Archive

Posts Tagged ‘run’

Starting with concurrency and a “Hello world” using pthreads

December 31, 2013 No comments


Nowadays is very common to see dual-core or quad-core CPUs, but there are still systems with one single core.
In this post, I want to about the very basis of concurrency, a brief introduction to this world, so some descriptions will be so so simple, but real world is a little bit more complex.

In multi-core or multi-CPU systems, the operating system (OS) distributes tasks within our CPUs, and we can run several tasks at the same time, but we have been enjoying multitasking for decades, long before multi-core appeared, so, how is it possible? Let’s think about single-core systems, they can only perform one task, but the operating system distributes the CPU control along all tasks very fast so it makes us think all of them are running at the same time. So single-core CPUs usually have one execution thread, dual-core system, will have two execution threads, so they can perform two tasks at the same time. But don’t worry about how it is distributed, because the operating system does it for you.

We’ve seen the OS is smart enough to take advantage of our CPUs or CPU cores, so does it have any sense to make our application use several cores, or make our application run two things at the same time?
Of course it does, let’s see some examples:

  • Imagine a GUI program performing a heavy task. While this task is running, we may want to give the user some little control to cancel the process, for example, or to carry on using our software. We can run the heavy task in one thread and the GUI in another one.
  • A server which attends several clients simultaneously.
  • A program which does intense calculations, if we have dual-core, it can be twice faster

In the last case, we must think about something a little bit. Distributing tasks in time takes some time for the OS, so we must make sure the time we save (separating the process in several threads) is bigger than the time the OS wastes managing the new tasks and sometimes we must put all the results of all individual threads together (and it takes time too).
We may think it will be valid only for multi-core processors but the time taken by our tasks may vary depending on what we are doing, for example, calculations take all CPU Time, but if we access external devices (keyboard, screen, hard disk, USB, etc), even access a file, it makes our process wait for the data and while we are waiting, the CPU is wasting time, so we can take this time to perform calculations for some other process, and it is valid even on single-core CPUs. Sometimes running a tasks in two threads can squeeze our CPU better than one thread using CPU while the other thread is just waiting for data an vice-versa.

Ok, before playing a little bit with it we must know there are some ways to do that. We can create several processes, it’s like running some programs at the same time, and the other way is creating threads. The difference is that processes must store in RAM memory, data and stack among other things, and some threads can share code and data, so only stack and a little information about execution status belongs to each thread exclusively. We can say process initiates also a single execution thread.

Let’s code a bit, we will use pthreads (POSIX threads):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>

void *newtask(void *null)
{
   printf("Hello world! I'm a new thread\n");
   sleep(1);
   printf("Goodbye world! I'm a thread about to die\n");
   pthread_exit(NULL);
}

int main (int argc, char *argv[])
{
   pthread_t thread;
   int rc;

   printf ("Main process just started.\n");
   rc = pthread_create(&thread, NULL, newtask, NULL);
   if (rc)
     {
       printf("ERROR in pthread_create(): %d\n", rc);
       exit(-1);
     }

   printf ("Main process about to finish.\n");
   /* Last thing that main() should do */
   pthread_exit(NULL);
}

This example can be compiled including pthread library, this way:

$ gcc -o onethread onethread.c -lpthread

We are just executing a program that creates another thread that performs a task, this task is writing a couple of messages on screen and wait a bit between them. We can see a result like this:

Main process just started.
Main process about to finish.
Hello world! I’m a new thread
[ wait for a second ]
Goodbye world! I’m a thread about to die

The first two messages belongs to the main thread, one when it stats and the other one when it ends, but before ending we are invoking the other thread that writes the other two messages. As we can see, each thread is executed independently, but to see it a bit more clearer, let’s see the next code:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>

void *newtask(void *null)
{
   printf("Hello world! I'm a new thread\n");
   int i;

   for (i=0; i<10;++i)
     {
       printf (" THREAD: %d\n", i);
       usleep(60);
     }
   printf("Goodbye world! I'm a thread about to die\n");
   pthread_exit(NULL);
}

int main (int argc, char *argv[])
{
   pthread_t thread;
   int rc;
   int i;

   printf ("Main process just started.\n");
   rc = pthread_create(&thread, NULL, newtask, NULL);
   if (rc)
     {
       printf("ERROR in pthread_create(): %d\n", rc);
       exit(-1);
     }

   for (i=0; i<10;++i)
     {
       usleep(100);
       printf (" MAIN: %d\n", i);
     }
   printf ("Main process about to finish.\n");
   /* Last thing that main() should do */
   pthread_exit(NULL);
}

What we’ve done is that the main thread (MAIN) and secondary thread (THREAD) will write on screen several messages inside a for loop. They will be counting from 0 to 9, but waiting a little bit between each writing (instead of doing heavier time eater tasks we simulate them with sleeps).

We can see a execution result like this:

Main process just started.
Hello world! I’m a new thread
THREAD: 0
THREAD: 1
MAIN: 0
THREAD: 2
MAIN: 1
THREAD: 3
THREAD: 4
MAIN: 2
THREAD: 5
MAIN: 3
THREAD: 6
MAIN: 4
THREAD: 7
THREAD: 8
MAIN: 5
THREAD: 9
MAIN: 6
Goodbye world! I’m a thread about to die
MAIN: 7
MAIN: 8
MAIN: 9
Main process about to finish.

Messages are written alternatively both MAIN and THREAD, like interlaced, now we can see they are both running concurrently, at the same time. We can start making multi-thread software in C.

Photo: Toastyken (Flickr) CC-by

Top