Re: [MIPS] Add saa and saad instructions for octeon

[Richard Sandiford <rdsandiford at googlemail dot com>]

Andrew Pinski <andrew.pinski@caviumnetworks.com> writes:
> On Tue, Nov 15, 2011 at 10:39 AM, Richard Sandiford
> <rdsandiford@googlemail.com> wrote:
>> David Daney <ddaney@caviumnetworks.com> writes:
>>> On 11/15/2011 09:12 AM, Richard Sandiford wrote:
>>>> Richard Sandiford<rdsandiford@googlemail.com> Âwrites:
>>>>> Andrew Pinski<andrew.pinski@caviumnetworks.com> Âwrites:
>>>>>> Hi,
>>>>>> Â ÂThis patch adds the saa and saad instructions for Octeon. Â They are
>>>>>> not part of the original Octeon but are part of the Octeon+. ÂWe at
>>>>>> Cavium don't feel the need to add a new arch for it since this is the
>>>>>> only difference instruction wise and the original Octeon is no longer
>>>>>> in production.
>>>>>
>>>>> I think it's still worth having a separate CPU name for it.
>>>>
>>>> and (to be clear) CPU enum, so that the instructions can be rejected
>>>> for plain -march=octeon.
>>>>
>>>
>>> I guess that makes sense.
>>>
>>> I don't think we want a separate E_MIPS_MACH_OCTEON?? tag in the ELF
>>> flags EF_MIPS_MACH field for this though. ÂThe Cavium SDK toolchains
>>> simply supply E_MIPS_MACH_OCTEON for all Octeon+ objects, and I would
>>> like to keep it this way.
>>
>> OK, that sounds fair.
>
> And here is the new patch with the new -march=octeon+ .  Note there is
> a hack in the testsuite mips.exp's run_dump_test_arch as the mips64r2
> sync test for Octeon+ should use the Octeon's .d file.

Looks good, subject to Maciej's comments about the tc-mips.c changes.
Please add tests for some %reloc()s too, once they work.

Richard