This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: atomic ops for powerpc
>>>>> Geoffrey Keating writes:
>> Can you place the new patterns in sync.md like other ports?
Geoff> I considered that, but didn't do it because the patterns were only ~400
Geoff> lines.
Other targets are similar. I would greatly appreciate if we could
keep the rs6000 port consistent with other ports for this new
functionality. I ask again: please use sync.md.
>> Not all assemblers understand lwsync
Geoff> Which ones? Can we just say that you shouldn't use those assemblers if
Geoff> you want to use that function? Otherwise we'll have to say '.long
Geoff> xxxx'.
It will not work on recent versions of AIX or Linux. How do you
plan to ensure that those functions are not used when the assembler does
not support the instruction?
Please keep in mind the wide range of platforms using PowerPC.
>> and I think the pattern is missing TARGET_POWERPC.
Geoff> I understood that it does work on POWER, it just turns into a regular
Geoff> 'sync'.
It depends on your intent for the pattern. I assumed you were
trying to generate that processor-specific instruction, because you did
not give the pattern a generic name. The "lwsync" instruction is
PowerPC-specific.
David