[mpich-devel] Future of MPE?
balaji at mcs.anl.gov
Tue Oct 15 18:14:35 CDT 2013
With respect to the question on whether MPE will be distributed with mpich, the answer is it depends on what you mean by "distribute with mpich".
If you mean, can you download one tarball and have code for both mpich and mpe, then sure, we can do that.
If you mean, I'll do configure/make/make install once and both mpich and mpe will be built, then probably not. This requires a lot of build system hacking for MPE's configure to test for MPI features before the MPI implementation is actually built. That's nasty; we already do that for the mpich test suite and it's a nightmare to maintain. We don't want to do that for yet another package.
Option 1 is not really more useful than downloading MPE from a separate URL and building it separately. That's the reason why we didn't do it.
On Oct 15, 2013, at 5:29 PM, Jeff Hammond wrote:
> Thanks for the info. If I get time to work on MPE, I'll see what I can do about the blocking issues noted below.
> ----- Original Message -----
>> From: "Pavan Balaji" <balaji at mcs.anl.gov>
>> To: devel at mpich.org
>> Sent: Tuesday, October 15, 2013 5:07:47 PM
>> Subject: Re: [mpich-devel] Future of MPE?
>> There are a couple of things to be done in MPE:
>> 1. The const bindings need to be added to bring it to MPI-3. Without
>> this, it won't build.
>> 2. The new MPI functions need to be added. This is a lower priority
>> if people only need the MPI-2 functionality.
>> 3. The tools that come with MPE (e.g., the coll checker) need to be
>> updated for MPI-3. This is also a lower priority if people only
>> need MPI-2.
>> There is also some more cleanup that's needed (e.g., build system).
>> Unfortunately, it's not on our priority list right now, given that
>> there are many other MPI-3 features that need improvement. That's
>> taking up all of our development bandwidth. But if someone is
>> willing to work on it, we'll be happy to help out.
>> -- Pavan
>> On Oct 15, 2013, at 4:06 PM, Jeff Hammond wrote:
>>> What is there to resurrect? Are you having trouble building MPE?
>>> Why does it need to be distributed with MPICH to be viable? MPE
>>> sits strictly on top of any MPI as far as I know.
>>> I have on my TODO list to migrate the existing HTML docs to
>>> http://wiki.mpich.org/mpe/index.php/Main_Page but it's currently
>>> in slot 17562 in the priority queue.
>>> ----- Original Message -----
>>>> From: "Lisandro Dalcin" <dalcinl at gmail.com>
>>>> To: "mpich-devel" <devel at mpich.org>
>>>> Sent: Tuesday, October 15, 2013 3:42:50 PM
>>>> Subject: [mpich-devel] Future of MPE?
>>>> Is there any plan to resurrect MPE and distribute it within
>>>> in the near future?
>>>> Lisandro Dalcin
>>>> CIMEC (UNL/CONICET)
>>>> Predio CONICET-Santa Fe
>>>> Colectora RN 168 Km 472, Paraje El Pozo
>>>> 3000 Santa Fe, Argentina
>>>> Tel: +54-342-4511594 (ext 1016)
>>>> Tel/Fax: +54-342-4511169
>>> Jeff Hammond
>>> Argonne Leadership Computing Facility
>>> University of Chicago Computation Institute
>>> jhammond at alcf.anl.gov / (630) 252-5381
>>> ALCF docs: http://www.alcf.anl.gov/user-guides
>> Pavan Balaji
> Jeff Hammond
> Argonne Leadership Computing Facility
> University of Chicago Computation Institute
> jhammond at alcf.anl.gov / (630) 252-5381
> ALCF docs: http://www.alcf.anl.gov/user-guides
More information about the devel