On 3 Mar, 2012, at 04:36 AM, Brad Burkman wrote:

> I've built Sage on the LittleFe teaching cluster, made of six Intel
> Atoms in a case I can carry from class to class (littlefe.net, if
> you're interested).
> 
> Sage works just fine on the head node, but won't detect the other
> nodes.

If these processors are not "integrated" by the OS into a single view, I don't 
believe that Sage will be able to use the "slaves" without some help from you; 
since I don't know anything about  the OS in use, I can't say more.

> I ran "import sage.parallel.ncpus" and "sage.parallel.ncpus.ncpus()",
> and just get 2 [just the head node], not the expected 12 [the entire
> cluster].

I believe that Sage relies on what the OS tells it (e.g., via sysctl(2)).    
What does "sysctl hw.ncpu" tell you?

> I tried putting "-machinefile ~/machines" into the argument of
> sage.parallel.ncpus.ncpus(), which is what I do with mpirun, but just
> got errors.

As a general rule, you can use "?" and "??" after procedure names to get 
information and implementation details.  Try
   sage.parallel.ncpus.ncpus??

> Does Sage support distributed memory parallelism (like MPI), or just
> shared memory parallelism (like OpenMP)?

By default, Sage uses "fork()", so no shared memory (except read-only access to 
process space).  Try 
  parallel??

HTH

Justin

--
Justin C. Walker, Curmudgeon-At-Large
Institute for the Absorption of Federal Funds
--------
If you're not confused,
You're not paying attention
--------



-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to