#include using namespace std; typedef signed long long ll; #define _P(...) (void)printf(__VA_ARGS__) #define FOR(x,to) for(x=0;x<(to);x++) #define FORR(x,arr) for(auto& x:arr) #define FORR2(x,y,arr) for(auto& [x,y]:arr) #define ALL(a) (a.begin()),(a.end()) #define ZERO(a) memset(a,0,sizeof(a)) #define MINUS(a) memset(a,0xff,sizeof(a)) template bool chmax(T &a, const T &b) { if(a bool chmin(T &a, const T &b) { if(a>b){a=b;return 1;}return 0;} //------------------------------------------------------- int N,M; const int mo=998244353; const int NUM_=400001; static ll fact[NUM_+1],factr[NUM_+1],inv[NUM_+1]; ll modpow(ll a, ll n = mo-2) { ll r=1; a%=mo; while(n) r=r*((n%2)?a:1)%mo,a=a*a%mo,n>>=1; return r; } ll comb(ll N_, ll C_) { if(C_<0 || C_>N_) return 0; return factr[C_]*fact[N_]%mo*factr[N_-C_]%mo; } ll hcomb(ll P_,ll Q_) { return (P_==0&&Q_==0)?1:comb(P_+Q_-1,Q_);} template using vec=vector; //using vec=valarray; template vec fft(vec v, bool rev=false) { int n=v.size(),i,j,m; for(int m=n; m>=2; m/=2) { T wn=modpow(5,(mo-1)/m); if(rev) wn=modpow(wn); for(i=0;i=mo) v[j1]-=mo; w=(ll)w*wn%mo; } } } for(i=0,j=1;j>1;k>(i^=k);k>>=1); if(i>j) swap(v[i],v[j]); } if(rev) { ll rv = modpow(n); FOR(i,n) v[i]=(ll)v[i]*rv%mo; } return v; } template vec MultPoly(vec P,vec Q,bool resize=false) { if(resize) { int maxind=0,pi=0,qi=0,i; int s=2; FOR(i,P.size()) if(norm(P[i])) pi=i; FOR(i,Q.size()) if(norm(Q[i])) qi=i; maxind=pi+qi+1; while(s*2 R(s*2); for(int x=0;x<=pi;x++) for(int y=0;y<=qi;y++) (R[x+y]+=P[x]*Q[y])%=mo; return R; } vec P2(s*2),Q2(s*2); FOR(i,pi+1) P2[i]=P[i]; FOR(i,qi+1) Q2[i]=Q[i]; swap(P,P2),swap(Q,Q2); } P=fft(P), Q=fft(Q); for(int i=0;i>N>>M; if(N==1||M==1) { cout< F,C={1}; FOR(i,N+1) F.push_back(factr[i]*factr[i]%mo); x=M-2; while(x) { if(x%2) { C=MultPoly(F,C,1); C.resize(N+1); } F=MultPoly(F,F,1); F.resize(N+1); x/=2; } ll ret=0; for(i=2;i<=N;i++) { ll a=comb(N,i)*comb(N,i)%mo; a=a*modpow(2,2*i-3)%mo*(i-1)%mo; a=a*fact[N-i]%mo*fact[N-i]%mo; a=a*C[N-i]%mo; ret+=a; } ret=ret%mo*comb(M,2)%mo; cout<