[cairo] Pycairo and ImageSurface

[Glenn Murphy]

> Now, if that's _still_ too slow, then the next thing to address is
> perhaps in the X server itself, (ideally, it would be able to use the
> video hardware for the rotation, and not require much CPU at all). So,
> it's a question of how good the support is for rotated images in your
> server's implementation of the Render extension.

Switching to use an Xlib surface didn't really help.  So you're saying that
it's a problem with my X server setup?  Cairo itself should be using the
Render extension already?  Sorry for the simple questions, I'm still new to
cairo.

> Are you using a cairo snapshot between 1.3.2 and 1.3.12?

Yep, that was the problem.  Thanks a lot for the help.


On 2/20/07, Carl Worth <cworth at cworth.org > wrote:
>
> On Tue, 20 Feb 2007 16:03:05 -0600, "Glenn Murphy" wrote:
> > When I added the rotation is when I first hit this problem. I was
> rendering
> > everything on each refresh, so the application was sucking up a lot of
> CPU
> > time. I changed it to draw everything to a ImageSurface the one time and
> > then paint that to the screen when the expose-event was signaled.
> Whenever
> > the expose-event signal is received, all it does is create a cairo
> context
> > (using gtk.gdk.Window.cairo_create()), translate to the center, rotate 1
> > degree and paints the ImageSurface. However, after doing all that, it
> still
> > sucks up a lot of CPU time from redrawing the DrawingArea all the time.
> > expose-event is getting signaled a lot (about 30 times a second) and
> each
> > time it has to redraw the entire DrawingArea.
>
> If you're painting an image surface every time, then you will be
> transferring all of the image data from your application to the X
> server every time.
>
> So, better would be to have your intermediate surface as an xlib
> surface, (or whatever surface type your destination is). You can
> create one of those with cairo_surface_create_similar. Given a
> cairo_t* cr from gdk_cairo_create that would be something like:
>
> surface = cairo_surface_create_similar (cairo_get_target (cr),
>                                         CAIRO_CONTENT_COLOR[_ALPHA],
>                                         width, height);
>
> Hopefully it's obvious to you how to translate that from C to python.
>
> Now, if that's _still_ too slow, then the next thing to address is
> perhaps in the X server itself, (ideally, it would be able to use the
> video hardware for the rotation, and not require much CPU at all). So,
> it's a question of how good the support is for rotated images in your
> server's implementation of the Render extension.
>
> > Another problem I've been having is from rotating the ImageSurface.
> Whenever
> > I rotate it, the image created is moved outside of the clipping region.
> > Every 90 degrees, it moves back to the correct position. I haven't been
> able
> > to find anything about this problem. I have three images showing this
> > problem:
>
> Are you using a cairo snapshot between 1.3.2 and 1.3.12? That looks an
> awful lot like a bug that was fixed in the 1.3.14 snapshot which was
> described as follows in the release notes:
>
>         Fix corrupted results when rendering a transformed source image
>         surface to an xlib surface. This was a regression that had been
>         present since the 1.3.2 snapshot.
>
>         [ http://cairographics.org/news/cairo-1.3.14]
>
> I hope that helps. And have fun with cairo,
>
> -Carl
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freedesktop.org/archives/cairo/attachments/20070221/8ff2e5dc/attachment.htm