diff options
| author | Rupert Smith <rupertlssmith@apache.org> | 2007-10-02 16:29:15 +0000 |
|---|---|---|
| committer | Rupert Smith <rupertlssmith@apache.org> | 2007-10-02 16:29:15 +0000 |
| commit | 87cf93aba115c215cd0ca16e9f944e96d43a863b (patch) | |
| tree | d2e37870e4d26b1cea68eb4801c3c5687c95f665 /java/perftests/src | |
| parent | 6329b7a3a7ccd60ec42cb6d1348693f37fab58d0 (diff) | |
| download | qpid-python-87cf93aba115c215cd0ca16e9f944e96d43a863b.tar.gz | |
QPID-615, Added patched version of MINAs VM Pipe cleanup thread. Will replace once bug fix is in newer version of MINA.
git-svn-id: https://svn.apache.org/repos/asf/incubator/qpid/branches/M2.1@581293 13f79535-47bb-0310-9956-ffa450edef68
Diffstat (limited to 'java/perftests/src')
| -rw-r--r-- | java/perftests/src/main/java/org/apache/qpid/requestreply/PingPongProducer.java | 10 |
1 files changed, 0 insertions, 10 deletions
diff --git a/java/perftests/src/main/java/org/apache/qpid/requestreply/PingPongProducer.java b/java/perftests/src/main/java/org/apache/qpid/requestreply/PingPongProducer.java index d114c29756..ed77332e19 100644 --- a/java/perftests/src/main/java/org/apache/qpid/requestreply/PingPongProducer.java +++ b/java/perftests/src/main/java/org/apache/qpid/requestreply/PingPongProducer.java @@ -119,21 +119,11 @@ import java.util.concurrent.atomic.AtomicLong; * <tr><td> Provide command line invocation to loop the ping cycle on a configurable broker url.
* </table>
*
- * @todo Make the message listener a static for all replies to be sent to? It won't be any more of a bottle neck than having
- * one per PingPongProducer, as will synchronize on message correlation id, allowing threads to process messages
- * concurrently for different ids. Needs to be static so that when using a chained message listener and shared
- * destinations between multiple PPPs, it gets notified about all replies, not just those that happen to be picked up
- * by the PPP that it is atteched to.
- *
* @todo Use read/write lock in the onmessage, not for reading writing but to make use of a shared and exlcusive lock pair.
* Obtain read lock on all messages, before decrementing the message count. At the end of the on message method add a
* block that obtains the write lock for the very last message, releases any waiting producer. Means that the last
* message waits until all other messages have been handled before releasing producers but allows messages to be
* processed concurrently, unlike the current synchronized block.
- *
- * @todo Get rid of pauses between batches, it will impact the timing statistics, and generate meanigless timings.
- * Instead make mina use a bounded blocking buffer, or other form of back pressure, to stop data being written
- * faster than it can be sent.
*/
public class PingPongProducer implements Runnable /*, MessageListener*/, ExceptionListener
{
|