Community
Participate
Working Groups
We have been profiling the creation of a Servlet using WTP 3.0 on top of Eclipse 3.4. It appears that over 50% of the time that it takes to create a Servlet is spent in the JDT, with most of that time being taken by the JDT indexing. This is a major performance difference from WTP 1.5.x (which was based on Eclipse 3.2).
Looking at 3.3.2 and 3.4RC4 JDT/Core performance results, indexing didn't suffer such a regression since 3.2... Looking at the 2 existing tests (indexing the JDT/Core plugin and indexing all Eclipse workspace), I can see that for windows boxes we did improve speed of indexing between 3.2 and 3.3 (from 4.5% to 11.8%) and we are pretty stable between 3.3 and 3.4 (from -1.2% to 0.3%). Didn't WTP use a specific IndexManager implemented on top of JDT/Core IndexManager? Could it be possible that this strong regression comes from it? Do you have performance tests with more specific numbers to measure how big is the regression? Also perhaps some Yourkit snapshots showing in which precise area the time is spent?
Closing as WORKSFORME as per comment 1 and no feedback since then. Please reopen with requested information if you're still getting this issue, thanks.
Verified for 3.5M4 using I20081209-0100
