Programming with POSIX® Threads

56 strcpy (request->path, filepath);

57 request->string = search;

58 request->next = NULL;

59 if (crew->first == NULL) {

60 crew->first = request;

61 crew->last = request;

62 } else {

63 crew->last->next = request;

64 crew->last = request;

65 } 66

67 crew->work_count++;

68 status = pthread_cond_signal (&crew->go);

69 if (status != 0) {

70 free (crew->first);

71 crew->first = NULL;

72 crew->work_count = 0;

73 pthread_mutex_unlock (&crew->mutex);

74 return status;

75 }

76 while (crew->work_count > 0) {

77 status = pthread_cond_wait (&crew->done, &crew->mutex);

78 if (status != 0)

79 err_abort (status, "waiting for crew to finish");

80 }

81 status = pthread_mutex_unlock (&crew->mutex);

82 if (status != 0)

83 err_abort (status, "Unlock crew mutex");

84 return 0;

85 }

Part 5 shows the initial thread (main) for the little work crew sample.

10-13 The program requires three arguments — the program name, a string for which to search, and a path name. For example, "crew butenhof ~"

15-23 On a Solaris system, call thr_setconcurrency to ensure that at least one LWP (kernel execution context) is created for each crew member. The program will work without this call, but, on a uniprocessor, you would not see any concurrency. See Section 5.6.3 for more information on "many to few" scheduling models, and Section 10.1.3 for information on "set concurrency" functions.

24-30 Create a work crew, and assign to it the concurrent file search.

■ crew.c part 5 main

1 /*

2 * The main program to "drive" the crew...

3 */

4 int main (int argc, char *argv[])

5 {

6 crew_t my_crew;

7 char line[128], *next;

8 int status;

9

10 if (argc < 3) {

11 fprintf (stderr, "Usage: %s string path\n", argv[0]);

12 return -1;

13 }

14

15 #ifdef sun

16 /*

17 * On Solaris 2.5, threads are not timesliced. To ensure

18 * that our threads can run concurrently, we need to

19 * increase the concurrency level to CREW_SIZE.

20 */

21 DPRINTF (("Setting concurrency level to %d\n", CREW_SIZE));

22 thr_setconcurrency (CREW_SIZE);

23 #endif

24 status = crew_create (&my_crew, CREW_SIZE);

25 if (status != 0)

26 err_abort (status, "Create crew");

27

28 status = crew_start (&my_crew, argv[2], argv[l]);

29 if (status != 0)

30 err_abort (status, "Start crew");

31

32 return 0;

33 }

4.3 Client/Server

But the Judge said he never had summed up before;

So the Snark undertook it instead,

And summed it so well that it came to far more

Than the Witnesses ever had said!

— Lewis Carroll, The Hunting of the Snark

In a client/server system, a "client" requests that a "server" perform some operation on a set of data (Figure 4.3). The server performs the operation independently — the client can either wait for the server or proceed in parallel and look for the result at a later time when the result is required. Although it is simplest to have the client wait for the server, that's rarely very useful — it certainly doesn't

FIGURE 4.3Client/Server

provide a speed advantage to the client. On the other hand, it can be an easy way to manage synchronization for some common resource.