//#pragma GCC optimize("Ofast")
//#pragma GCC optimize("unroll-loops")
//#pragma GCC target("avx,avx2,fma")
#include <bits/stdc++.h>
#define sze(x) (ll) x.size()
#define idx(x, a) get<x>(a)
#define LID(x) (x << 1LL)
#define RID(x) (x << 1LL) + 1LL
#define ID(x) (x + MAXN)
#define CONV(x) (x - MAXN)
#define countbit(x) __builtin_popcountll(x)
#define pb push_back
#define mp make_pair
#define fi first
#define se second
using namespace std;
typedef long long ll;
typedef unsigned long long ull;
typedef long double ld;
typedef pair<ll, ll> pi;
mt19937_64 rng(chrono::steady_clock::now().time_since_epoch().count());
ll rand(ll l, ll r)
{
return uniform_int_distribution<ll>(l, r)(rng);
}
const ll MAXN = 100'003;
const ll MAXS = 2003;
const ll INF = 1e18;
const ld EPS = 1e-9;
ll n, s;
ll dp0[MAXS], dp1[MAXS];
vector<pi> a[MAXN];
void solve()
{
fill(dp0, dp0 + s + 1, 0);
for (ll wi = 1; wi <= s; ++wi)
{
if (a[wi].empty()) continue;
sort(a[wi].begin(), a[wi].end(), greater<pi>());
vector<ll> sumval((s / wi) + 2);
for (ll i = 1, r = 0; i < (ll)sumval.size(); ++i)
{
sumval[i] = sumval[i - 1] + a[wi][r].fi * max(0LL, min(1LL, a[wi][r].se));
--a[wi][r].se;
while (r + 1 < (ll)a[wi].size() && a[wi][r].se == 0) ++r;
}
for (ll sn = 0; sn <= s; ++sn)
{
dp1[sn] = 0;
for (ll i = sn, cnt = 0; i >= 0; i -= wi, ++cnt)
{
dp1[sn] = max(dp1[sn], dp0[i] + sumval[cnt]);
}
}
// for (ll sn = 0; sn <= s; ++sn) cerr << dp1[sn] << " ";
for (ll sn = 0; sn <= s; ++sn) dp0[sn] = dp1[sn];
// cerr << endl;
}
cout << dp1[s];
}
int main()
{
ios_base::sync_with_stdio(0);
cin.tie(0);
#ifdef OFFLINE
freopen("input.inp", "r", stdin);
#endif
cin >> s >> n;
for (ll i = 1; i <= n; ++i)
{
ll vi, wi, ki; cin >> vi >> wi >> ki;
a[wi].pb({vi, ki});
}
solve();
}
