[mpich-discuss] question about valgrind errors I am getting

[Kenneth Raffenetti]

Have you tried the same with MPICH 3.1.4?

Ken

On 07/31/2015 12:52 PM, Grindeanu, Iulian R. wrote:
> hello,
> I am trying to see where is the leak in our code;
> valgrind shows some leaks, like these
>
> ==4262== 1,512 bytes in 3 blocks are still reachable in loss record 6 of 6
> ==4262==    at 0x4C2CC70: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
> ==4262==    by 0x56A6555: set_collops.part.0 (in /home/iulian/3rdparty/mpich-3.1/install/lib/libmpich.so.12.0.0)
> ==4262==    by 0x56A73E2: MPIR_Comm_commit (in /home/iulian/3rdparty/mpich-3.1/install/lib/libmpich.so.12.0.0)
> ==4262==    by 0x56DFAD3: MPID_Init (in /home/iulian/3rdparty/mpich-3.1/install/lib/libmpich.so.12.0.0)
> ==4262==    by 0x5796878: MPIR_Init_thread (in /home/iulian/3rdparty/mpich-3.1/install/lib/libmpich.so.12.0.0)
> ==4262==    by 0x5796288: PMPI_Init (in /home/iulian/3rdparty/mpich-3.1/install/lib/libmpich.so.12.0.0)
> ==4262==    by 0x40B20B: main (read_nc.cpp:18)
>
>
> in our code, line 18 is MPI_Init(..)
>
> commenting out portions of our code gets rid of the leaks;
> do you have a suggestion what are we doing wrong, or what is the leak related to? Or  could it be a false positive from valgrind?
>
>
> Thanks,
> Iulian
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
>
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss