Denis Auroux twisted the bytes to say: I created a testbed for our experiments.
https://github.com/dmgerman/testDrawing and please see this for more context: https://mail.gnome.org/archives/gtk-app-devel-list/2015-April/msg00002.html Denis> Hi all, Denis> I don't know anything about how GTK3 works, but I would be inclined to Denis> agree with the responses Daniel got on that mailing list: lack of Denis> smoothness of the shapes means that the *events* are not sufficiently Denis> fine-grained. Denis> Excessive rendering slowness would cause a perfectly nice shape to be Denis> drawn well behind the position of the pen, with a very noticeable lag Denis> but no loss of smoothness (as is already the case if you try to draw Denis> very fast in xournal when e.g. you're working over a remote X11 Denis> connection and/or the CPU kept 100% occupied by other processes (try Denis> e.g. running ssh -X machine from one system to another one, not on the Denis> same fast local network, and running xournal within the ssh -- possibly Denis> further overloading the network and/or CPU with other stuff going on -- Denis> and you'll see what I mean). Denis> The description I found of gdk_window_set_event_compression() sounds Denis> exactly like what would cause polygonal lines to appear instead of round Denis> circles -- this seems to be a new feature in gtk3. I have never Denis> benchmarked carefully but I do imagine that xournal typically receives a Denis> number of motion events (and creates the corresponding line segments and Denis> asks gnomecanvas to draw them, which I assume is NOT done synchronously Denis> -- in any case gnomecanvas does not actually push the display surface to Denis> the screen each time) in sequence before any display refresh actually Denis> takes place, and during the time when the display gets refreshed the Denis> motion events get queued and will be processed once the display refresh Denis> is done. The GTK3 event compression would eliminate all intermediate Denis> motion events that occurred during the display refresh and just give you Denis> a single position, losing all the intermediate ones; turning it off Denis> should help a lot. Denis> Actually, this is exactly what it says here: Denis> ======== Denis> void gdk_window_set_event_compression (GdkWindow *window, gboolean event_compression); Denis> Determines whether or not extra unprocessed motion events in the event Denis> queue can be discarded. If TRUE only the most recent event will be Denis> delivered. Denis> Some types of applications, e.g. paint programs, need to see all motion Denis> events and will benefit from turning off event compression. Denis> By default, event compression is enabled. Denis> ======== Denis> We are a paint program and need to see all motion events :) Denis> I'm having a horrible two weeks at work and will acquire a new child in Denis> about a month, so this is not the right time for to engage in any Denis> serious xournal-related work, but my suggestion would be to see if just Denis> disabling event compression helps with stroke smoothness. If it fixes Denis> smoothness but causes display lag issues, then you can go back and Denis> complain about the rendering performance again :-) Denis> Best, Denis> Denis Denis> On 03/30/2015 01:09 PM, D M German wrote: >> Bernhard Reiter twisted the bytes to say: >> >> Bernhard> I've found a thread which seems to indicate that Daniel found the reason Bernhard> for the performance issues and might've resolved them afterwards [1] -- Bernhard> hence my questions if it might be time to start an "official" gtk3 Bernhard> branch and release an alpha into the wild for further testing. After Bernhard> all, Daniel's work now almost 1.5 years old -- would be too bad to just Bernhard> leave it lying around. >> Bernhard> (CC'ing Daniel) >> >> it is actually more complicated than that. >> >> I started looking at the gtk3 branch of xournalpp (it is now in working >> condition) and it suffers the exact same problem than my code using >> goocanvas. What is special is that xournalpp does not use any >> library. it implements rendering directly. >> >> So I started digging and so far I think this is the most promising lead: >> >> See this and the responses: >> >> https://mail.gnome.org/archives/gtk-app-devel-list/2015-March/msg00074.html >> >> I am still following up with it. Denis, you are a much better expert at >> the event handling than I am, so your feedback might be valuable. >> >> -dmg >> >> Bernhard> Bernhard >> Bernhard> [1] Bernhard> https://mail.gnome.org/archives/goocanvas-list/2013-November/msg00007.html >> Bernhard> On 2015-03-27 18:14, Denis Auroux wrote: >> >> As far as I understand, there are still performance issues on larger >> >> documents, caused by the gtk3 canvas libraries being much less efficient >> >> than libgnomecanvas. Or have they been fixed? (I haven't followed >> >> closely). >> >> >> >> -- Denis >> >> >> >> On 03/27/2015 01:12 PM, Bernhard Reiter wrote: >> >>> Hi, >> >>> >> >>> I've tried out Daniel German's Gtk3 port [1] of xournal, and found it to >> >>> work quite well -- everything that I normally use xournal for seemed to >> >>> work. I was curious if there were plans to merge his branch into the >> >>> official repo any time soon and/or maybe release an alpha version for >> >>> public testing, as the port has been around for a while. >> >>> >> >>> Regards >> >>> Bernhard >> >> >> -- >> Daniel M. German "The greatest dangers to liberty >> lurk in insidious encroachment >> by men of zeal, >> well-meaning >> Louis Brandeis -> but without understanding." >> http://turingmachine.org/ >> http://silvernegative.com/ >> dmg (at) uvic (dot) ca >> replace (at) with @ and (dot) with . >> >> >> Denis> -- Denis> Denis Auroux Denis> UC Berkeley, Department of Mathematics Denis> 817 Evans Hall, Berkeley CA 94720-3840, USA Denis> aur...@math.berkeley.edu Denis> ------------------------------------------------------------------------------ Denis> Dive into the World of Parallel Programming The Go Parallel Website, sponsored Denis> by Intel and developed in partnership with Slashdot Media, is your hub for all Denis> things parallel software development, from weekly thought leadership blogs to Denis> news, videos, case studies, tutorials and more. Take a look and join the Denis> conversation now. http://goparallel.sourceforge.net/ Denis> _______________________________________________ Denis> Xournal-devel mailing list Denis> Xournal-devel@lists.sourceforge.net Denis> https://lists.sourceforge.net/lists/listinfo/xournal-devel -- Daniel M. German "A coin symbolizes our free will" El ZahÃr, Jorge Luis Borges http://turingmachine.org/ http://silvernegative.com/ dmg (at) uvic (dot) ca replace (at) with @ and (dot) with . ------------------------------------------------------------------------------ One dashboard for servers and applications across Physical-Virtual-Cloud Widest out-of-the-box monitoring support with 50+ applications Performance metrics, stats and reports that give you Actionable Insights Deep dive visibility with transaction tracing using APM Insight. http://ad.doubleclick.net/ddm/clk/290420510;117567292;y _______________________________________________ Xournal-devel mailing list Xournal-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xournal-devel
