On Tuesday, 11 March 2014 07:29:50 AM Paul Emsley wrote:
> On 11/03/14 06:29, Ethan A Merritt wrote:
> > On Sunday, 02 March, 2014 16:17:22 Ethan Merritt wrote:
> >> I recently acquired a new machine and found that my previous
> >> coot executables will not run on it, and neither will the
> >> stable or "nightlies" built for Fedora/RHEL or Ubuntu.
> >> No problem, I thought, I'll just rebuild it from current source.
> >>
> >> That worked, but the graphics display speed of the resulting
> >> executable is very slow. I have not noticed rendering speed
> >> issues with any other programs I am running.
> >>
> >> The recent mention here of buggy Mesa versions makes me wonder
> >> if that is what I'm suffering from.
> >> I can't find any likely pointers on the web.
> >> Does anyone have more detail?
> >> Is there a work-around?
> > Following up on my own problem report....
> >
> > I eventually poked something (don't ask me what!) in one of several
> > graphics configuration tools that made a large difference.
> > As best as I can tell by comparing the Xorg.0.log files before and
> > after, the difference is this:
> >
> > Xorg.0.log:[1690259.966] (==) AIGLX enabled
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled GLX_MESA_copy_sub_buffer
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled GLX_INTEL_swap_event
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled GLX_ARB_create_context
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled GLX_ARB_create_context_profile
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled
> > GLX_EXT_create_context_es2_profile
> > Xorg.0.log:[1690259.978] (II) AIGLX: enabled GLX_SGI_swap_control and
> > GLX_MESA_swap_control
> > Xorg.0.log:[1690259.978] (II) AIGLX: GLX_EXT_texture_from_pixmap backed by
> > buffer objects
> > Xorg.0.log:[1690259.978] (II) AIGLX: Loaded and initialized i965
> >
> > So enabling AIGLX makes a huge difference.
> > Unfortunately I can't find any changes in xorg.conf or anywhere else that
> > would explain why it is now enabled whereas before it wasn't.
> > Anyhow I now get very fast rendering of lines, shaded objects
> > (ball-and-stick) etc.
> >
> > HOWEVER....As soon as even a single label (atom name or distance)
> > appears on the screen it goes back to being slow. Clearing the labels
> > makes it fast; clicking an atom makes it slow; 100% reproducible.
> > So it really doesn't like rendering characters.
> > Any ideas on what else I might try?
> >
> >
>
> The font rendering of Coot has been niggling me for a while. I have
> noticed a slowdown in the FPS when labels are added, but not perhaps as
> much as you suggest.
>
> The standard coot spin test is to load the tutorial data and map, zoom
> in until the map touches the edge of the screen, turn on the
> Frames/Second counter and Draw -> Spin View and watch the status bar.
Following that recipe:
Tutorial map spin in square window with full screen height: 16.6 frames/sec
Click on one atom to display label: 1.55 frames/sec
So more than a factor of 10 speed difference with/without text.
Ethan
