[mvapich-discuss] problem with jobstartup with mvapich 0.96/0.97 and mpiexec

[Weikuan Yu]

Hi, Jimmy,

Thanks for this problem report w.r.t mpiexec. We will consider this  
situation and bring back our thoughts on possible actions.

Weikuan

On Mar 15, 2006, at 7:04 PM, Jimmy Tang wrote:

> Hi,
>
> with the latest release of 0.97 I was testing it against some codes  
> that
> we have locally on our systems etc... and after some frustration that
> mpiexec hasnt been working since 0.96. I decided to investigate why
> mpiexec fails with 0.96/0.97 of mvapich, which lead me to email the
> mvapich mailing list for a solution.
>
> the mpiexec developer pointed me to  the source code
>
> 	mpid/vapi/process/pmgr_client_mpirun_rsh.c
>
> there is a block of code which repeats itself 3 times (which also  
> breaks
> things) removing code allows mvapich to function correctly with mpiexec
>
> I'd like to ask if its possible to remove the block of code from the
> source? or at least put an ifdef in to disable that code by default?
>
> I would imagine most people probably run (torque|openpbs) + mpiexec +
> mvapich which is where mvapich fails to startup with mpiexec with that
> code block in this setup. it would be nice if mvapich played nice with
> mpiexec out of the box without the need to remove "cruft" that breaks
> things.
>
> attached is a diff with the code block removed.
>
> Thanks,
> Jimmy
>
>
> --  
> Jimmy Tang
> Trinity Centre for High Performance Computing,
> Lloyd Building, Trinity College Dublin.
> http://www.tchpc.tcd.ie/
> <pmgr_client_mpirun_rsh.c.patch>_______________________________________ 
> ________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>
--
Weikuan Yu, Computer Science, OSU
http://www.cse.ohio-state.edu/~yuw