Well that's true, but some objects continue to eat CPU until they're collected by garbage collection, that is a long as they persist in memory. There is an assigned bug open regarding AudioHostRenderer consuming CPU while in pause, but the WHATWG HTML5 specification does not specify a stop method for media elements and there doesn't seem to be a mechanism in place to inform an object all references to it have been removed so how would it know to stop consuming CPU cycles?
On Feb 24, 6:52 am, Erik Corry <[email protected]> wrote: > The bug here is to use the garbage collector to collect anything other > than memory. You can't collect CPU use with a memory manager. > > 2010/2/24 Bjorn Tipling <[email protected]>: > > > > > Hi, > > > This seems like an incorrect way to determine garbage collection: > > > const int lowUsageMB = 256; // If memory usage is below this > > threshold, do not bother forcing GC. > > const int highUsageMB = 1024; // If memory usage is above this > > threshold, force GC more aggresively. > > const int highUsageDeltaMB = 128; // Delta of memory usage growth > > (vs. last workingSetEstimateMB) to force GC when memory usage is high. > > > This is in /thirdy_party/WebKit/WebCore/bindings/v8/V8GCController.cpp > > > The problem with this is that according to my tests, there is never > > any garbage collection until you hit lowUsageMB if the tab stays > > focused or if you are running this code in a Chrome extension in > > background.html. It's not just about memory though, because some > > things that stay in memory such as an instance of AudioHostRenderer > > eat a lot of CPU, even when paused (a separate bug). > > > The tests and bug are below: > > >http://code.google.com/p/chromium/issues/detail?id=36237 > > > Seems to me that v8 should be a little smarter about when to start > > collecting garbage, because if I can quickly eat a lot of CPU with CPU > > intensive objects that persist when they should no longer or hit the > > upper cap on the low memory limit by creating objects that have a big > > memory footprint. Idle tab detection does not save a chrome extension > > from this issue. > > > Thank you. > > > -- > > v8-users mailing list > > [email protected] > >http://groups.google.com/group/v8-users -- v8-users mailing list [email protected] http://groups.google.com/group/v8-users
