Re: mpi 1.1.1 released

[Michael Creel]

Hi Carlo,

You might like to have a look at MPITB for Octave. This allowed use of mpi interactively, and worked very well. It used LAM/MPI, and if I recall correctly, it didn't work with Open MPI. I suppose that this framework could be used. http://www.ugr.es/~jfernand/mpitb.html

Cheers,

M.

On Thu, Jan 2, 2014 at 10:16 AM, c. <address@hidden> wrote:

On 2 Jan 2014, at 05:12, Sukanta Basu <address@hidden> wrote:

>>
>> Hi Sukanta,
>>
>> Sorry for the late reply.
>>
>> The problem with mpich2 is given by the following lines in the Makefile:
>>
>> MPICC := mpic++
>> JUNK := $(shell $(MPICC) -showme:compile)
>> MPIINC := $(shell echo $(JUNK) | sed -e "s/-pthread/-lpthread/g")
>> JUNK := $(shell $(MPICC) -showme:link)
>> MPILIBS := $(shell echo $(JUNK) | sed -e "s/-pthread/ /g")
>>
>> which try to extract compile and link flags needed to build against
>> your mpi library by running
>>
>> mpic++ -showme:compile
>>
>> and
>>
>> mpic++ -showme:link
>>
>> respectively. It appears mpich2 does not support these commands and I'm actually not even
>> sure it provides the mpic++ wrapper either.
>>
>> So the best option for the moment is the following, I changed the variable definitions in
>> the Makefile as follows:
>>
>> MPICC     ?= mpic++
>> OFMPIINC  ?= $(shell $(MPICC) -showme:compile | sed -e "s/-pthread/-lpthread/g")
>> MPIINC    := $(OFMPIINC)
>> OFMPILIBS ?= $(shell $(MPICC) -showme:link | sed -e "s/-pthread/ /g")
>> MPILIBS   := $(OFMPILIBS)
>>
>> with this change you can set the include and link flags for your mpi yourself before installing
>> the package:
>>
>>>> setenv ("OFMPIINC", "-I/opt/openmpi/1.6.5/include")
>>>> setenv ("OFMPLIBS", "-L/opt/openmpi/1.6.5/lib -lmpi_cxx -lmpi -lm")
>>>> pkg install mpi.tar.gz
>>
>> I currently made this change in the subversion repository, it's not released yet, so you
>> must either download the package from there or unpack and patch the package yourself.
>>
>> I'd be glad to here feedback from you if you try this out.
>>
>> HTH,
>> c.
>>
>>
>>
>

> Hi Carlo et al.,
>
> Happy New Year to all of you!
>
> During the holidays, I spent some time on the memory leak issue
> associated with Octave's mpi package (aka openmpi_ext package). I
> performed several simulations on various platforms. Please see the
> attached report (in doc and pdf format). Now, I am more or less
> convinced that the mpi package is causing severe memory leaks.
> Unfortunately, I cannot identify the location of the leaks. I would
> greatly appreciate your help in this matter.
>
> Best regards,
> Sukanta

Hi Sukanta,

Thanks for the thorough testing, glad to hear that Octave MPI + MPICH2 works for you!

As for the memory leak issue, can you provide a simple test to reproduce the problem locally and try to debug it?

I am considering proposing a GSoC project about improvements to the MPI package, in particular I'd like to add the
ability for users to start parallel jobs and collect the output in an interactive Octave CLI/GUI session.

If your have a non trivial application built on Octave MPI it would be great to use it for testing.
Would it be possible to use your MATLES application for this purpose? Is it Free Software?

In addition to solving the memory leak issue do you have any other improvements that could be part of the project?
Would you like to be a mentor for the project?

Happy new year!
c.