This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: GAS: Handling option parsing for ARM co-processor extensions
- From: Richard Earnshaw <rearnsha at arm dot com>
- To: Phil Blundell <pb at nexus dot co dot uk>
- Cc: Richard dot Earnshaw at arm dot com, binutils at sources dot redhat dot com
- Date: Tue, 15 Jan 2002 13:03:09 +0000
- Subject: Re: GAS: Handling option parsing for ARM co-processor extensions
- Organization: ARM Ltd.
- Reply-to: Richard dot Earnshaw at arm dot com
> On Tue, 2002-01-15 at 12:12, Richard Earnshaw wrote:
> > Well, not really, since AFAICT from the Cirrus web pages, Maverick is just
> > a co-processor extension on a CPU. There's nothing much to tie it to a
> > particular processor, so using -mmaverick to imply a particular CPU looses
> > flexibility.
>
> Um, yeah, I didn't make that very clear. I meant that you would end up
> with "-marmv4 -mmaverick" or something like that, along the lines of the
> "-marmv3 -mfpa10" that we have now. XScale was a dumb thing to pick as
> an example.
I don't like that, for one because, as I said to Lars, you have to parse
the entire command line before you can know that you have a valid
combination of flags. Plus, how do you turn Maverick off if that is one
of the default flags set up for a build? The only reasonable way is to
start adding -mno-maverick etc, which becomes clumsy. IMO, selecting the
instruction set to assemble for should be just one flag that allows you to
fully describe what you want: -m<base-cpu>+<option> allows for that.
>
> > Incidentally, I'm a bit concerned about -mxscale as a processor. From the
> > documents I've read XScale seems to be more of a specification for an
> > architecture than a single implementation.
>
> Yes, I think that's right. I'm not sure that the distinction is
> particularly important right now, though.
Hm, except that when it does become important there will be a much larger
user base, so the resistance to change will become greater. However, I'm
not sure what I would rename it to...
R.