Recently I've encountered a still movie showing high CPU usage and did some research about it to find that any looping timeline would be considered invalidated and thus fully redrawn, no matter if things change or not.
Here's a launcher for my test case: http://strk.keybit.net/tmp/graph.sh It comes from a WordPress plugin page. The SWF shows a graph, fixed after the first load, but rendered 15 times per second, keeping CPU very busy. Since invalidated bounds detection isn't tested _very_ much (but enough to prevent me from doing a too naive fix) I haven't pushed my optimization of it to master but in a "fast" branch. With the "fast" branch the "graph" case can be played with CPU around 8% rather than 100%. Not bad... Also, the testsuite gives an XPASS, for a test about inv. bounds that was there since a long time. So, this mail is to ask you to test the "fast" branch against your usual movies, where you can easily catch any problems with your eyes. In particular, a regression (if any) would be in partially rendered frames, wherease a refresh would show the complete rendering. I don't expect any regression, but would like a broader confirmation about it. Also, would be nice to hear if you can see other real-world cases of CPU relief with this. Thanks in advance. --strk; () Free GIS & Flash consultant/developer /\ http://strk.keybit.net/services.html _______________________________________________ Gnash-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash-dev
