wpool.h3

cc_wpool* cc_wpool_construct (intnumworkers)

Construct worker pool.

void cc_wpool_destruct (cc_wpool *pool)

Destruct worker pool. Will wait for all jobs in progress to finish

int cc_wpool_get_num_workers (cc_wpool *pool)

Returns the number of workers in the pool.

void cc_wpool_set_num_workers (cc_wpool *pool, intnewnum)

Sets the number of workers in the pool.

void cc_wpool_wait_all (cc_wpool *pool)

Wait for all pool workers to finish working and go into idle state. This method should only be called by the thread controlling the pool. A pool worker should not call this method, since it will obviously never return from here (it will never go idle).

SbBool cc_wpool_try_begin (cc_wpool *pool, intnumworkersneeded)

Locks the pool so that workers can be started using the cc_wpool_start_worker() method. numworkersneeded should contain the minumum number of workers that is needed. If workers are available, the pool will be locked and TRUE is returned. Otherwise FALSE is returned.

Usage pseudocode:

int numworkers = 5;
if (cc_wpool_begin(pool, numworkers)) {
  for (int i = 0; i < numworkers; i++) {
    cc_wpool_start_worker(my_work[i], my_closure[i]);
  }
  cc_wpool_end(pool);
}

Important! If too few workers are available, the pool will not be locked and cc_wpool_end() should not be called.

See also:

void cc_wpool_begin (cc_wpool *pool, intnumworkersneeded)

Wait for numworkersneeded workers to become idle. When returning from this call, the pool will be locked, and up to numworkersneeded can be started using the cc_wpool_start_worker() method. Remember to call cc_wpool_end() to unlock the pool again.

See also:

void cc_wpool_start_worker (cc_wpool *pool, cc_wpool_f *workfunc, void *closure)

Starts a worker. The pool must be locked (using cc_wpool_begin()) before calling this method.

See also:

void cc_wpool_end (cc_wpool *pool)

Unlocks the pool after a cc_wpool_begin(), cc_wpool_start_worker() sequence.

Please note that if cc_wpool_begin() returns 0, you should not call cc_wpool_end().

See also: