On Thu, 3 Apr 2008 17:47:05 -0300, "bulia byak" wrote: > (Sorry but that reminds me of your claims of outperforming Cairo by > some insane margin, but when Carl Worth asked you for specifics of > your testing, AFAIK you never came up with the real data.) Since my name got mentioned here, I should say that I don't think that's quite accurate. I didn't ever see the actual benchmark program, but Charles did point me to people, (he hadn't run the benchmark himself, unsurprisingly), who did describe what the benchmark was actually testing. There are definitely things that are slow in cairo, (particularly in its software fallbacks). We try to avoid using those, (preferring to get hardware-accelerated performance through the X server, etc.---but even then things aren't always don in the X server either). And sometimes you really do need to use the software backend. We've got our own benchmarks on the cairo list showing that liboil code can do much faster compositing than what's currently in cairo/pixman and several experimental scanline-based polygon rasterizers, (presumably similar to the approach in Xara), that are much faster for software rendering than cairo's current tessellate-then-rasterize-trapezoids approach. So I'm not immediately dismissive of the benchmark results. There's no doubt that the Xara demos I saw running were impressively fast. But sure, I would have preferred to see the actual benchmark source, (and even more I would have loved to see all the Xara source). I still would love to be able to "apt-get install xaralx" just like I can "apt-get install inkscape". Oh well. -Carl
Attachment:
pgpV1o6fQlQEV.pgp
Description: PGP signature