1. Allocating

Refrain from declaring big arrays as local variables, as it will often cause you to run out of stack space and fail with a Runtime Error.

Instead of doing:

int main()
{
    int N;
    scanf("%d", &N);
    int arr[N];
    for(int i = 0; i < N; i++) scanf("%d", &arr[i]);
}

consider:

int arr[100001];
int main()
{
    int N;
    scanf("%d", &N);
    for(int i = 0; i < N; i++) scanf("%d", &arr[i]);
}

Declaring big arrays in global scope is a much safer approach as long as you know the maximum bound of ~N~ (and almost all problems give you upper bounds). Be wary of out of bounds array indices, though.

2. Input and Output

It is recommended for C++ users to use C-style input and output, namely scanf and printf instead of cin and cout for performance reasons.

If you must use cin and cout, you can put these two lines of code at the top of your main function:

int main()
{
    cin.sync_with_stdio(0);
    cin.tie(0);
    ...
}

to speed up the cin stream. This will unsync cin with scanf and cout. Note that you should not use scanf after unsyncing with stdio.

Additionally, you should not use endl, but rather \n to output newlines when flushing is not required. endl's flushing behavior can potentially cause your program to receive TLE instead of AC.

Finally, if the problem only requires unsigned integral data types to be read, you can prepend this macro to the top of your source:

#define scan(x) do{while((x=getchar())<'0'); for(x-='0'; '0'<=(_=getchar()); x=(x<<3)+(x<<1)+_-'0');}while(0)
char _;

Instead of std::cin >> n, or scanf("%d", &n), you would use scan(n).

3. int, long, and long long

On the judge, int is 32-bit, long long is 64-bit, and long can be either 32- or 64-bit depending on which judge your submission is graded on. Therefore, it is recommended to use either int or long long, but not long.