[Date Prev][Date Next][Thread Prev][Thread Next][Thread Index]

RE: [XaraXtreme-dev] Move from CDraw to AGG

I don't have the code (it's not dead easy to release it now), but I do
know the tests were pretty simple. We just rendered, repeatedly, a fixed
bezier filled shape (in fact this shape or something very like it,
attached), and timed how long it took to render it say 20,000 times on
top of itself into an off-screen bitmap.

We then compared various fill styles, flat colour, graduated fill (like
the attached) and with or without a stroke and with and without

So the tests of the different engines was pretty simple. I recall that
we found AGG to be quite a lot faster than Cairo (but this was probably
around the time of Cairo 1.0). So these tests were for the raw engine
speed, nothing to do with the drawing program (we have other tests for
that, that are just as easy - replicate said shape 20,000 times, and
then just time how long it takes to redraw. That's how we test basic
performance against Illustrator and other drawing programs).


> -----Original Message-----
> From: owner-dev@xxxxxxxxxxxxxxxx 
> [mailto:owner-dev@xxxxxxxxxxxxxxxx] On Behalf Of Carl Worth
> Sent: 28 February 2007 17:20
> To: dev@xxxxxxxxxxxxxx
> Cc: Alex Bligh
> Subject: Re: [XaraXtreme-dev] Move from CDraw to AGG
> On Wed, 28 Feb 2007 09:43:54 +0000, Luke Hart wrote:
> > AGG is currently much more efficient than Cairo (my simple tests 
> > indicated getting on for an order of magnitude), so may 
> well be better 
> > in the short term.
> Could you provide some of those tests to the cairo community, 
> (or even just point them out to me here, or privately).
> There was a Mozilla-sponsored Summer of Code project last 
> year to implement an AGG backend for cairo---with the hope 
> being that any performance benefit that AGG has could become 
> part of cairo itself.
> That backend did become functional, but the timings suggested 
> it wasn't significantly faster than the software in cairo 
> that it was replacing.
> So I'm curious to see why you came up with such different 
> results in your comparison.
> -Carl

Attachment: shape.png
Description: shape.png