#include <bits/stdc++.h>
#include <atcoder/all>
// url 
#define rep(i,a,b) for(int i=a;i<b;i++)
#define rrep(i,a,b) for(int i=b-1;i>=a;i--)
#define all(x) (x).begin(),(x).end()
#define pb(x) push_back(x);
template<class T>bool chmax(T &a, const T &b) { if (a<b) { a = b; return 1; } return 0; }
template<class T>bool chmin(T &a, const T &b) { if (b<a) { a = b; return 1; } return 0; }
typedef long long ll;
typedef long double lld;
using namespace std;
using namespace atcoder;
using mint = static_modint<998244353>;
const ll mod=998244353;
// using mint = static_modint<1000000007>;
//const ll mod=1e9+7;
int dx[4]={1,0,-1,0};
int dy[4]={0,1,0,-1};
const string zton="0123456789";
const string atoz="abcdefghijklmnopqrstuvwxyz";
const string ATOZ="ABCDEFGHIJKLMNOPQRSTUVWXYZ";
const ll inf=(1ll<<60);
// const int inf=(1<<30);
lld dist(lld x1,lld x2,lld y1,lld y2){
    lld res=(x1-x2)*(x1-x2)+(y1-y2)*(y1-y2);
    res=sqrt(abs(res));
    return res;
}
lld arg(lld x,lld y){
    const lld eps=1e-8;
    lld res=0;
    if(abs(x)+abs(y)<=eps)return 0.0;
    else if(abs(x)<=eps){
        if(y>=0.0)return (M_PI/2);
        else return (M_PI/2+M_PI);
    }
    else if(abs(y)<=eps){
        if(x>=0.0)return 0.0;
        else return M_PI;
    }
    res=atan2(abs(y),abs(x));
    if(x<=0&&y>=0)res=(M_PI-res);
    else if(x<=0&&y<=0)res+=(M_PI);
    else if(x>=0&&y<=0)res=(M_PI*2-res);
    return res;
}
ll gcd(ll a,ll b){
    if(a==0||b==0)return a+b;
    ll r;
    r=a%b;
    if(r==0){
        return b;
    }
    else{
        return gcd(b,r);
    }
}
std::ostream &operator<<(std::ostream &dest, __int128_t value) {
    std::ostream::sentry s(dest);
    if (s) {
        __uint128_t tmp = value < 0 ? -value : value;
        char buffer[128];
        char *d = std::end(buffer);
        do {
            --d;
            *d = "0123456789"[tmp % 10];
            tmp /= 10;
        } while (tmp != 0);
        if (value < 0) {
            --d;
            *d = '-';
        }
        int len = std::end(buffer) - d;
        if (dest.rdbuf()->sputn(d, len) != len) {
            dest.setstate(std::ios_base::badbit);
        }
    }
    return dest;
}
__int128 parse128(string &s) {
    __int128 ret = 0;
    for (int i = 0; i < s.length(); i++){
        if ('0' <= s[i] && s[i] <= '9'){
            ret = 10 * ret + s[i] - '0';
        }
    }
    return ret;
}
using l3 = __int128;
const int kaijouSize=501010;
ll inv_[kaijouSize];
mint inv[kaijouSize];
mint kaijou[kaijouSize];
mint kaijou_inv[kaijouSize];
void InitKaijou(){
    inv_[0]=1;
    inv_[1]=1;
    inv[0]=1;
    inv[1]=1;
    rep(i,2,kaijouSize) inv_[i]=inv_[mod%i]*(mod-mod/i)%mod;
    rep(i,2,kaijouSize) inv[i]=inv_[i];
    kaijou[0]=1;
    rep(i,1,kaijouSize){
        kaijou[i]=kaijou[i-1]*1ll*i;
    }
    kaijou_inv[0]=1ll;
    rep(i,1,kaijouSize){
        kaijou_inv[i]=kaijou_inv[i-1]*1ll*(inv[i]);
    }
}
mint nck(int n,int k){
    if(n<k)return 0;
    else return (((kaijou[n]*kaijou_inv[k]))*kaijou_inv[n-k]);
}
mint npk(int n,int k){
    if(n<k)return 0;
    return (kaijou[n]*kaijou_inv[n-k]);
}    
mint a[kaijouSize];
mint solve1(int N,int P){
    mint not_ans=0;
    rep(i,0,N){
        if((N-i)%P==0){
            mint tmp=nck(N,i);
            int ex=N-i;
            int k=ex/P;
            // cout << i << " " << k << endl;
            tmp*=kaijou[ex];
            tmp/=kaijou[k];
            tmp/=pow_mod(P,k,mod);
            not_ans+=tmp;
        }
    }
    mint ans=kaijou[N]-not_ans-1;
    return ans;
}

mint solve2(int N,int P){
    vector<int> A(N);rep(i,0,N)A[i]=i;
    int ans=0;
    do{
        vector<int> now(N);rep(i,0,N)now[i]=i;
        rep(i,0,P){
            vector<int> next(N);
            rep(j,0,N){
                next[A[j]]=now[j];
            }
            now=next;
        }
        bool ok=true;
        rep(i,0,N)if(now[i]!=i)ok=false;
        if(ok==false)ans++;

    }while(next_permutation(all(A)));
    return ans;
}

int main(void){
    ios::sync_with_stdio(false);
    cin.tie(nullptr);
    InitKaijou();
    a[0]=0;
    a[1]=0;
    a[2]=1;
    a[3]=2;
    rep(i,4,kaijouSize){
        a[i]=(a[i-1]+a[i-2])*(i-1);
    }
    int N,p;cin >> N >> p;
    cout << solve1(N,p).val() << endl;



}