2006/10/24, Lucas Nussbaum <[EMAIL PROTECTED]>: > Hi, > > What I'm reporting is known as bug #347225. I reported it a few months > ago, while using a PCI ATI graphic card in my setup to be able to get dual > head to work using my laptop's dockstation. > > What I reported back then is that switching between tabs in epiphany > "felt" a lot slower than in firefox. The problem only appeared when > using the 'ati' driver with the PCI card (which can be assumed to be > slow, even if it's quite recent). It wasn't noticeable when using the > fglrx driver with the laptop's internal video card.
The 'ati' driver has poor or no acceleration for drawing, the fglrx driver has poor or adequate acceleration for drawing. I would say it's expected for a less accelerated drawing to take longer than more accelerated ;) That doesn't totally rule out software probelms of course. > Even if I cannot reproduce the problem anymore (I changed my setup to > avoid using the PCI card), I'm still interesting in understanding it. > Also, somebody commented today on the bug, reporting a similar issue > when switching from the nvidia proprietary driver to the 'nv' driver. Same comment applies to 'nv' vs 'nvidia' drivers I guess. > So, my questions are: > - is this a known problem ? Rather ask at this point "what is the actual problem". > - whose fault (component-wise) is it ? (I'm not really familiar with > GNOME development) > - what are the likely causes for this ? Most likely the mozilla engine, but if the results differ much between ephy and ff... Could be some difference in caching the content? > - is this going to be improved ? Only if someone knows what's actually going on. > Others related questions: > - are people working on benchmarking the gnome frameworks with slow > hardware ? I do use a slowish laptop to benchmark stuff, though I'm not doing it as much as I would like... (and only on components I tend to use). > - Are there software tools to help diagnose such issues ? (based on > traces analysis maybe ?) Sysprof is a pretty easy-to-use one, though actually getting the debug symbols which are required for meaningful output might be harder depending on your platform... > - Are there software tools allowing to emulate a slower video card to > make the detection of such issues easier ? Use 'svga' as the X driver or use remote connection to the X come to mind. Though, one should be aware that it's not simply "this card is fast" vs "this card is slow" when you start talking about acceleration, specially if there is stuff that can be done entirely on the GPU without software getting in the way (like you see in the recent QT benchmarks. XRender, though accelerated in certain ways, got smacked in the head by OpenGL since the OpenGL implementation could basically just dump the stuff to GPU and let it do the job). So it's also crucial to know what the cards may support, simply pushing pixel data to the card as fast as possible is not always the fastest way. -- Kalle Vahlman, [EMAIL PROTECTED] Powered by http://movial.fi Interesting stuff at http://syslog.movial.fi _______________________________________________ Performance-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/performance-list
