Adam Bradley, this e-mail is probably most targeted for you. A week ago Jorrit pointed out to me a slight performance problem with the event subscription algorithm. This is seen with the "bouncy" demo in CEL as a glitch when adding lots of balls to the system (each ball has it's own event listener) in revision 2141 (CEL moved to a mega listener for all the balls after that). I have been reviewing the event system code and found the cause of the performance problem, but before I dig in to fix it I have some questions.
1) It doesn't appear that you are using the topological sort algorithm (topo sort is O(|E|+|V|) and csPO::Solve() is at least O(|V|^2)). Is there a particular reason you didn't use this? 2) What exactly is the policy for event handlers that get subscribed during event handling? I think that it is that the handler will get sent the current event if none of its post constraints have already handled the event, but I want to double check that with you. 3) Do you have a test suite for event handling? I can convert the system to use a topo sort algorithm, but a test suite would greatly help that. Michael D. Adams [EMAIL PROTECTED] P.S. The cause of the performance problem was principally the cost of doing a csPartialOrder::SanityCheck under debug mode, but even with the body of SanityCheck commented out there is still a small performance glitch with large numbers of balls in "bouncy". It appears to be mostly due to the SubscriberIterator operating in SI_GRAPH mode and its call to GetEnabled(). ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Crystal-main mailing list Crystal-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/crystal-main Unsubscribe: mailto:[EMAIL PROTECTED]
