This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC: [PATCH] MIPS: Implement built-in atomic memory operations.


David Daney <ddaney@avtrex.com> writes:
> Maciej W. Rozycki wrote:
>>>> For a Linux configuration with MIPS I this should also be true,
>>>> the kernel emulates ll/sc in those cases.
>>>>
>>>>   
>>>>       
>>> In this case could you "lie" to the compiler by passing -mips2 on the command
>>> line?  Otherwise we would have to either leave it as it is or add a new
>>> command line parameter to explicitly enable ll/sc.
>>>     
>>
>>  I would feel a bit uneasy about adding the implication of the 
>> availability of ll/sc based on the target operating system.  Using a 
>> switch like "-march=mips2" is unacceptable as it enables some other 
>> instructions.  I would be in favour to an additional option along the 
>> lines of -mbranch-likely.
>>
>>   
> Something like -mllsc.  The default value could be set at configure time 
> with --with-llsc=[true|false]. Perhaps mips*-*-linux* would imply 
> --with-llsc=true unless overridden.  Done in a similar manner to the 
> default -march and -msoft-float settings.
>
> Perhaps Richard could opine on the matter.

Sounds fine to me.  FWIW, if you'd rather not add this option yourself,
I think it's OK for the patch to go in without.

Richard


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]