4.6)  How do I sleep() in a C program for less than one second?

      The first thing you need to be aware of is that all you can
      specify is a MINIMUM amount of delay; the actual delay will
      depend on scheduling issues such as system load, and could be
      arbitrarily large if you're unlucky.

      There is no standard library function that you can count on in
      all environments for "napping" (the usual name for short
      sleeps).  Some environments supply a "usleep(n)" function which
      suspends execution for n microseconds.  If your environment
      doesn't support usleep(), here are a couple of implementations
      for BSD and System V environments.

      The following code is adapted from Doug Gwyn's System V emulation
      support for 4BSD and exploits the 4BSD select() system call.
      Doug originally called it 'nap()'; you probably want to call it
      "usleep()";

      /*
            usleep -- support routine for 4.2BSD system call emulations
            last edit:  29-Oct-1984     D A Gwyn
      */

      extern int        select();

      int
      usleep( usec )                            /* returns 0 if ok, else -1 */
            long                usec;           /* delay in microseconds */
            {
            static struct                       /* `timeval' */
                    {
                    long        tv_sec;         /* seconds */
                    long        tv_usec;        /* microsecs */
                    }   delay;          /* _select() timeout */

            delay.tv_sec = usec / 1000000L;
            delay.tv_usec = usec % 1000000L;

            return select( 0, (long *)0, (long *)0, (long *)0, &delay );
            }

      On System V you might do it this way:

      /*
      subseconds sleeps for System V - or anything that has poll()
      Don Libes, 4/1/1991

      The BSD analog to this function is defined in terms of
      microseconds while poll() is defined in terms of milliseconds.
      For compatibility, this function provides accuracy "over the long
      run" by truncating actual requests to milliseconds and
      accumulating microseconds across calls with the idea that you are
      probably calling it in a tight loop, and that over the long run,
      the error will even out.

      If you aren't calling it in a tight loop, then you almost
      certainly aren't making microsecond-resolution requests anyway,
      in which case you don't care about microseconds.  And if you did,
      you wouldn't be using UNIX anyway because random system
      indigestion (i.e., scheduling) can make mincemeat out of any
      timing code.

      Returns 0 if successful timeout, -1 if unsuccessful.

      */

      #include <poll.h>

      int
      usleep(usec)
      unsigned int usec;                /* microseconds */
      {
            static subtotal = 0;        /* microseconds */
            int msec;                   /* milliseconds */

            /* 'foo' is only here because some versions of 5.3 have
             * a bug where the first argument to poll() is checked
             * for a valid memory address even if the second argument is 0.
             */
            struct pollfd foo;

            subtotal += usec;
            /* if less then 1 msec request, do nothing but remember it */
            if (subtotal < 1000) return(0);
            msec = subtotal/1000;
            subtotal = subtotal%1000;
            return poll(&foo,(unsigned long)0,msec);
      }

      Another possibility for nap()ing on System V, and probably other
      non-BSD Unices is Jon Zeeff's s5nap package, posted to
      comp.sources.misc, volume 4.  It does require a installing a
      device driver, but works flawlessly once installed.  (Its
      resolution is limited to the kernel HZ value, since it uses the
      kernel delay() routine.)

      Many newer versions of Unix have a nanosleep function.