On Mon, 18 May 2026 12:44:40 GMT, Vladimir Kozlov <[email protected]> wrote:
>> When running very small Java programs in the AOT training mode, the 3 lambda >> expressions in ModuleBootstrap.java are evaluated. However, when running >> without the AOT training mode, these lambda expressions are not evaluated. >> >> As a result, the AOT training mode captures a much large profile (due to the >> complex implementation of the lambda framework). This causes the AOT cache >> of very small Java programs to be much larger than necessary. >> >> The fix is to rewrite these 3 lambda expressions (streams) back to tradition >> iterative code (contributed by @AlanBateman) >> >> Changes in ModuleBootstrap.java revealed a bug in >> `AOTArtifactFinder::find_artifacts()`: some AOT-initialized classes may be >> incorrectly excluded. >> >> >> --------- >> - [x] I confirm that I make this contribution in accordance with the >> [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai). > > Good. > > Does this change performance for small app without AOT? Thanks @vnkozlov @liach for the review ------------- PR Comment: https://git.openjdk.org/jdk/pull/31142#issuecomment-4491219186
