On sexta-feira, 9 de agosto de 2013 19:04:56, Oswald Buddenhagen wrote: > i'll note that the qml guys identified that as a major bottleneck a few > years ago and wrote their own timer code.
Probably for the same reason why we identified the glib GTimeout as a major bottleneck (and grew worse for a long-lived application[1]) back in 2006, so we decided not to use one GTimeout for each QTimer, but simply keep our own list and tell glib of one timer only: impedance mismatch. Sure, the glib code back in 2006 was a singly-linked priority list where everything got appended to the end (O(n) appending) and leaked. But we still keep our own list today for the simple reason that we don't have to tell glib about each timer, and have it allocate more resources in addition to ours. [1] that's actually the reason why QT_NO_GLIB exists: I needed to benchmark and debug the same Qt applications with and without glib. The application in question, back in Qt pre-4.2 days before KDE 4 got released, was the Teambuilder monitor. It was the only application I left running for a long time. -- Thiago Macieira - thiago.macieira (AT) intel.com Software Architect - Intel Open Source Technology Center
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Development mailing list [email protected] http://lists.qt-project.org/mailman/listinfo/development
