Why system time larger than real time?

Question

It is a program that calculates the number of divisors of 10000 as the maximum number of threads created by 1000.

I thought realtime was larger and turned.

n_Threads is 1000

thread.c

struct tms mytms;
    clock_t t1, t2;

    if( (t1 = times( &mytms )) == -1 )
    {
        perror( "times 1" );
        exit( 1 );
    }
    for( int i = 0; i < n_Threads; i++ )
    {
        if( pthread_create( &tid[i], NULL, &thread_func, NULL ) != 0 ){
            fprintf( stderr, "pthread_create error!\n" );
            fprintf( stderr, "%s\n", strerror( errno ) );
            exit( 1 );
        }
    }
    for( int i = 0; i < n_Threads; i++ )
    {
        pthread_join( tid[i], NULL );
    }

    if( (t2 = times( &mytms )) == -1 )
    {
        perror( "times 2" );
        exit( 1 );
    }

    fprintf( stdout, "%d 의  약 수 의  개 수 : %d \n", MAX_NUMBER, shared_Value );

    printf( "Real time : %.5f sec\n", (double)(t2 - t1) / CLK_TCK );
    printf( "User time : %.5f sec\n", (double)mytms.tms_utime / CLK_TCK );
    printf( "System time : %.5f sec\n", (double)mytms.tms_stime / CLK_TCK );

void* thread_func( void* arg )
{
    while(loopValue<=MAX_NUMBER){

        pthread_mutex_lock( (&mutex) );
        if( MAX_NUMBER % loopValue == 0 ) {
            shared_Value++;
        }
        loopValue++;
        pthread_mutex_unlock( (&mutex) );
    }
    return NULL;
}

result

10000 number of proper divisor : 25
Real time : 0.40000 sec
User time : 0.09000 sec
System time : 0.59000 sec

why system time larger than real time?

Please edit your question to show which is the exact command you have used to measure time and post also output of `uname -a`. — pifor, Jun 12 '20 at 11:10
Maybe there is a rounding issue when converting to `double`: what is the output if you measure the number of tick clocks as `intmax_t` and convert to seconds using `sysconf(_SC_CLK_TCK)` ? — pifor, Jun 12 '20 at 11:48
System time counts *total CPU time* spent inside the kernel. It can be higher than real time if you have more than one CPU or core. — Nate Eldredge, Jun 12 '20 at 18:40
Did you read [time(7)](https://man7.org/linux/man-pages/man7/time.7.html) and [clock_gettime(2)](https://man7.org/linux/man-pages/man2/clock_gettime.2.html) and [times(2)](https://man7.org/linux/man-pages/man2/times.2.html)? — Basile Starynkevitch, Jun 12 '20 at 19:15

score 1 · Answer 1 · answered Jun 12 '20 at 18:36

The system time is the amount of time your program spends in the operating system; as opposed to user time, which is the time your program spends running your direct code ( + libraries ).

You seem to have discovered that creating and deleting threads takes a lot of system time. You may also have discovered that incrementing a number 10000 times takes a lot less time than creating and deleting threads. These are good lessons.

Also, depending upon whether your mutex is properly initialized; that it takes very little time to execute mutex operations on an invalid mutex; or that you are spending a lot of time having 1000 threads thrashing on a single mutex. To figure out which, you could check the return values from these functions; often illuminating.

Maybe start with a smaller number of threads. For a program like yours, there is nothing to be gained by having more threads than CPUs. That isn't true of all programs; some threads periodically wait for I/O operations and the like; but there is seldom much justification for 1000s of threads.

Why system time larger than real time?

1 Answers1