Community
Participate
Working Groups
Using 3.1M2, the performance of code completion is sometimes unacceptable. I created a thread dump and will attach it...
Created attachment 14893 [details] stack trace while codecompletion runs
Created attachment 14894 [details] different looking thread dump
Moving to JDT/Core for comments. Jörg, do you have a scenario / numbers that show a slowdown of 3.1 M2 with respect to earlier builds?
(just had a bugzilla "mid air collision", what I just wanted to say is:) What is even less understandable: once the list of completions has popped up, in my case with only one found completion, I press return to chose it and then CPU usage jumps to the ceiling again for quite a while until the code is inserted at the edit location. Please see third attachment for the corresponding thread dump. Just to have mentioned it: when I say "unacceptable performance", I really meant high CPU usage for nearly 1/2 minute on a 1,6 Ghz machine (its not that my jars were on a network drive or something)
Created attachment 14896 [details] third dump, while inserting completed code
Markus, when I edit the same code in the same way using 3.0 (Release 3.0, not 3.0.1), I don't get this long pause. There is a little perceivable delay, but it's absolutely acceptable and I certainly wouldn't have the time to click on the console window and take a thread dump...
this bug got stuck 'between door' with its component being jdt ui and owner jdt-core-inbox moving to jdt core
Some improvement have been done in this area since build 3.1 M2. Please try with last milestone build 3.1 M4 and let us know if it always happen, thanks
No, this one doesn't happen anymore for me. Thanks!
Reopen to set as duplicate...
*** This bug has been marked as a duplicate of 76027 ***