[mpich-discuss] Problem with spawning child with same executable name

[Yauheni Zelenko]

Hi, Pavan!

It's necessary to run "mpiexec -n 1 /Spawn" in configuration with several hosts (universe size > 1).

Eugene
________________________________________
From: Pavan Balaji [balaji at mcs.anl.gov]
Sent: Saturday, February 05, 2011 2:03 AM
To: mpich-discuss at mcs.anl.gov
Cc: Yauheni Zelenko
Subject: Re: [mpich-discuss] Problem with spawning child with same executable name

I'm not able to reproduce this problem. This is what I did (attached is
a slightly modified Spawn.c):

% mpicc Spawn.c -o Spawn

% mpicc SpawnChild.c -o SpawnChild

% mpiexec ./Spawn ./Spawn
Start process 0 of 1 on snowflake in universe of 1

% mpiexec ./Spawn ./SpawnChild
Start process 0 of 1 on snowflake in universe of 1

% mpiexec -n 2 ./Spawn ./Spawn
Start process 1 of 2 on snowflake in universe of 1
Start process 0 of 2 on snowflake in universe of 1

% mpiexec -n 2 ./Spawn ./SpawnChild
Start process 0 of 2 on snowflake in universe of 1
Start process 1 of 2 on snowflake in universe of 1

% mpiexec -n 2 -hosts localhost,127.0.1.1 ./Spawn ./Spawn
Start process 1 of 2 on snowflake in universe of 2
Start process 0 of 2 on snowflake in universe of 2

% mpiexec -n 2 -hosts localhost,127.0.1.1 ./Spawn ./SpawnChild
Start process 1 of 2 on snowflake in universe of 2
Start process 0 of 2 on snowflake in universe of 2

What am I missing?

  -- Pavan

On 02/04/2011 03:46 PM, Yauheni Zelenko wrote:
> Hi!
>
> I'm trying to use MPI_Comm_spawn to run child process and then use MPI_Intercomm_merge. This steps works when spawned children have different executable name then parent process but failed on MPI_Intercomm_merge step if executable names are same.
>
> Attached files could demonstrate a problem. Code is based on spawn_merge example.
>
> If difference in executable names is requirement, will be good idea to include in into  MPI_Comm_spawn documentation. Otherwise it's bug.
>
> Eugene.
>
>
>
> _______________________________________________
> 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