Subject: Re: Relevance of ARM merge (was: Re: port-arm26 and port-arm32 into
To: Chris Gilbert <chris@paradox.demon.co.uk>
From: Reinoud Zandijk <imago@kabel065011.kabel.utwente.nl>
List: port-arm32
Date: 02/14/2001 13:27:05
Hi Chris,

On Wed, 14 Feb 2001, Chris Gilbert wrote:
> Hmm, SMP Xscale boards anyone :)  (however that'd take a lot of work to pull
> the kernel core up to a level to support smp on the arm.)

a *lot* of work i must say :( but a chalange too :)

> The effort is to provide support for ELF across all arm platforms.
> Bearing in mind that this will bring ELF onto arm32 I see that as a
> plus.  Currently Ben seems to be doing a fair amount of work on this
> front (I believe that others would help but lack the time, knowledge
> or other commitments stops them), and he's the arm26 port-master.
> Perhaps you've also missed the fact that ben has done some
> improvements to both the arm32 and arm26 ports to actually merge them
> together.

I agree... the arm32 port is one of the few if not the only one to still
have a.out instead of ELF and this is breaking more and more stuff that
wants to link dynamically etc...

> For what purpose, I've seen no sign of any new architectures being
> merged into the main NetBSD tree, I believe that someone has SA1100's
> running NetBSD, and there maybe other devices.  There's been talk of
> moving the different architectures into their own sys/arch/ dir.  But
> nothing concrete has actually happened.

Well I would like these SA1100 patches/support to be included into the
source tree for other ppl would love to work on those too and it makes the
user base a lot bigger for there are lots of small SA1100 boards around
esp. around universities nowadays and they all run Linux :(

> Perhaps ben should be considering if we wants to merge with a port that's not
> been seriously worked on for at least a year?  If you look at the current
> state of mainline there's 2 platforms where the current kernel fails to run,
> it runs out of VM space.  There's a patch for this sat in the PR system,
> along with some other patches from Mike to do with networking.

Would like to have a look at those again ... are the patches reliable
enough now to keep the system going ? i.e. you havent noticed any oddities
(anymore) ?

> Most people aren't interested in the NetBSD/arm32 port now, linux is
> the one with all the hype and all the cool new toys ported to.
> Having said that though we still sell a few CD's at shows.  They
> recently announced they've got linux/arm running on an xscale eval
> board, they've got ipaq support etc At one time NetBSD/arm was ahead
> of linux in it's support for hardware.  At one time we even had boards
> that were ahead of Risc-Os.  The CATS added PCI etc to a SA, and now
> 2-3 years later we've got companies doing RISC-OS systems based around
> similair hardware.

Wouldn't say it that strong but lots of ppl are working on Linux yeah ...
and some even started at a FreeBSD port ... so we have lots to do :) The
fact that the arm32 port is getting more and more speed the last couple of
weeks is very promising and will hopefully improve the interest in it
more.

> Hmm, this seems to have turned into a personal moan about the state of the
> arm32 port, which wasn't my intent, but perhaps just me showing that I'm
> frustrated with the current state of things.

Well I've allways enjoyed your feedback Chris :) ... please keep doing so!
.. I'm currently working on the Kinetic card support (wich is progressing)
to give Kinetic machines CATS performance and working on a small
modification that also allows the new bootloader to work on RISC OS 3.7
wich seems to be different in the OS_Memory call... (snif)

If all goes well, I'll be starting Xscale support in not to long a time
and hopefully it wont take long. The main challenge / design choise is the
amount of compatibility... i.e. how to use the special features and
optimalisations for this processor.

As a side project i plan to work on a small patch that allows ARM6/7
RiscPC machines to be able to reboot again :)

Cheers,
Reinoud