Performance/MemShrink/Meetings/2011-10-11

From MozillaWiki
Jump to: navigation, search

We completed triage and are tracking follow-up activites for the following bugs until we achieve their resolution:


Bug 693527 - add Watchpoint support to the cycle collector

  • Andrew is working on this; looks like it might be blocked on testing.

Bug 693404 - Enable jemalloc on MacOS 10.5 

  • jemalloc doesn't accurately measure RSS on OSX.
  • Justin to follow up with graphics people who know people elancaster and legneato to sleuth this with Apple people, too. elancaster checking in with Joel about this.
  • Could keep a list of every page we'd freed, then check if they are there or not, but that would be pretty unpleasant, and also not be reflected in top. Not clear if anybody else does this and gets it right. jemalloc with decommit enabled is correct, but very slow. Various crazy strategies for possibly measuring it, but probably wouldn't want to run them in a normal build.

Bug 693247 - [Endurance] Memory usage spike detected in Mozmill endurance tests - All OSs

  • We need to understand whether or not this spike is real. If it is, fix it. If it isn't, let's audit the tests

Bug 693088 - Fennec does not nuke tabs when backgrounded

  • elancaster tracking this as part of the release plan for the Native UI

Bug 693016 - tbpl leaks memory badly

  • need to understand see if this leak happens in chrome suspicion is that this is tbpl sucking, Need to repro with two browsers side by side

Bug 692748 - Increasing memory due to RSS feeds in bookmarks folder

  • we are churning through some stuff like js which would lead to unused;15 MBs more heap unclassified. Nick to run tools good to look at why heap unclassified is so high to start with… it goes down the more he runs the tests, it will tell us this issus is likely an edge case.

Bug 683953 - Browser-chrome mochitests should show statistics about leaked DOMWindows and DocShells

  • khuey has a better idea on how to get things done, here. Likejust capturing regression in formation

Bug 683290 - We won't discard any images on the current tab even if they are not in the DOM

  • khuey is on this setting up an android test environment

678859 - Runaway memory usage with DOM Animation (Windows only?)

  • JS to bug bas

670596- GC: Allow unused arenas to be marked as uncommitted

  • This is a P1; need to check in about what needs to happen next

411894 - Storage service should obey "memory-pressure" and tell sqlite to free memory

  • This is important for mobile; need to understand what needs to happen next

Bug 386451 - Too much memory used by Thebes Images to display 4.5Mb Image

  • We are not using 32bits per pixel; we are not sure that people actually encounter this very much. We need to understand this better.

Bug 631811 - [tracking] Turn Valgrind builds into a full-fledged acceptable test

  • Andrew to do a post letting people know that they can use this on their code.

Bug 672443 - [Jetpack] Too many compartments!

  • Luke is going to add asserts to make sure that we are good in fx… this will happen for beta

Bug 678037 - most JSScripts created in the browser are never run

  • P1; need to figure out what needs to happen

Bug 681201 - Not-yet-restored empty tabs take a lot of memory

  • P1; need to push this bug along