Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 14 May 2006 10:56:14 -1000
From:      Jim Thompson <jim@netgate.com>
To:        gnn@freebsd.org
Cc:        freebsd-small@freebsd.org
Subject:   Re: Embedded FreeBSD Presentation...
Message-ID:  <141C7316-5D6A-4582-A609-3C8AA341BF94@netgate.com>
In-Reply-To: <m2slncsj91.wl%gnn@neville-neil.com>
References:  <6.2.3.4.0.20060514091549.01e2a4d8@pop3.retena.com> <m2slncsj91.wl%gnn@neville-neil.com>

next in thread | previous in thread | raw e-mail | index | archive | help

On May 14, 2006, at 5:45 AM, gnn@freebsd.org wrote:

> At Sun, 14 May 2006 09:16:48 +0200,
> Eduardo wrote:
>> Is this really possible:? Sorry for be a bit unoptimist, but in the
>> embeded space the o.s.must follow some rules that freebsd (nor
>> linux, *bsd, windows, etc..) can't:
>>
>> - size: The o.s. must be minimal. Freebsd kernel, now, is a bit huge.
>> - realtime: The o.s. must do some tasks at fixed times, this tasks
>> can't wait for nothing.
>>
>> The size one can be fixed, but the realtime not. It needs a new
>> scheleude, irq manager,...; so a great kernel rework.
>
> Perhaps you mis-understand the thrust of this drive, in that we are
> not talking about turning FreeBSD into an RTOS, at least not in the
> short term, but about making it more amenable to embedded.
>
> There is much to do but it will be a gradual process.  Size and
> configuration are the first things to address.
>
>> Also, you forget the PowerPC chips. They are in a lot of embedded
>> devices and now freebsd has support for them (6.x).
>
> We did not forget them but amongst those who have shown interest in
> this project ARM and MIPS are the clear leaders.  If we find people
> who wish to address the PowerPC chips as well, all the better.  One
> other important component in this work is focus.  We cannot be all
> things to all people, at least not at the outset, so two different
> processors and two or three reference boards for each are where we
> plan to start.

In the embedded world, most people who build their own "SOC" tend to  
use ARM7, ARM9 or MIPS cores.
For instance, Atheros and Broadcom both use a MIPS 4KC core in their  
WiFi parts.   Realtek uses an LX50 core, which is "MIPS minus the  
patented instructions" (there is a port of GCC which won't issue  
these instructions).

TI uses ARM7 in OMAP, but that platform is probably "too small" for  
FreeBSD's resource requirements in the immediate future.  (Read: lots  
of work to squeeze FreeBSD onto platforms using OMAP.)

Folks who are not ODMs of one of these companies tend to use PowerPC  
(either IBM's 440 or one of the Freescale parts) or Xscale.   Intel  
did a real "number" on the market starting in 2003 (similar to what  
they did to the WiFi chip market with Centrino) and essentially drove  
suppliers of MIPS core parts (such as IDT) into a corner of the market.

That said, there are very few "commodity priced" PowerPC boards.

I think the most "bang for the buck" will be found porting FreeBSD to  
Xscale (especially the ixp42x parts) first, followed by PowerPC  
(because FreeBSD is already running on a (slightly different) PowerPC  
chipset), and then the various MIPS-based parts that support WiFi or  
the little switches-with-embedded-MIPS parts.

Jim



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?141C7316-5D6A-4582-A609-3C8AA341BF94>