Bug#895748: Default MPI

[Drew Parsons]

clone 895748 -1
retitle -1 Should we switch the standard default MPI to MPICH?

On Mon, 2018-04-16 at 13:21 +0100, Alastair McKinstry wrote:
> 
> On 16/04/2018 09:42, Drew Parsons wrote:
> > 
> > Thanks Alastair. OpenMPI 3 has just arrived in unstable.  Is it
> > still
> > the case that openmpi 3 does not provide failed image support?
> 
> Yes :-(


Interesting.  Given the mpich test failures on riscv64 that Manuel
reported, let's mark riscv64 for openmpi for the time being.  We don't
want to drastically change the entire mpi-defaults until the openmpi3
transition is finished anyway.

I'll clone this bug so we can continue discussing the Default MPI and
review Michael's testsuites (update the email bug number in any
replies!).

Drew

-- 
debian-science-maintainers mailing list
debian-science-maintainers@alioth-lists.debian.net
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers

Bug#895748: Default MPI

[Alastair McKinstry]

On 16/04/2018 09:42, Drew Parsons wrote:
> On Mon, 2018-04-16 at 08:27 +0100, Alastair McKinstry wrote:
>
> Thanks Alastair. OpenMPI 3 has just arrived in unstable.  Is it still
> the case that openmpi 3 does not provide failed image support?
Yes :-(
> Drew
>

-- 
Alastair McKinstry, , , 
https://diaspora.sceal.ie/u/amckinstry
Commander Vimes didn’t like the phrase “The innocent have nothing to fear,”
 believing the innocent had everything to fear, mostly from the guilty but in 
the longer term
 even more from those who say things like “The innocent have nothing to fear.”
 - T. Pratchett, Snuff

-- 
debian-science-maintainers mailing list
debian-science-maintainers@alioth-lists.debian.net
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers

Bug#895748: Default MPI

[Drew Parsons]

On Mon, 2018-04-16 at 11:14 +0200, Michael Banck wrote:
> 
> One thing I could do is upload a new mpi-testsuite package, which
> tries
> to run (the MPICH testsuite mostly, IIRC) for both MPICH and OpenMPI;
> that might give some input on how good/bad they are on the various
> arches.

That's an excellent idea Michael.  We've got time before making the
decision on MPI defaults while we wait for the openmpi3 transition to
complete.  That's a good time for collecting data from testsuites.

Drew

-- 
debian-science-maintainers mailing list
debian-science-maintainers@alioth-lists.debian.net
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers

Bug#895748: Default MPI

[Michael Banck]

Hi,

On Mon, Apr 16, 2018 at 04:42:31PM +0800, Drew Parsons wrote:
> On Mon, 2018-04-16 at 08:27 +0100, Alastair McKinstry wrote:
> > > Now that mpich builds on all arches, it could be a good opportunity
> > > to 
> > > review and discuss whether we want to keep openmpi as the default
> > > default, or whether we want to switch to mpich.
> > > 
> > I'd favour switching over to mpich for all archs for this release.
> > This
> > is because MPICH has the failed image support, OpenMPI hasn't.
> 
> Thanks Alastair. OpenMPI 3 has just arrived in unstable.  Is it still
> the case that openmpi 3 does not provide failed image support?

One thing I could do is upload a new mpi-testsuite package, which tries
to run (the MPICH testsuite mostly, IIRC) for both MPICH and OpenMPI;
that might give some input on how good/bad they are on the various
arches.


Michael

-- 
debian-science-maintainers mailing list
debian-science-maintainers@alioth-lists.debian.net
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers