关于迭代加深，个人认为和dfs的区别在于如果知道答案的深度可以用dfs，相反如果不知道深度的话可以考虑用迭代加深。

#pragma GCC optimize(2)
#include<bits/stdc++.h>
using namespace std;
typedef long long ll;
const int maxn = 1e5+5;
int Case = 1;
ll n, m;
ll up;
ll res[maxn], v[maxn];
ll ask_min(ll a, ll b) {
	for(ll i = 2;; i++) {
		if(i*a >= b) return i;
	}
}

ll gcd(ll a, ll b) {
	return (b==0)?a:gcd(b, a%b);
}

bool better(int d) {
	for(int i = d; i >= 1; i--) {
		if(res[i] != v[i])return res[i] == -1||v[i]<res[i];
	}
	return false;
}

bool dfs(ll dep, ll mi, ll a, ll b) {
	if(dep == up) {
		if(b%a) return false;
		v[dep] = b/a;
		if(better(dep)) {
			for(int i = 1; i <= dep; i++) {
				res[i] = v[i];
			}
		}
		return true;
	}
	bool ok = false;
	mi = max(mi, ask_min(a, b));
	for(ll i = mi;; i++) {
		if(b*(up-dep+1) <= i*a) break;
		v[dep] = i;
		ll b2 = b*i, a2 = a*i-b, g = gcd(a2, b2);
		if(dfs(dep+1, i, a2/g, b2/g)) ok = true;
	}
	return ok;
}

void solve() {
	scanf("%lld%lld", &n, &m);
	if(m%n == 0) {
		printf("%lld\n", m/n);
		return;
	}
	ll x = ask_min(n, m);
	memset(res, -1, sizeof(res));
	for(up = 2;; up++) {
		if(dfs(1, x, n, m)) break;
	}
	for(int i = 1; i <= up; i++) 
		printf("%lld%c", res[i], i == up?'\n':' ');
    return;
}

int main() {
    while(Case--) {
        solve();
    }
    return 0;
}