Skip to content

Memory footprint: what happens inside µBlock after installation

Raymond Hill edited this page Feb 13, 2015 · 31 revisions
  1. µBlock will load the default selection of filter lists:
    • This causes short-term memory churning (loading/parsing/sorting/storing), which short-term memory will be garbage-collected eventually
    • All this short-term memory churning causes µBlock's baseline memory footprint to grow further
  2. A few minutes later, µBlock will look whether one or more filter lists need to be updated
  3. If one or more filter lists need to be updated, µBlock will launch a background task which will fetch an updated version of whatever filter list is deemed obsolete
  4. Once all filter lists are brought up to date, µBlock will flush from memory all filters and reload with the latest version
    • This will causes another round of short-term memory churning, which short-term memory will be garbage-collected eventually
    • Again, all this short term memory churning causes µBlock's baseline memory footprint to grow further
    • You can disable auto-update if you want, but it is optimal to let µBlock take care of this
  5. A few minutes minutes later, assuming no change in selection of filter lists, or change in the content of selected filter lists, µBlock will make a selfie
    • A selfie allows µBlock to skip the parsing/sorting of data next time it loads
    • Generating the selfie also causes short-term memory churning, which short-term memory will be garbage-collected eventually
    • Again, this short term memory churning causes µBlock's baseline memory footprint to grow further
    • Any change in the selection of filter lists, or change in the content of selected filter lists will invalidate µBlock's selfie
  6. Even after the growth in memory baseline, µBlock's own memory footprint is still quite smaller than that of Adblock Plus -- once the garbage collector does its job.
  7. µBlock's much smaller contributed memory footprint to web pages is much smaller than that of ABP
    • The contributed footprint to web pages is part of the memory footprint of the web pages themselves
    • As opposed to an extension's own memory footprint, visible using Chromium's "Task Manager", the contributed memory footprint to web pages cannot be easily seen by users
    • Though this measure is not readily visible, it's where you get the biggest bang for the buck with µBlock relative to ABP -- because µBlock does not inject thousands of CSS rules into pages and embedded frames.

Now if you have reached the point where there is a valid µBlock selfie, this is when µBlock will run the most efficiently.

The next time you launch µBlock and there is a valid selfie, the load time will be a fraction of the load time when no selfie is available, and µBlock's baseline memory footprint will be smaller than when µBlock launches without a selfie available.

So my point is that µBlock will perform best efficiency wise if you leave it time to optimize itself after installation: In subsequent launches, µBlock will perform more efficiently than what you may have observed right after you installed it.


Also, if you want to look at memory figures, aside the above notes regarding garbage collection, keep in mind:

  • The Chromium bug which causes systemetically a memory leak when opening an extension popup UI.
  • Currently opened extension option page(s) contribute to increase the extension's memory use
  • Sometimes a browser's garbage collector ("GC") is lazier than others, i.e. it may takes longer to kick in.
    • In my benchmarks, it has happened I had to force a GC cycle using dev tools.

Clone this wiki locally