C++ With the Command Line

Command Line Basics

General

Linux

Mac

Should be mostly the same as Linux ...

Open the Terminal application and familiarize yourself with some basic commands. Upgrade to zsh if you haven't already.

Windows

Installing GCC

USACO (and most contests) use GCC's g++ to compile and run your code. You'll need g++ specifically to use the #include <bits/stdc++.h> header file; see Running Code Locally for details.

On Linux

GCC is usually preinstalled on most Linux distros. You can check if it is installed with

whereis g++

If it is not preinstalled, you can probably install it using your distro's package manager.

On Windows

There are multiple ways to install C++ command line tools on windows, depending on your needs. To find more information, go to Running Code Locally with Windows.

On Mac

1. Install XCode command line tools.

   xcode-select --install
   

   If you previously installed these you may need to update them:

   softwareupdate --list # list updates
   softwareupdate -i -a # installs all updates

   After this step, clang should be installed (try running clang --version in Terminal).

2. Install Homebrew.

3. Install gcc with Homebrew.

   brew install gcc
   

   According to this if brew doesn't seem to finish for a long time then

   brew install gcc --force-bottle
   

   probably suffices.

4. You should be able to compile with g++-#, where # is the version number (e.g., 10). Running the following command

   g++-10 --version
   

   should display something like this:

   g++-10 (Homebrew GCC 10.2.0_2) 10.2.0
   Copyright (C) 2020 Free Software Foundation, Inc.
   This is free software; see the source for copying conditions.  There is NO
   warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
   

5. If you want to be able to compile with just g++, write a shell alias! Put the following lines into your shell's rc file (~/.bashrc if you use bash, and ~/.zshrc if you use zsh).

   alias g++=g++-10

   Once you do so, g++ --version should now output the same thing as g++-10 --version.

   Note: avoid overriding the system g++ with symlinking or hard-linking as that will almost surely cause problems. Don't worry if you don't know what those terms mean.

C++ with the Command Line

Basics of Compiling & Running

Consider a simple program such as the following, which we'll save in name.cpp.

#include <iostream>
using namespace std;

int main() {
	int x; cin >> x;
	cout << "FOUND " << x << "\n";
}

It's not hard to compile & run a C++ program. First, open up Powershell on Windows, Terminal on Mac, or your distro's terminal in Linux. We can compile name.cpp into an executable named name with the following command:

g++ name.cpp -o name

Then we can execute the program:

./name

If you type some integer and then press enter, then the program should produce output. We can write both of these commands in a single line:

g++ name.cpp -o name && ./name

Note that && ensures that ./name only runs if g++ name.cpp -o name finishes successfully.

Redirecting Input & Output

If you want to read standard input from inp.txt, use the following:

./name < inp.txt

If you want to write standard output to out.txt, then use the following:

./name > out.txt

They can also be used in conjunction, as shown below:

./name < inp.txt > out.txt

See Input & Output for how to do file input and output within the program.

Compiler Options (aka Flags)

Use compiler flags to change the way GCC compiles your code. Usually, we use something like the following in place of g++ name.cpp -o name:

g++ -std=c++17 -O2 name.cpp -o name -Wall

• -O2 tells g++ to compile your code to run more quickly while increasing compilation time (see here).
• -std=c++17 allows you to use features that were added to C++ in 2017. USACO currently uses -std=c++0x (aka C++11).
• -Wall checks your program for common errors. See Debugging for more information.

You should always compile with these flags.

Mac-Specific

Adding Shortcuts

Retyping the flags above can get tedious. You should define shortcuts so you don't need to type them every time!

First, create your .zshrc if it doesn't already exist.

touch ~/.zshrc

Open your .zshrc with a text editor.

open ~/.zshrc

or some text editor (ex. sublime text with subl).

subl ~/.zshrc

You can add aliases and functions here, such as the following to compile and run C++ on Mac.

co() { g++ -std=c++17 -O2 -o "${1%.*}" $1 -Wall; }
run() { co $1 && ./${1%.*} & fg; }

Now you can easily compile and run name.cpp from the command line with co name.cpp && ./name or run name.cpp. Note that all occurrences of $1 in the function are replaced with name.cpp, while ${1%.*} removes the file extension from $1 to produce name.

What is & fg for?

Let prog.cpp denote the following file:

#include <iostream>
#include <vector>
using namespace std;

int main() {
	vector<int> v;
	cout << v[-1];
}

According to the resource above, the & fg is necessary for getting zsh on Mac to display crash messages (such as segmentation fault). For example, consider the running the first prog.cpp above with run prog.cpp.

If & fg is removed from the run command above then the terminal displays no message at all. Leaving it in produces the following (ignore the first two lines):

[2] 30594
[2]  - running    ./${1%.*}
zsh: segmentation fault  ./${1%.*}

Measuring Time & Memory Usage

For example, suppose that prog.cpp consists of the following:

#include <bits/stdc++.h>
using namespace std;

const int BIG = 1e7;
int a[BIG];

int main() {
	int sum = 0;
	for (int i = 0; i < BIG; ++i) sum += a[i];
	cout << sum;
}

Then co prog.cpp && gtime -v ./prog gives the following:

  Command being timed: "./prog"
  User time (seconds): 0.01
  System time (seconds): 0.01
  Percent of CPU this job got: 11%
  Elapsed (wall clock) time (h:mm:ss or m:ss): 0:00.22
  Average shared text size (kbytes): 0
  Average unshared data size (kbytes): 0
  Average stack size (kbytes): 0
  Average total size (kbytes): 0
  Maximum resident set size (kbytes): 40216
  Average resident set size (kbytes): 0
  Major (requiring I/O) page faults: 91
  Minor (reclaiming a frame) page faults: 10088
  Voluntary context switches: 3
  Involuntary context switches: 38
  Swaps: 0
  File system inputs: 0
  File system outputs: 0
  Socket messages sent: 0
  Socket messages received: 0
  Signals delivered: 0
  Page size (bytes): 4096
  Exit status: 0

Note that $10^7$ integers require $4\cdot 10^7\cdot 10^{-3}\approx 40000$ kilobytes of memory, which is close to $40216$ in the above output as expected.

Adjusting Stack Size

Let A.cpp denote the following program:

#include <iostream>
using namespace std;

int res(int x) {
	if (x == 200000) return x;
	return res(x+1);
}

int main() {
	cout << res(0) << "\n";
}

If we compile and run this with g++ A.cpp -o A && ./A, this outputs 200000. However, changing 200000 to 300000 gives a segmentation fault. Similarly,

#include <iostream>
using namespace std;

int main() {
	int arr[2000000];
	cout << arr[0] << "\n";
}

runs, but changing 2000000 to 3000000 also gives a segmentation fault. This is because the stack size on Mac appears to be limited to 8 megabytes by default.

Method 1

ulimit -s 65532 will increase the stack size to about 64 MB. Unfortunately, this doesn't work for higher numbers.

Method 2

To get around this, we can pass a linker option. According to the manual for ld (enter man ld in Terminal), the option -stack_size size does the following:

Specifies the maximum stack size for the main thread in a program. Without this option a program has a 8MB stack. The argument size is a hexadecimal number with an optional leading 0x. The size should be a multiple of the architecture's page size (4KB or 16KB).

So including -Wl,-stack_size -Wl,0x10000000 as part of your compilation command will set the maximum stack size to $16^7$ bytes $\approx$ $256$ megabytes, which is usually sufficient. However, running the first program above with 200000 replaced by 1e7 still gives an error. In this case, you can further increase the maximum stack size (ex. by adding another 0 to the end).