This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] A few ppc assembler fixes
- From: Alan Modra <amodra at gmail dot com>
- To: Peter Bergner <bergner at vnet dot ibm dot com>
- Cc: Binutils <binutils at sourceware dot org>, Bill Schmidt <wschmidt at linux dot vnet dot ibm dot com>, Hal Finkel <hfinkel at anl dot gov>
- Date: Thu, 23 Apr 2015 14:43:31 +0930
- Subject: Re: [PATCH] A few ppc assembler fixes
- Authentication-results: sourceware.org; auth=none
- References: <1429746727 dot 21947 dot 60 dot camel at otta> <20150423031444 dot GM12627 at bubble dot grove dot modra dot org> <1429761126 dot 21947 dot 74 dot camel at otta>
On Wed, Apr 22, 2015 at 10:52:06PM -0500, Peter Bergner wrote:
> On Thu, 2015-04-23 at 12:44 +0930, Alan Modra wrote:
> > If I'm not mistaken the dcbt/dcbtst change will affect the following
> > -m options.
> >
> > -m601 +server
> > -mppc +server
> > -mppc32 +server
> > -m603 +server
> > -m604 +server
>
> Yes, although these old cpus only ever used the two operand
> version.
>
>
> > -m403 +server
> > -m405 +server
>
> My bad, I thought these were caught with the BOOKE usage.
> I'll have to move these back to the embedded side.
>
>
>
> > -m7400 +server
> > -m7410 +server
> > -m7450 +server
> > -m7455 +server
>
> These only use the two operand version of the instruction, so are
> probably a don't care. However, I should probably move these
> back to the embedded side too.
>
>
> > -m750cl +server
>
> This also only uses the two operand version of the instruction,
> but since this is an IBM part, it's probably safest to use the
> server ordering.
>
>
> > -mppc64 +server
> > -m620 +server
> > -mppc64bridge +server
>
> Yes, I think we want server ordering here too.
>
>
> > -me300 +server
>
> Another case I thought BOOKE would catch, but didn't. I'll change it.
>
>
> > -me5500 -server
> > -me6500 -server
>
> These are embedded processors, and they use the three operand
> version of the instruction and expect the embedded ordering,
> so yes, these are correct.
>
>
> > -mspe +server
>
> This is more an instruction category rather than a cpu, isn't it?
> That said, it's probably safest to lump this in with the embedded
> cpus, since that's where it seems to be used. Do you agree?
Yes. I probably shouldn't have included that option in my tests.
BTW, the following is what I used
for z in `cat ppcgasopt`; do echo "$z"; gas/as-new $z -mregnames dcbt.s && objdump -sj.text a.out | grep -q 7d09 && echo " server"; done
ppcgasopt is a selection of -m options taken from tc-ppc.c, and dcbt.s
dcbt r9,r6,8
--
Alan Modra
Australia Development Lab, IBM