On Sun, 2006-01-08 at 10:43 +0000, MJ Ray wrote:
Thanks for the explanations. So, if I understand correctly, Cairo is in a similar space to Display PostScript and Display PDF, but uses GL to draw and use the graphics processor as much as possible.
Yeah, you're basically spot-on, although Cairo tends to be used with software renderers at the moment (the GL stuff is newer, and where people want to go with it).
[...] Are there similar sites to r300.sf.net for other manufacturers, or any overview site?
No, not really, it's a mess. [...]
Are others willing to help create one? I think I could program it fairly quickly.
I think others have tried; but I guess previous efforts have tried to be a general db on all hardware. A specific graphics hw site could work, I don't think it's really been tried.
The large issue is the model name -> hardware mapping; it's basically a many-to-many relation (a given model could have a number of different chips inside; and vice-versa - it's basically a silicon yield thing). E.g., the Radeon naming system: http://dri.freedesktop.org/wiki/ATIRadeon - check out the X series in particular (as chips as more expensive to produce, and yield goes down, there is a tendency to use high-performance devices in cheap hardware when the device has to be clocked down/is partially faulty in some way, if you see what I mean). Different chips in the same model are not always operable by the same driver.
The other issue is the non-brand boards; you buy a board with ATi/NV chips or whatever, but it's made by some Taiwanese manufacturer or something. Though the chips are the same, they are sometimes wired up in different ways - doesn't usually affect the 3D support per se, but can mean that you can't run the card in certain configurations (eg., dual head, TV out, DVI support or other hardware-y features) with standard free software drivers. Sometimes you just need to fix the PCI ID db, sometimes it's more work are requires a patch to the driver (as my Radeon 9200 SE needed to get DVI support).
It probably isn't worth trying to capture all those permutations though; probably worth concentrating on capturing the basic 3D support and (what would be good would be) a performance metric. I'm sure plenty of people out there would be willing to help, me included.
Cheers,
Alex.