[mpich-discuss] Porting MPICH

Jeff Hammond jhammond at alcf.anl.gov
Fri Jun 7 22:10:30 CDT 2013


I looked into this more thoroughly.  The simple solution is to use
"--with-thread-package=no --enable-threads=single" together.  The
latter will disable the use of all thread-related functions that lead
to missing symbols due to the former.

In short, a no-op implementation is not necessary, as I found out
_after_ I had created one.

Jeff

On Mon, May 20, 2013 at 6:46 AM, jhonatan alves
<jhonatanalves1 at gmail.com> wrote:
> Hello,
> Today we will try the strategies proposed by you and see how it goes.
>
> Thank you for helping
>
>
> 2013/5/19 Jeff Hammond <jhammond at alcf.anl.gov>
>>
>> Yeah, I would use the abort strategy to detect if MPICH is using
>> thread-related functions in MPI_THREAD_SINGLE (ignoring
>> MPICH_ASYNC_PROGRESS=1) and the noop implementation for practical
>> purposes.
>>
>> Jeff
>>
>> On Sun, May 19, 2013 at 2:49 PM, Pavan Balaji <balaji at mcs.anl.gov> wrote:
>> >
>> > Thanks.  Yes, they can be mostly no-ops (except TLS checks which can be
>> > any static variable since there's only one thread by definition).
>> > Implementing them as "not-implemented" aborts might be OK too.
>> >
>> >  -- Pavan
>> >
>> > On 05/19/2013 02:43 PM US Central Time, Jeff Hammond wrote:
>> >> Presumably those functions can be noops if no threads are going to be
>> >> used.  Am I wrong?
>> >>
>> >> It might be worth implementing those functions as stubs that abort
>> >> with UNIMPL error and see how far that goes.  I'll try to get to this
>> >> later today.
>> >>
>> >> Jeff
>> >>
>> >> On Sun, May 19, 2013 at 2:36 PM, Pavan Balaji <balaji at mcs.anl.gov>
>> >> wrote:
>> >>>
>> >>> On 05/19/2013 02:28 PM US Central Time, jhonatan alves wrote:
>> >>>> But i believe that the missing of POSIX
>> >>>> threads can be the most trouble part. So we need to port to the
>> >>>> thread
>> >>>> implementation in EPOS.
>> >>>
>> >>> Yup, I too believe that'll be the blocker.  MPICH supports multiple
>> >>> threading packages, but currently requires at least one to function
>> >>> correctly:
>> >>>
>> >>> https://trac.mpich.org/projects/mpich/ticket/231
>> >>>
>> >>> It's been a while since I looked into this issue, but I could look
>> >>> into
>> >>> it if you are running into it for your platform.
>> >>>
>> >>>  -- Pavan
>> >>>
>> >>> --
>> >>> Pavan Balaji
>> >>> http://www.mcs.anl.gov/~balaji
>> >>> _______________________________________________
>> >>> discuss mailing list     discuss at mpich.org
>> >>> To manage subscription options or unsubscribe:
>> >>> https://lists.mpich.org/mailman/listinfo/discuss
>> >>
>> >>
>> >>
>> >
>> > --
>> > Pavan Balaji
>> > http://www.mcs.anl.gov/~balaji
>>
>>
>>
>> --
>> Jeff Hammond
>> Argonne Leadership Computing Facility
>> University of Chicago Computation Institute
>> jhammond at alcf.anl.gov / (630) 252-5381
>> http://www.linkedin.com/in/jeffhammond
>> https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond
>> ALCF docs: http://www.alcf.anl.gov/user-guides
>
>



-- 
Jeff Hammond
Argonne Leadership Computing Facility
University of Chicago Computation Institute
jhammond at alcf.anl.gov / (630) 252-5381
http://www.linkedin.com/in/jeffhammond
https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond
ALCF docs: http://www.alcf.anl.gov/user-guides



More information about the discuss mailing list