[mpich-discuss] MPI_Barrier() failed

[王凯]

Hi! Pavan Balaji
Thanks for your quick reply!
Here is the information on the error:
--Fatal error in PMPI_BARRIER:Other MPI error, error stack
--MPI_Barrier(MPI_COMM_WORLD) failed
--Failure during collective
--connect failed -- the semaphore timeout period has expired.


The difference of the time consuming of the two process is about several minutes.
Is it the reason that caused the connect failed?

--


Email:    yogikai at 163.com
Address: Room 921, Automation Building,
No. 95 Zhongguancun East Road,
Haidian District, Beijing 100190, China
Cell Phone:15210370340
The State Key Laboratory for Intelligent Control and Management of Complex Systems
Institute of Automation, Chinese Academy of Sciences





At 2011-06-29 10:15:46，"Pavan Balaji" <balaji at mcs.anl.gov> wrote:

>
>There's no time limit in barrier. The problem might be something else. 
>Which version of MPICH2 are you using? Also can you send some more 
>information on the error?
>
>  -- Pavan
>
>On 06/28/2011 09:11 PM, 王凯 wrote:
>> Hi!
>> I have two computer in the MPI program, and I use the function
>> mpi_barrier(MPI_COMM_WORLD) to synchronize the two process on the two
>> computer.
>> One of the process is running more slower than the other one because two
>> different GPUs is on the two computer. I got "fatal error in
>> PMPI_Barrier" and "connect failed--the semaphore timeout period has
>> expired".
>> Is there a time limitation in the mpi_barrier()?
>>
>> --
>>
>> Email: yogikai at 163.com <mailto:yogikai at 163.com>
>> Address: Room 921, Automation Building,
>> No. 95 Zhongguancun East Road,
>> Haidian District, Beijing 100190, China
>> Cell Phone:15210370340
>> The State Key Laboratory for Intelligent Control and Management of
>> Complex Systems
>> Institute of Automation, Chinese Academy of Sciences
>>
>>
>>
>>
>>
>> _______________________________________________
>> mpich-discuss mailing list
>> mpich-discuss at mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
>-- 
>Pavan Balaji
>http://www.mcs.anl.gov/~balaji
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20110629/343e798f/attachment-0001.htm>