#include <runnable_batch.h>
The class operates using a small pool of threads, so new tasks are run in existing free threads, not new ones. Overhead will only be due to locking in MessageQueue and shared_ptr.
Definition at line 17 of file runnable_batch.h.
Public Member Functions | |
RunnableBatch (unsigned int p) | |
void | join () |
void | schedule (std::tr1::shared_ptr< Runnable > r) |
~RunnableBatch () | |
Classes | |
class | RunMessageInThread |
CVD::RunnableBatch::RunnableBatch | ( | unsigned int | p | ) |
Construct a job manager pallelism Number of threads to use.
Note that 1 thread will cause jobs to be executed in one new thread. 0 threads will cause the jobs to be executed in the current thread. This will prevent thread spawning and should make debugging easier. This is the behaviour if CVD is compiled without threading support.
CVD::RunnableBatch::~RunnableBatch | ( | ) |
Destruct the job manager.
This will wait for all threads to finish and terminate.
void CVD::RunnableBatch::join | ( | ) |
Wait until all pending tasks have been run, then terminate all threads.
void CVD::RunnableBatch::schedule | ( | std::tr1::shared_ptr< Runnable > | r | ) |
Put a task on the queue.
This will be run at some point in the future. Job lifetime is managed by shared_ptr, so they may be managed wither by RunnableBatch or by the caller.