#pragma GCC optimize ("O3")
#pragma GCC target ("avx2")
#include <bits/stdc++.h>
using namespace std;
using ll = long long;
#define mid ((l+r) >> 1)
const int INF = 2e9, N = 5e5+1;
int n, u, d, s, mx[500500], t1[2002002], t2[2002002];
vector<array<int, 2>> v[500500];
void update(int t[], int id, int x, int node = 1, int l = 1, int r = N) {
if(id < l || r < id) return;
if(l == r) {
t[node] = max(t[node], x);
return;
}
if(id <= mid) update(t, id, x, node*2, l, mid);
else update(t, id, x, node*2+1, mid+1, r);
t[node] = max(t[node*2], t[node*2+1]);
}
int find(int t[], int nl, int nr, int node = 1, int l = 1, int r = N) {
if(nr < l || r < nl) return -INF;
if(nl <= l && r <= nr) return t[node];
int re = -INF;
if(nl <= mid) re = max(re, find(t, nl, nr, node*2, l, mid));
if(mid < nr) re = max(re, find(t, nl, nr, node*2+1, mid+1, r));
return re;
}
int main() {
cin.tie(0)->sync_with_stdio(0);
cin >> n >> u >> d >> s;
for(int i=1;i<=n;i++) {
int t, x, y; cin >> t >> x >> y;
v[t].push_back({x, y});
}
fill(t1, t1+4*N+1, -INF), fill(t2, t2+4*N+1, -INF);
fill(mx, mx+N+1, -INF);
update(t1, s, s*u), update(t2, s, -s*d);
for(int t=1;t<=N;t++) if(!v[t].empty()) {
sort(v[t].begin(), v[t].end());
for(auto [x, y] : v[t]) mx[x] = max(find(t1, 1, x-1) - x*u, x*d + find(t2, x+1, N)) + y;
for(auto [x, y] : v[t]) {
update(t1, x, x*u+mx[x]);
update(t2, x, mx[x]-x*d);
}
for(auto [x, y] : v[t]) {
auto k = find(t1, 1, x-1) - x*u + y;
mx[x] = max(mx[x], k), update(t1, x, x*u+k);
}
reverse(v[t].begin(), v[t].end());
for(auto [x, y] : v[t]) {
auto k = x*d + find(t2, x+1, N) + y;
mx[x] = max(mx[x], k), update(t2, x, k-x*d);
}
for(auto [x, y] : v[t]) {
update(t1, x, x*u+mx[x]);
update(t2, x, mx[x]-x*d);
}
}
int ans = 0;
for(int i=1;i<=N;i++) {
ans = max(ans, find(t1, 1, s-1) - s*u);
ans = max(ans, s*d + find(t2, s+1, N));
}
cout << ans << "\n";
}
# |
Verdict |
Execution time |
Memory |
Grader output |
1 |
Correct |
70 ms |
29784 KB |
Output is correct |
2 |
Correct |
74 ms |
30032 KB |
Output is correct |
3 |
Correct |
73 ms |
29832 KB |
Output is correct |
4 |
Correct |
68 ms |
29788 KB |
Output is correct |
5 |
Correct |
81 ms |
29968 KB |
Output is correct |
6 |
Correct |
136 ms |
30300 KB |
Output is correct |
7 |
Correct |
188 ms |
31160 KB |
Output is correct |
8 |
Correct |
255 ms |
32940 KB |
Output is correct |
9 |
Correct |
327 ms |
34520 KB |
Output is correct |
10 |
Correct |
508 ms |
39208 KB |
Output is correct |
11 |
Correct |
712 ms |
39204 KB |
Output is correct |
12 |
Correct |
845 ms |
42336 KB |
Output is correct |
13 |
Correct |
788 ms |
42336 KB |
Output is correct |
14 |
Correct |
998 ms |
45468 KB |
Output is correct |
15 |
Correct |
850 ms |
45468 KB |
Output is correct |
16 |
Correct |
985 ms |
45468 KB |
Output is correct |
17 |
Correct |
72 ms |
29784 KB |
Output is correct |
18 |
Correct |
62 ms |
29784 KB |
Output is correct |
19 |
Correct |
76 ms |
29784 KB |
Output is correct |
20 |
Correct |
74 ms |
29784 KB |
Output is correct |
21 |
Correct |
98 ms |
29788 KB |
Output is correct |
22 |
Correct |
79 ms |
30032 KB |
Output is correct |
23 |
Correct |
78 ms |
29784 KB |
Output is correct |
24 |
Correct |
83 ms |
29788 KB |
Output is correct |
25 |
Correct |
217 ms |
30800 KB |
Output is correct |
26 |
Correct |
384 ms |
32080 KB |
Output is correct |
27 |
Correct |
562 ms |
33084 KB |
Output is correct |
28 |
Correct |
671 ms |
33752 KB |
Output is correct |
29 |
Correct |
778 ms |
34688 KB |
Output is correct |
30 |
Correct |
763 ms |
34896 KB |
Output is correct |