[MITgcm-support] xmitgcm: python package for reading mitgcm mds files
Matthew Mazloff
mmazloff at ucsd.edu
Tue Oct 18 10:34:51 EDT 2016
Hi Ryan
Just a note — hope not to distract on this great package — There are a lot of machines (e.g. stampede) and setups where one needs useSingleCpuIO=.FALSE.
This is primarily so each node can write to its local /tmp disk instead of the global file system. Makes for a huge performance increase.
Also, if ones setup gets huge one node may not be able to receive all the data to I/O. It may just not have enough memory to hold the global array.
Thanks for the package!
Matt
> On Oct 18, 2016, at 7:27 AM, Ryan Abernathey <ryan.abernathey at gmail.com> wrote:
>
> Great question Liam. The package definitely assumes useSingleCpuIO=.TRUE. However, it would probably be pretty straightforward to accommodate tile-based files. In fact, this would mesh quite well with the concept of dask "chunks" (http://dask.pydata.org/en/latest/array-creation.html#chunks <http://dask.pydata.org/en/latest/array-creation.html#chunks>) and could be implemented pretty far upstream.
>
> If this is something that would be useful to you, please open an issue:
> https://github.com/xgcm/xmitgcm/issues <https://github.com/xgcm/xmitgcm/issues>
>
> I am curious why anyone uses useSingleCpuIO=.FALSE. Are there any performance benefits? To me it just seems like a hassle dealing with all the files, and my models run faster with useSingleCpuIO=.TRUE. So it's a win-win.
>
> Note that Andrea also raised an issue related to reading diagnostics that use the `levels` option in data.diagnostics to select only certain vertical levels for output. (It doesn't work with xmitgcm...yet)
> https://github.com/xgcm/xmitgcm/issues/8 <https://github.com/xgcm/xmitgcm/issues/8>
>
> -Ryan
>
>
>
>
> On Tue, Oct 18, 2016 at 10:10 AM, Liam Brannigan <liam.brannigan at misu.su.se <mailto:liam.brannigan at misu.su.se>> wrote:
> Hi Ryan
>
> Thanks for developing this. I was wondering how you produce the output files for parallel runs - do you have useSingleCpuIO=.TRUE. or do you glue together the mds output to produce the files read by xmitgcm?
>
> Liam
> ________________________________________
> From: mitgcm-support-request at mitgcm.org <mailto:mitgcm-support-request at mitgcm.org> <mitgcm-support-request at mitgcm.org <mailto:mitgcm-support-request at mitgcm.org>>
> Sent: 15 October 2016 18:00
> To: mitgcm-support at mitgcm.org <mailto:mitgcm-support at mitgcm.org>
> Subject: MITgcm-support Digest, Vol 160, Issue 7
>
> Send MITgcm-support mailing list submissions to
> mitgcm-support at mitgcm.org <mailto:mitgcm-support at mitgcm.org>
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://mitgcm.org/mailman/listinfo/mitgcm-support <http://mitgcm.org/mailman/listinfo/mitgcm-support>
> or, via email, send a message with subject or body 'help' to
> mitgcm-support-request at mitgcm.org <mailto:mitgcm-support-request at mitgcm.org>
>
> You can reach the person managing the list at
> mitgcm-support-owner at mitgcm.org <mailto:mitgcm-support-owner at mitgcm.org>
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of MITgcm-support digest..."
>
>
> Today's Topics:
>
> 1. xmitgcm: python package for reading mitgcm mds files
> (Ryan Abernathey)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sat, 15 Oct 2016 00:19:45 -0400
> From: Ryan Abernathey <ryan.abernathey at gmail.com <mailto:ryan.abernathey at gmail.com>>
> To: "mitgcm-support at mitgcm.org <mailto:mitgcm-support at mitgcm.org>" <mitgcm-support at mitgcm.org <mailto:mitgcm-support at mitgcm.org>>
> Subject: [MITgcm-support] xmitgcm: python package for reading mitgcm
> mds files
> Message-ID:
> <CAJAZx5DG=vWDD-G9nyN+RqvekzUpxdZsZm3_dTTuM+61_QM6Hg at mail.gmail.com <mailto:vWDD-G9nyN%2BRqvekzUpxdZsZm3_dTTuM%2B61_QM6Hg at mail.gmail.com>>
> Content-Type: text/plain; charset="utf-8"
>
> Dear Colleagues,
>
> I have released a first version of my python package called xmitgcm:
>
> xmitgcm is a python package for reading MITgcm <http://mitgcm.org/ <http://mitgcm.org/>> binary
> MDS
> <http://mitgcm.org/public/r2_manual/latest/online_documents/node277.html <http://mitgcm.org/public/r2_manual/latest/online_documents/node277.html>> files
> into xarray <http://xarray.pydata.org/ <http://xarray.pydata.org/>> data structures. By storing data in
> dask <http://dask.readthedocs.org/ <http://dask.readthedocs.org/>> arrays, xmitgcm enables parallel,
> out-of-core <https://en.wikipedia.org/wiki/Out-of-core_algorithm <https://en.wikipedia.org/wiki/Out-of-core_algorithm>> analysis
> of MITgcm output data.
>
> For more information, please consult
>
> - The online documentation: http://xmitgcm.readthedocs.io <http://xmitgcm.readthedocs.io/>
> - The GitHub repository: https://github.com/xgcm/xmitgcm <https://github.com/xgcm/xmitgcm>
>
> Among many features, the package includes basic support for reading llc mds
> data, which is something we have been lacking in python. xmitgcm does not
> depend on any existing MITgcm scripts like rdmds.
>
> xmitgcm is not a complete analysis framework like gcmfaces. It is just a
> tool for reading mds data. Nevertheless, once the data is in xarray form,
> many powerful options become available.
>
> Early in its development, xmitgcm used to be called "xgcm". Those of you
> who were using the earlier "xgcm" package should switch to this one. It
> should be largely backwards compatible, but xmitgcm will continue to
> receive udpates, bug fixes, and new features.
>
> I welcome your feedback on your experience with xmitgcm via email, or, even
> better, as github issues
> https://github.com/xgcm/xmitgcm/issues <https://github.com/xgcm/xmitgcm/issues>
> and pull requests
> https://github.com/xgcm/xmitgcm/pulls <https://github.com/xgcm/xmitgcm/pulls>
>
> Happy pythoning!
>
> -Ryan
> http://rabernat.github.io/ <http://rabernat.github.io/>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20161015/050b0b75/attachment.html <http://mitgcm.org/pipermail/mitgcm-support/attachments/20161015/050b0b75/attachment.html>>
>
> ------------------------------
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
> http://mitgcm.org/mailman/listinfo/mitgcm-support <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
>
> End of MITgcm-support Digest, Vol 160, Issue 7
> **********************************************
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
> http://mitgcm.org/mailman/listinfo/mitgcm-support <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20161018/1d3035d0/attachment.htm>
More information about the MITgcm-support
mailing list