C++ I/O Optimization for Competitive Programming — Fast Read, Fast Write, and More

#beginners #cpp #debugging #tutorial

I'm AlanWu. I spent way too long on Luogu wondering why my solution was TLE even though the algorithm was right. Then I learned about I/O.

The two lines that fix 90% of it

ios::sync_with_stdio(false);
cin.tie(nullptr);

Put these at the top of main. That's it. You just made cin about as fast as scanf.

What they do: the first line unties C++ streams from C streams (so they don't sync buffers). The second line unties cin from cout (so cout doesn't flush before every cin). Together, reading 1 million ints drops from ~2 seconds to ~0.3.

endl is a trap

cout << x << endl;   // flushes every time — slow
cout << x << '\n';   // doesn't flush — fast

endl does '\n' AND flush. You almost never need the flush part during a contest. Use '\n'.

When you need even more speed: fast read

For problems with 10^6+ numbers, cin with sync off is still not enough. You need integer parsing by hand:

inline int read() {
    int x = 0, f = 1;
    char ch = getchar();
    while (ch < '0' || ch > '9') {
        if (ch == '-') f = -1;
        ch = getchar();
    }
    while (ch >= '0' && ch <= '9') {
        x = x * 10 + (ch - '0');
        ch = getchar();
    }
    return x * f;
}

And a matching fast write:

inline void write(int x) {
    if (x < 0) putchar('-'), x = -x;
    if (x > 9) write(x / 10);
    putchar(x % 10 + '0');
}

Use these when every millisecond counts. Most problems don't need them.

freopen for local testing

Typing test cases by hand is slow. Use file redirection:

#ifdef LOCAL
    freopen("input.txt", "r", stdin);
    freopen("output.txt", "w", stdout);
#endif

Now cin reads from input.txt and cout writes to output.txt when you compile with -DLOCAL. On the judge, LOCAL isn't defined, so it reads from stdin like normal. You never need to change your code between local testing and submission.

Speed tier summary

Method	1 million ints	When to use
`cin >> x` (no optimization)	~2.0s	Never for contests
`cin >> x` (sync off)	~0.3s	Default for most problems
`scanf("%d", &x)`	~0.3s	Same tier as optimized cin
`read()` (custom)	~0.15s	10^6+ integers, strict time limits