)%=Mo;
		fo(i,0,k-1) c[i]=g[i];
	}

	void work(int n,int *a) {
		h[cnt=mx=0].clear();
		fo(i,1,n) {
			int now=-a[i];
			for(unsigned j=0;j<h[cnt].size();j++) (now+=(ll)a[i-j-1]*h[cnt][j]%Mo)%=Mo;
			d[i]=now;if (!now) continue;
			fail[cnt]=i;
			if (!cnt) {
				h[++cnt].clear();
				h[cnt].resize(i);
				continue;
			}
			vector<int> r;r.resize(i-fail[mx]-1);
			int mul=Mo-(ll)now*pwr(d[fail[mx]],Mo-2)%Mo;
			r.pb(-mul);for(unsigned j=0;j<h[mx].size();j++) r.pb((ll)h[mx][j]*mul%Mo);
			if (r.size()<h[cnt].size()) r.resize(h[cnt].size());
			for(unsigned j=0;j<h[cnt].size();j++) (r[j]+=h[cnt][j])%=Mo;
			if (i-fail[mx]+h[mx].size()>=h[cnt].size()) mx=cnt;
			h[++cnt]=r;
		}
		k=h[cnt].size();
		p[k]=1;fo(i,1,k) p[k-i]=Mo-h[cnt][i-1];
		if (k==1) trs[0][0]=Mo-p[0];
		else trs[0][1]=1;
		fo(i,1,k<<1) f[i]=a[i];
		fo(i,1,29) mult(trs[i-1],trs[i-1],trs[i]);
	}

	int calc(int n) {
		if (n<=k) return f[n];n-=k;
		fo(i,0,k-1) an[i]=0;an[0]=1;
		fo(i,0,29) if (n>>i&1) mult(an,trs[i],an);
		int ans=0;
		fo(i,0,k-1) (ans+=(ll)an[i]*f[i+k]%Mo)%=Mo;
		return ans;
	}
}