[mpich-discuss] exit without calling finalize（windows xp pro 64bit）

[yangjunbo686]

Hi,Jayesh

Just as what you say, there is something wrong with my code, rather than MPI. 
I removed the code with respect to MPI in the program, and run it again. Three hours later, an error which was not found in test phase occurred,
  Unhandled exception at 0x78078705 (msvcp100.dll) in xx.exe: 0xC0000005: Access violation writing location 0x00000000.

That means the issue has nothing to do with MPI.
Although the error has yet to been found, the problem has become less difficult. 

Thanks,
Junbo
At 2012-09-19 00:33:10,"Jayesh Krishna" <jayesh at mcs.anl.gov> wrote:
>Hi,
> The error message says that one of the MPI processes exited before calling MPI_Finalize. This usually occurs when one of the MPI processes exits prematurely (segfault etc). I would recommend debugging your code further to find the source of the problem.
>
>(PS: The error message can also occur if there is a bug in MPICH that causes the application to abort - however I would recommend debugging the application first to find the exact source of the problem.)
>
>Regards,
>Jayesh
>
>----- Original Message -----
>From: "yangjunbo686" <yangjunbo686 at 163.com>
>To: mpich-discuss at mcs.anl.gov
>Sent: Friday, September 14, 2012 7:58:22 PM
>Subject: [mpich-discuss] exit without calling finalize（windows xp pro 64bit）
>
>
>
>
>
>
>Dear all ， 
>
>
>
>I recently have written a program which calls MSC.Patran and MSC.Nastran over and over again until find an optimal solution, in the visual studio 2010 C++ environment on Windows XP professional 64bit. And the version of mpich2 is 1.4.1p1( 32bit). 
>
>Build, run the program and then wait. Sometimes, I can get an optimum solution when the parallel is done. But mostly, the optimization terminates unexpectedly, and error message follows: 
>
>
>
>job aborted: 
>
>rank: node: exit code[: error message] 
>
>0: yang: 123 
>
>1: yang: -1073741819: process 1 exited without calling finalize 
>
>2: yang: 123 
>
>
>
>Any advice? 
>
>Thanks for your help! 
>
>
>
>Yang Junbo 
>
>
>_______________________________________________
>mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
>To manage subscription options or unsubscribe:
>https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss