Community
Participate
Working Groups
There was a performance regression in 2.4 from 2.3 in reading tests from 4-10% in the emulated database run. There was no significant regression in the Oracle database run, other than the dynamic JPQL test of ~10% which was expected due to Hermes (dynamic JPQL is quite rare because of our parse cache).
Compared EclipseLink versions: 2.3.0.v20110604-r9504 and 2.4.0.v20120601-r11584
Created attachment 217307 [details] some micro optimizations
SVN trunk commit: bug#382528 - isolated performance https://bugs.eclipse.org/bugs/show_bug.cgi?id=382528 A few micro optimizations. Code review: Gord (pending) Changes: - Cache if null values require conversion in direct mapping - Data optimize BigDecimal in accessor - Avoid locks for isolated cache keys - Avoid reflection for creation of standard collection types in container policy - Inline index in DatabaseField - Avoid creating wrapper aggregate cache key if not protected
Created attachment 217551 [details] a few more optimizations
Changing target - missed 2.4.1 release.
Deferring to 2.5
The Eclipselink project has moved to Github: https://github.com/eclipse-ee4j/eclipselink