C11 Threads

Posted at — Aug 3, 2020 | Last Modified on — May 11, 2023

Introduction

this is a simple introduction to threading in C. I will be using inbuild C threading library threads.h which is a feature of C11 standard.

the advantages of using C11 threads over platform-specific solutions like pthread is obvious, code portability. write it once, compile it anywhere with the compiler that supports C11 std.

Counting to a Billion

Single-threaded

For the demonstration, we take a simple program the counts to 1 billion (1 billion x 5 times)


#include <stdio.h>
#include <stdlib.h>

void count_to_billion()
{
    int at = 0;
    for (int i = 0; i < 1000000000; i++)
    {
        at++;
    }

    printf("counted to 1 billion! at=%d\n", at);
}

int main()
{

    // count to 5 billion 5 times
    count_to_billion();
    count_to_billion();
    count_to_billion();
    count_to_billion();
    count_to_billion();

    return EXIT_SUCCESS;
}

let’s compile the program

gcc main.c

we use a Unix utility time to measure the execution time of the program.

run the compiled program with time as the prefix

time ./a.out

Output:


~/Documents/Projects/CWS » time ./a.out
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
./a.out  8.68s user 0.00s system 99% cpu 8.686 total

as we can see from the output it took 8.686 secs to complete

MultiThreading

from cppreference.com:

The type thrd_start_t is a typedef of int(*)(void*), which differs from the POSIX equivalent void*(*)(void*)


#include <stdio.h>
#include <threads.h>
#include <stdlib.h>

int count_to_billion(void* arg)
{
    int at = 0;
    for (int i = 0; i < 1000000000; i++)
    {
        at++;
    }

    printf("counted to 1 billion! at=%d\n", at);
}

int main()
{

    int n_threads = 5;

    thrd_t thread[n_threads];

    for (int i = 0; i < n_threads; i++)
    {
        thrd_create(&thread[i], count_to_billion, NULL);
    }

    for (int i = 0; i < n_threads; i++)
    {
        thrd_join(thread[i], NULL);
    }
    return EXIT_SUCCESS;
}

let’s compile our program and run it.

gcc main.c -lpthread

don’t forget to link it with thread library of your OS, in my case, it’s pthreads

~/Documents/Projects/CWS » time ./a.out
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
counted to 1 billion! at=1000000000
./a.out  8.92s user 0.00s system 497% cpu 1.794 total

Conclusion

as it’s quite evident from the execution time of both programs, Multi-threaded program is much faster.

1.794(MT) > 8.686(ST)

<stdin> |

My Thoughts, Trials and Adventures