#10869: Optional package openmpi-1.4.3 fails to install on Solaris 10 (SPARC)
---------------------------------+------------------------------------------
Reporter: drkirkby | Owner: tbd
Type: defect | Status: new
Priority: major | Milestone: sage-4.7
Component: optional packages | Keywords:
Author: | Upstream: N/A
Reviewer: | Merged:
Work_issues: |
---------------------------------+------------------------------------------
Comment(by drkirkby):
Replying to [comment:3 vbraun]:
> Its a pretty low-level C library, so I don't think Sage developers
should write any documentation. We don't have documentation for libntl,
say, either. You need to know MPI and write Cython code to use it.
I can't feel that it would be better if there were some examples of how to
use a cluster. If you take a look at Mathematica's documentation
http://reference.wolfram.com/mathematica/ParallelTools/tutorial/Overview.html
it is pretty clear about how to set up a cluster of machines. For cores on
the same machine, one does not need to do anything
{{{
drkirkby@hawk:~$ math
Mathematica 7.0 for Sun Solaris x86 (64-bit)
Copyright 1988-2009 Wolfram Research, Inc.
In[1]:= ParallelEvaluate[$ProcessID]
LaunchKernels::launch: Launching 8 kernels...
Out[1]= {17926, 18049, 18172, 18295, 18418, 18541, 18664, 18787}
In[2]:= ParallelSum[i^2, {i, 1000}]
Out[2]= 333833500
}}}
For kernels on other machines, it's a bit more complex, but I got it set
up in about 30 minutes, using a GUI where you select the kernels.
It seems Sage has the OpenMPI library, but without some examples of how to
use it, then it is going to be next to impossible for someone other than
the people that implemented it in Sage to know how to use it.
Do you know any Sage '''user''' that has used the library to do anything,
who is '''not a developer''' involved in implementing it in Sage?
I've got a book on OpenMPI here (got it free after reviewing another
book), but I feel even if I knew how to use OpenMPI, I fear doing anything
in Sage would be hard.
> You don't necessarily need a cluster, you can also use it to distribute
work over multiple cores of a single computer. Although not quite as
efficient as threads, that gives you the opportunity to run your code
unmodified on a cluster should you need to in the future.
Same with Mathematica. The kernels can be local, remote, or a mix of the
two.
> Since its quite a mature package I don't see a reason why it shouldn't
be in optional. Most of the experimental spkgs are completely broken by
comparison. I also have never seen a SPARC cluster. Though we should fix
it eventually, of course. Given that Solaris uses OpenMPI as its official
MPI library it must be possible to compile it :-)
Yes, it must be possible to compile it. Not sure why it does not build.
I just tried the package on my !OpenSolaris machine using Sun Studio,
rather than gcc. That builds OK. Which probably means it would on SPARC
too, though I've not tried it. I updated #10866 to indicate the library
builds with the Sun compiler if using the Sun compiler. It took 9 minutes
to build on my 3.33 GHz !OpenSolaris machine, so I hate to think how long
it would take on t2.math. It would not surprise me if it took an hour or
more, as that machine is very slow.
Dave
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/10869#comment:4>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
