|
From: | ZouJQ |
Subject: | RE: [Gnash-dev] gnash on mips performance issue |
Date: | Wed, 3 Feb 2010 20:30:03 +0800 |
Something about frame dropping. There is #define SKIP_RENDERING_IF_LATE 1 in gui.cpp. It checks whether the frame should be dropped. I have also checked Gui::advanceMovie() in gui.cpp, and found: bool advanced = m->advance(); -> movie_root::advance() -> movie_root::advanceMovie() There is advanceLiveChars(); this function seems consuming much cpu timing. Can we add some aditional logic in movie_root::advance(). try { const size_t elapsed = now - _lastMovieAdvancement; if (elapsed >= _movieAdvancementDelay) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Does it necessary? I think it waste time. it don't skip any frame here. { advanced = true; advanceMovie(); // To catch-up lateness we pretend we advanced when // was time for it. // NOTE: // now - _lastMovieAdvancement // gives you actual lateness in milliseconds // _lastMovieAdvancement += _movieAdvancementDelay; } //log_debug("Latenss: %d", now-_lastMovieAdvancement); executeAdvanceCallbacks(); executeTimers(); } > Date: Fri, 29 Jan 2010 10:49:34 +0100 > Subject: Re: [Gnash-dev] gnash on mips performance issue > From: address@hidden > To: address@hidden > CC: address@hidden > > 2010/1/29 ZouJQ <address@hidden>: > > I am use --enable-gui=sdl with agg as renderer. > > After run gprof, it shows the bottleneck is agg renderer. > > > > Now I do a test. In function "Gui::display(movie_root* m)" gui.cpp, remove > > "m->display();"(Render the frame) and "renderBuffer();"(show frame on > > screen). The fps is the same as x86 !!! > > Could be a false alarm, but I would check if SDL GUI deals correctly > with invalidated regions. > It migth be faster to compare the SDL gui with the GTK gui first. If > GTK is faster then we > would have found a culprit. > > > 1. Is there any way to optimize agg? Can agg do something changing on mips > > for better performance? > > Most likely. > Do you have fundings or expertise to help with that ? > > Also frame dropping (which I've been thinking about) could reduce the > problem with lower > effort, but improving AGG use is surely something worth doing. > > > 2. It is an exciting news that the next release will come soon, thank you > > very much. Does this release modify renderer much, such as agg and cairo ? > > No significant changes in rendering performance. > More details on changes here: http://wiki.gnashdev.org/Release_0.8.7 > > --strk; > > () Free GIS & Flash consultant/developer > /\ http://foo.keybit.net/~strk/services.html > > > _______________________________________________ > Gnash-dev mailing list > address@hidden > http://lists.gnu.org/mailman/listinfo/gnash-dev 使用新一代 Windows Live Messenger 轻松交流和共享! 立刻下载! |
[Prev in Thread] | Current Thread | [Next in Thread] |