#pragma GCC optimize("O3")
#include <bits/stdc++.h>
using namespace std;
using i64 = long long;
#define L(i, j, k) for (int i = (j); i <= (k); ++i)
#define R(i, j, k) for (int i = (j); i >= (k); --i)
#define all(x) x.begin(), x.end()
const int nax = 100050;
const int kax = 204;
int N, K;
int a[nax];
int from[kax][nax];
i64 dp[2][nax];
i64 pre[nax];
int q[nax];
int L = 1, R = 1;
int main() {
ios::sync_with_stdio(false);
cin.tie(nullptr); cout.tie(nullptr);
cin >> N >> K;
L(i, 1, N) {
cin >> a[i];
pre[i] = pre[i - 1] + a[i];
}
L(i, 0, N) dp[0][i] = 0;
L(k, 1, K) {
q[R++] = 0;
int nw = k & 1;
int pv = nw ^ 1;
auto Y = [&](int x) { return dp[pv][x]; };
auto X = [&](int x) { return pre[x]; };
auto dY = [&](int x, int y) { return Y(x) - Y(y); };
auto dX = [&](int x, int y) { return X(x) - X(y); };
auto DP = [&](int x, int y) { return dp[pv][y] + (pre[x] - pre[y]) * (pre[N] - pre[x]); };
L(i, 1, N) {
while (R - L > 1 && dY(q[L + 1], q[L]) >= (pre[N] - pre[i]) * dX(q[L + 1], q[L])) L++;
int j = q[L];
dp[nw][i] = DP(i, j);
from[k][i] = j;
while (R - L > 1 && dY(i, q[R - 1]) * dX(q[R - 1], q[R - 2]) >= dY(q[R - 1], q[R - 2]) * dX(i, q[R - 1])) R--;
q[R++] = i;
}
L(i, 1, N) dp[pv][i] = dp[nw][i];
L = R = 1;
}
i64 mx = -1;
int start = -1;
L(i, 1, N) {
if (mx < dp[K & 1][i]) {
mx = dp[K & 1][i];
start = i;
}
}
cout << mx << '\n';
L(i, 0, K - 1) {
cout << start << ' ';
start = from[K - i][start];
}
}
/*
contribution = sum in segment * sum after the segment
dp[i][j] = considered until i and have j splits
for all p < i
dp[i][j] = max(dp[p][j - 1] + (pre[i] - pre[p]) * (pre[N] - pre[i]))
dp[i][j] = max(dp[p][j - 1] + - pre[p] * (pre[N] - pre[i])) + (pre[i]) * (pre[N] - pre[i])
m = -pre[p]
x = pre[N] - pre[i]
c = dp[p][j - 1]
CHT:
if x is better than y:
dp[x][j - 1] - pre[x] * (pre[N] - pre[i]) >= dp[y][j - 1] - pre[y] * (pre[N] - pre[i])
dp[x][j - 1] - dp[y][j - 1] >= (pre[x] - pre[y]) * (pre[N] - pre[i])
Y(x) = dp[x]
X(x) = pre[x]
Y(x) - (y) >= (X(x) - X(y)) * (pre[N] - pre[i])
slope(i, q[r]) <= slope(q[r], q[r - 1]) pop
*/
