On Thu, Feb 05, 2004 at 12:08:55AM +1100, Benjamin Herrenschmidt wrote:
> On Wed, 2004-02-04 at 22:01, Guido Guenther wrote:
> > On Wed, Feb 04, 2004 at 10:09:05AM +1100, Benjamin Herrenschmidt wrote:
> > > Can you go the whole way and kill the switch/case for arch too ?
> > Hmmm. I'd rather not. riva_get_arch() seems reliable (and we can keep in
> > sync with XFee86 easily so we don't trip on wrong RIVA_ARCH bugs as I
> > did recently). I don't see any additional benefit from encoding these in
> > the pci_dev structure. Having information such as 2 Crts, etc in there
> > would be very good though.
>
> The benefit is having a single list and so less crap in the driver...
I intend to get rid of all references to PCIIDs in the driver except for
nv_get_arch. Other options will be encoded in the pci_dev struct.
> > While were at this: is there any benefit of having riva_chips encoded in
> > pci_dev? We have the riva.{Architecture,Chipset} already. So can we kill
> > this and keep the room for something "useful" in there, thereby removing
> > the need of keeping riva_chips and rivafb_pci_tbl in sync?
>
> The whole riva_chips is shit and should be killed. My idea was to replace
> it with the architecture.
Done that already in my tree. I'll send a new patch with this and some
other cleanups once I synced with James.
Cheers,
-- Guido
Attachment:
signature.asc
Description: Digital signature