Subject: Re: mods for proposed port 'tsarm'
To: Jesse Off <joff@embeddedARM.com>
From: Nathan J. Williams <email@example.com>
Date: 12/06/2004 13:31:40
"Jesse Off" <joff@embeddedARM.com> writes:
> > Would it have any advantage over evbarm other than avoiding the
> > slightly disparaging tone of the phrase "evaluation board"? I don't
> > think that warrants creating a new port.
> Probably not a lot, but like was brought up earlier, there are a lot of
> other boards we may want to support now and in the future that are also
> not evaluation boards. Check out a copy of PC104 Embedded Solutions
> magazine and thumb through the pages of embedded systems vendors to see
> what I mean.
Right, I'm aware of them; I've done a port to a pc104 ARM system for a
customer, and I made it an evbarm port.
> Another thing worth mentioning is that the Cirrus processors it uses does
> have a non-standard ARM FPU coprocessor that the latest GCC toolchains
> 3.4+ do support. Conceivably, you would want an appropriately compiled
> userland to take advantage when the toolchain catches up. Would a port be
> appropriate if it were to the Cirrus ARM variant, e.g. "cirrus", "clarm"
> (cirrus-logic arm)?
This is an interesting question. Right now evbarm is all little-endian
and (I think) no FP. Other ABIs for eval board systems, with FP of
some form, or big-endian versions, are concievable, and it's not clear
where they'd go in the current universe (exacerbating the problem that
port->ABI is considered to be a many-to-one mapping). We'd want a
better way to say "This kernel, but that userland".