[mpich-discuss] Can I run MPI program without mpirun/mpiexec?

[junli gu]

I did as what you said., get the value of argc and argv[],  set them as
global variables  in the source code,also set the number pf processes as a
global variable,

Then I compile it again and run it with mpiexec successfully.
 then I try to run it as a binary like this: ./mpi_array. I got the
following error:

"Quitting. Number of MPI tasks must be divisible by 4.
application called MPI_Abort(MPI_COMM_WORLD, 0) - process 0
"
THis is because the program requires there should be 4 processes., while
there is actually 1 process created. I t seems that
MPI_Comm_size(MPI_COMM_WORLD, &numtasks) DOES NOT accept the prevalued
numtasks; Maybe the numtasks can only be got from the command line?

If I delete the function MPI_Comm_size(MPI_COMM_WORLD, &numtasks); I compile
it again and run it with mpiexec successfully.

Then I try to run it as a binary like this: ./mpi_array

And i got the following errors:

MPI task 0 has started...
Initialized array sum = 1.280000e+14
Fatal error in MPI_Send: Invalid rank, error stack:
MPI_Send(174): MPI_Send(buf=0xbfeaee88, count=1, MPI_INT, dest=1, tag=2,
MPI_COMM_WORLD) failed
MPI_Send(99).: Invalid rank has value 1 but must be nonnegative and less
than 1

It seems that MPI has only boot one process. So when process 1 tries to send
message to process 2, there is an error.

 MPI_Comm_size and  MPI_Comm_rank(MPI_COMM_WORLD,&taskid) start the fixed
number of processes together.

How can we create more that one processes as needed without the
MPI_Comm_size and  MPI_Comm_rank?  I  wonder also that even if we can create
some number of processes. Can the MPI_send(), MPI_receive .etc. work without
mpiexec?

2009/11/23 Mário Costa <mario.silva.costa at gmail.com>

> Hi,
>
> I have never done it but, I would say try it with mpiexec and print
> the arv/arc contents then use the values sent by mpiexec to initialize
> them inside your code, then pass them to MPI_Init()  ...
>
> Let me know if it works ....
>
> Regards,
> Mário
>
> 2009/11/20 junli gu <gujunli at gmail.com>:
> > I need to run multi-process mpi programs. Is it possible that run them
> > without mpirun command? to include everything needed into one stand alone
> > binary?
> >
> > 2009/11/20 <chan at mcs.anl.gov>
> >>
> >> ----- "junli gu" <gujunli at gmail.com> wrote:
> >>
> >> > Hi everyone:
> >> >
> >> >  I want to run mpi program like a normal binary without
> >> > mpirun/mpiexec
> >> > command, like this: ./mpi_hello . Is this possible?
> >>
> >> Yes. but without mpiexec, you can only run 1 process.
> >>
> >> > This is possible only that I can compile mpi program and put all the
> >> > runtime
> >> > information into a stand alone binary. I don't know if it is
> >> > possible.
> >>
> >> Hmm... Since you are using a cross-compiler, alpha-gcc, not sure how
> >> you built it...  Trying adding -static during linking to see if it
> >> links OK.
> >>
> >> A.Chan
> >>
> >> >  Thank you very much!
> >> >
> >> > --
> >> > ************************************************
> >> > Junli Gu--谷俊丽
> >> > Coordinate Science Lab
> >> > University of Illinois at Urbana-Champaign
> >> > ************************************************
> >> >
> >> > _______________________________________________
> >> > mpich-discuss mailing list
> >> > mpich-discuss at mcs.anl.gov
> >> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> >> _______________________________________________
> >> mpich-discuss mailing list
> >> mpich-discuss at mcs.anl.gov
> >> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> >
> >
> >
> > --
> > ************************************************
> > Junli Gu--谷俊丽
> > Coordinate Science Lab
> > University of Illinois at Urbana-Champaign
> > ************************************************
> >
> > _______________________________________________
> > mpich-discuss mailing list
> > mpich-discuss at mcs.anl.gov
> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> >
> >
>
>
>
> --
> Mário Costa
>
> Laboratório Nacional de Engenharia Civil
> LNEC.CTI.NTIEC
> Avenida do Brasil 101
> 1700-066 Lisboa, Portugal
> Tel : ++351 21 844 3911
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>



-- 
************************************************
Junli Gu--谷俊丽
Coordinate Science Lab
University of Illinois at Urbana-Champaign
************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20091123/40734bff/attachment.htm>