On Sat, 16 Dec 2000, Reinoud Zandijk wrote:
> Hi Brian and  Kyle,
>
> On Sat, 16 Dec 2000, Chris Gilbert wrote:
> > I believe that we don't currently support the Xscale chips.  Probably
> > simply because there's no systems with them (or at least I know of plans
> > for some, but none that actually use them).
>
> thats right... i dont know any other system next to Millipede's
> forthcoming Imago and the Microdigital Omega with Xscale option ...

Nod, has anyone been in contact with either company?  I know that a while 
back I tried one of them, but they never replied.  I'd certainly like to see 
us getting a port onto decently specced arm hardware (comparing this to the 
RiscPC)  Currently I need to talk to sort out the netstation thing. (I've 
just not had time!)

> > However they are being considered for the plans to make arm platforms use
> > ELF at some point in the not too distant future. I'm not sure how much
> > work is required for Elf support, I believe that toolchain and kernel
> > work is needed. There was some brief discussions that we need to change
> > the ABI as well so that we use something more sane, and that'll work with
> > Thumb chips as well, oh and have the same binaries for arm26 and arm32
> > would be good. :)
>
> Dont really see the connection to ELF really... i mean, its just another
> processor... not really that huge a port ... unless we want to change the
> FP code to more suit the capabilities of the ARMv5 instruction set etc...

I'm sure there was something about the alignment requirements of something 
(maybe it was thumb not xscale...)   There was talk on one of the tech lists 
of creating libraries that included the arcitecture type, so you could have 
an armv4 optimsed lib, etc.  But I think it might have to be more based on 
available features, eg RiscPC's can't do the half work loads etc.

> > It's also a matter of time for these things, particularly developer time.
> >  Oh and getting developers communicating.  Currently we seem to have 3
> > different bootloaders floating around, the one that Reinoud has done (not
> > looked at it yet) the original BtRiscBSD done in C and the arm26 BBBB,
> > which seems a little bit mad, (oh and technically the other open firmware
> > boot systems possibly counts as another bootloader, but we don't have
> > much control over that)
>
> Well i must say that, as the author of one of them, there was
> communication about it ... at least with Ben Harris about the BBBB
> bootloader for arm26 ... and i consulted others about the 32 bootloader
> wishes... Pity it still isnt fully ok now since it still wont load
> debugged kernel (i.e. kernels with the inkernel debugger swicht on - dont
> know why as yet... but have no access to check it now ....) ... but at
> least  we now also got the NC's booting!!!!!!

Nod, I've not had time to look yet (just finished a week doing Panto.)  Have 
you had much time to look into the debugger thing?

> My new bootloader is more a rewrite from scratch of the origional arm32
> bootloader for that one suffered a dedicated link between RISC OS version
> and  caused a lot of hairy code in the kernel at boottime .... if you're
> interested see the difference between teh rpc_machdep.c between the two.

Are you planning on doing a front end of some kind?  Perhaps it's possible to 
retro fit the new bootloader with the old front end (although that does 
require a copy of the Acorn C compiler, unless gcc on riscos would do the 
trick)

> Hope to  have given some clarity .... i hope :)

No probs, I'm just trying to get communication going on the arm32 list, it 
was fairly quiet for a long long time, I can't code (It's a contractural 
thing, as aposed to lack of skill)  but I can at least attempt to make sure 
we don't duplicate work between different people.

Cheers,
Chris