Just to be clear, what is happening is that a recent feature of Chrome is that if a mouse event triggers a setTimeout, the system is monitored for how long it takes to process, and if that process time exceeds some threshold, then from that time forward setTimeout processing is given a lower priority. In our case, it means that all updating of the display is put off until the mouse is let go. From
https://code.google.com/p/chromium/issues/detail?id=568725 #20 <https://code.google.com/p/chromium/issues/detail?id=568725#c20> alexcla...@chromium.org <https://code.google.com/u/109345702659122941885/> As much as possible chrome tries to do scrolling and animation on the compositor thread to try and make them smooth even when the mainthread is very unresponsive. Various things that can't be done by the compositor, e.g, running the layout pipeline (which has to be done on the main thread). In the context of this patch, a 'frame' is generated when the compositor asks the main thread to run the layout pipeline by posting a task to run BeingMainFrame. If animations stop and no more frames are needed the compositor will tell the main thread by calling BeginFrameNotExpectedSoon. When that happens the scheduler stops trying to optimize for fps. Under some circumstances Chrome will block loading and or timer tasks, if it thinks there's more important work pending. The details of this are complicated (I hope to document this properly in a public doc/blogpost), and it's quite possible it's not working well in all cases. What would be really useful is a link to the web application (even a cut down version) or a trace. Then we can investigate and potentially make fixes if the scheduler is doing the wrong thing. PS the scheduler team is now out till January, but I will be sure to follow up if you can send us anything more.
------------------------------------------------------------------------------
_______________________________________________ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
