On Fri, 21 Jun 2002, Dr Andrew C Aitchison asked:
> >Is there enough pattern to the ATI PCI chip ids to guess which
> >driver to use for unknown chips?
On Sat, 22 Jun 2002, Mike A. Harris wrote:
> The ATI hardware documentation provides enough information to
> know what family (Mach64/R128/Radeon R100/RV100/R200/RV200)
> wether or not it is a Mobility chip, and which one, etc. I
> haven't noticed anywhere in the docs which show the actual board
> names used to market the video hardware such as "Xpert@Work 98",
> "FireGL 8700", etc. A fair number of these ID's we can
> autodetect and point to the right driver, since if XFree86
> supports the card, then it has the PCI ID listed internally, and
> we can just use that info to provide the right driver.
>
> If I'm unsure if a card will work or not, and don't have one, I
> look at the docs, and the code, and try to add support that will
> hopefully work (and mostly has so far). However I may not know a
> given card is a "ATI Radeon 7200" per se. (random example). So
> users wont actually see "ATI Radeon 7200" show up as
> autodetected currently, they'll see "ATI Radeon QD" or whatever
> it happens to be.
>
> By the way... what is the "lspci -vn" output of a Radeon 7000,
> 7200 so I can make these autodetect with proper names. ;o)
My Radeon 7000 gives this:
01:00.0 Class 0300: 1002:5159
Subsystem: 174b:7112
Flags: bus master, stepping, 66Mhz, medium devsel, latency 32, IRQ 11
Memory at e8000000 (32-bit, prefetchable)
I/O ports at c800
Memory at ff8f0000 (32-bit, non-prefetchable)
Expansion ROM at ff8c0000 [disabled]
Capabilities: [58] AGP version 2.0
Capabilities: [50] Power Management version 2
Have you come across
http://www.yourvote.org/pci
? While the data is supplier by its readers, I've found it reliable
as far as it goes.
> >For example we could have guessed that the 1002:5148 would user
> >the same driver as 1002:5144 - 1002:5148. When we add the ID to
> >the hardware database it isn't as if we actually change the
> >driver in any way.
>
> Well, when it is added to XFree86 xf86PciInfo.h, the driver does
> need to change, so whoever is making those changes needs to know
> if it is a Radeon or whatever, or what driver it should get added
> to, as the conditional code paths in the given driver will need
> to be updated for the new card.
Ah yes.
I was hoping for a pattern to the PCI IDs which would allow us to
guess a similar chip ID to use when we encounter a new ID.
Most of the changes to the atimisc and r128 drivers for new IDs
just added the ID to an existing code-path, but I see that the
for Radeon, some of the additions have had more content, and simply
guessing and using a similar ID isn't really good enough.
Even so, it would be nice if instead of a message saying
"unsupported chip/card found, please send the output of 'lspci -vn'
to ... ", the server could guess a plausible line like
ChipId 0xWXYZ
to add to the config file, or even try it automatically.
We would still want the user to email us with the details so that we could
add proper support, but it would get users going more quickly.
> I'm looking into getting a more proper official list of the PCI
> ID's and the marketing names they map into, so users will see the
> name of their actual card as it says on the box rather than
> seeing "ATI Radeon QW" or similar in dialogs, etc.
I suspect that the name on the box is more closely related to the
"Subsystem" ID (which /var/log/XFree86 calls the "card" ID) than
to the chip ID. It is hard to tell since many of my ATi cards are
"white box", but I've found that with ATI the mapping between
pci chip ID and the marketing name is many <-> many :-(.
--
Dr. Andrew C. Aitchison Computer Officer, DPMMS, Cambridge
[EMAIL PROTECTED] http://www.dpmms.cam.ac.uk/~werdna
_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert
