Anuga-user Digest, Vol 71, Issue 2

[anuga-use...@lists.sourceforge.net]

Send Anuga-user mailing list submissions to
anuga...@lists.sourceforge.net

To subscribe or unsubscribe via the World Wide Web, visit
https://lists.sourceforge.net/lists/listinfo/anuga-user
or, via email, send a message with subject or body 'help' to
anuga-use...@lists.sourceforge.net

You can reach the person managing the list at
anuga-us...@lists.sourceforge.net

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Anuga-user digest..."


Today's Topics:

1. Anuga-Parallel not compatible with cluster configuration
(Jos? Galaz)
2. Parallel anuga not compatible with cl?ster configuration
(Jos? Galaz)


----------------------------------------------------------------------

Message: 1
Date: Thu, 16 Oct 2014 18:23:28 -0300
From: Jos? Galaz <jdg...@uc.cl>
Subject: [Anuga-user] Anuga-Parallel not compatible with cluster
configuration
To: anuga...@lists.sourceforge.net
Message-ID:
<CALTyBR=vmqvKEFAUgtM_39Wz6FpJwc2cno=vTUvsJC...@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Dear mailing-list,

I am currently having some troubles trying to use AnuGA in a cluster
installed at my university. The administrator of the cluster tells me that
pypar handles the memory in a different way than it is currently being done
here, and hence, it is not compatible.

One solution they suggested was to run my program using mpi4py, but I
assume that it is not an easy thing to do with AnuGA. Has anyone had faced
a problem like this before?

I'm totally open to suggestions, and please, feel free to ask about any
other detail you may consider important.

Best regards,

Jos?
-------------- next part --------------
An HTML attachment was scrubbed...

------------------------------

Message: 2
Date: Thu, 16 Oct 2014 18:18:39 -0300
From: Jos? Galaz <jdga...@gmail.com>
Subject: [Anuga-user] Parallel anuga not compatible with cl?ster
configuration
To: anuga...@lists.sourceforge.net
Message-ID:
<CALTyBR=gxDB7-6Kemzp_+TwFHTLsoAgpe2Bot+=4+67xbzPy=w...@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Dear mailing-list,

I am currently having some troubles trying to use AnuGA in a cluster
installed at my university. The administrator of the cluster tells me that
pypar handles the memory in a different way than it is currently being done
here, and hence, it is not compatible.

One solution they suggested was to run my program using mpi4py, but I
assume that it is not an easy thing to do with AnuGA. Has anyone had faced
a problem like this before?

I'm totally open to suggestions, and please, feel free to ask about any
other detail you may consider important.

Best regards,

Jos?
-------------- next part --------------
An HTML attachment was scrubbed...

------------------------------

------------------------------------------------------------------------------
Comprehensive Server Monitoring with Site24x7.
Monitor 10 servers for $9/Month.
Get alerted through email, SMS, voice calls or mobile push notifications.
Take corrective actions from your mobile device.
http://p.sf.net/sfu/Zoho

------------------------------

_______________________________________________
Anuga-user mailing list
Anuga...@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/anuga-user


End of Anuga-user Digest, Vol 71, Issue 2
*****************************************

[Stephen Roberts]

Hi Jos,

I know of a user who wrote a wrapper around mpi4py to mimic pypar and so
was able to run anuga parallel using mpi4py.

I will try to obtain his code.

But it was not too difficult.

The only tricky part was realising that pypar checks for contiguous data
(when sending numpy arrays) and mpi4py doesn't, so the wrapper needed to
check for that and copy to contiguous data before sending.

Cheers
Steve

--
+--------------------------------------------------------------------+
| Steve Roberts | My Office: (61)(2) 6125 4445 |
| Department of Mathematics | Math Office: (61)(2) 6125 2908 |
| Mathematical Sciences Institute| Fax: (61)(2) 6125 4984 |
| John Dedman Building #27 | mailto:stephen...@anu.edu.au |
| Australian National University | http://www.maths.anu.edu.au/~steve|
| Canberra ACT 0200 AUSTRALIA | ANU CRICOS # 00120C |
+--------------------------------------------------------------------+