Meta-issue: Improve JabRef's performance - Reduce CPU / RAM usage, increase speed and detect bottlenecks

[ThiloteE]

This meta-issue is here to provide a starting point for anybody who wants to work on performance issues and point them to related literature and utilities. It also keeps track of certain bugs related to performance degradation.

I would ask you to NOT START A DISCUSSION here. Comments may be hidden that do not add to the tools and documentation section.

---

JabRef's performance problems:

• minimum requirements to run JabRef are comparatively high
• High CPU usage --> high usage of electricity
• Sometimes JabRef is slow (context: large databases)

Solution

• Decrease CPU / RAM usage
• Detect performance Bottlenecks
• Do some performance checks from time to time and change code. Make JabRef become better and better.

Tools and documentation:

• devdocs about how to test JabRef's code
• devdocs about benchmarks
• devdocs about measure performance
• A Script for showing memory usage on Linux
• VisualVM, which is a tool to capture CPU, processes, threads and memory performance, as well as usage. It was used in https://github.com/JabRef/jabref/issues/8416#issuecomment-1086531925. According to https://questdb.io/blog/billion-row-challenge-step-by-step/, the plugin "VisualGC" should be used to see what the garbage collector does.
• Beginners guide to analyze java performance issues (from 2016, but seems well explained).
• For looking into virtual memory: On windows 10, there is the "performance monitor" tool available. I also found VMMap that apparently can analyse (virtual) memory usage patterns.

How to test:

• Disable/Avoid known causes for performance drops (unless working on a fix/improvement for these specific cases). Where to find these specific causes? --> Check issues on GitHub with label "performance"
• Also test with large databases.

---

How to create a large database:

1. Install Python3. E.g., via https://www.python.org/downloads/
2. Download https://github.com/JabRef/jabref/blob/main/scripts/bib-file-generator.py
3. Open terminal
4. cd into your Download directory
5. Run python3 bib-file-generator.py

[HoussemNasri]

Running heap dump analysis using jxray provided some interesting insights into JabRef's memory usage. The most interesting for me was that we have ~138MB of off-heap memory allocated (still don't know how and when it gets allocated) and many duplicated SimpleStringProperty and SimpleBooleanProperty objects around ~40MB of overhead.

It also discovered some memory leak candidates:

• 102,834Kb (29.0%) Object tree for GC root(s) Java Static org.jabref.gui.preferences.general.GeneralTabViewModel.fontSizeValueFactory
• 64,401Kb (18.2%) Object tree for GC root(s) Java Static org.jabref.gui.autosaveandbackup.BackupManager.runningInstances
• 47,018Kb (13.3%) Object tree for GC root(s) Java Static [org.jabref.gui.JabRefGUI.preferences], Java Local(org.jabref.gui.preferences.JabRefGuiPreferences) [1 object(s)] (1 thread(s))
• 34,475Kb (9.7%) Object tree for GC root(s) Java Static org.jabref.gui.JabRefGUI.mainFrame
• 33,309Kb (9.4%) Object tree for GC root(s) Java Static org.jabref.logic.citationstyle.CitationStyle.STYLES

jabref-xray-heap-analysis.zip