[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 20040116: Support for Unidata products on Cray systems

>To: address@hidden
>From: Kevin Thomas <address@hidden>
>Subject: Support for Unidata products on Cray systems
>Organization: Cray Inc.
>Keywords: 200401161952.i0GJqlp2017040 netCDF Cray

Hi Kevin,

> I work in the Applications Support group at Cray.  One
> of my jobs is to try to provide users of our systems
> with the software that they need.  I'm interested in
> getting changes put into some of the Unidata software
> to make it easier to use for our customers.
> 
> In particular, I have made some modifications to netCDF
> to port it to Cray's most current system, Cray X1.
> I'd like to get these changes integrated back into the
> netCDF source distribution, such that Cray X1 can be
> a supported platform.  Could you let me know how I
> might go about getting this done?

Thanks for the generous offer of help!  We would like to be able to
get your changes into the distribution also, but we currently have
no access to a Cray X1 platform for testing.

We're currently working on testing and releasing netCDF 3.5.1, based
on version 3.5.1-beta13, available from:

  ftp://ftp.unidata.ucar.edu/pub/netcdf/netcdf-beta.tar.Z

The best thing for us would be for you to provide patches to that
version or even better to the 3.5.1 release candidate that should be
ready soon, assuming changes are still needed.  It's also necessary to
make sure any changes made to support the Cray X1 don't affect the
performance on other platforms adversely.  We're currently revising
the way we configure and test netCDF on multiple platforms, so that
may be affected also.  Finally, it would be good to have a contact to
whom we could forward support questions about netCDF on the Cray X1 if
they were clearly platform-specific.

We're also developing a version 3.6.0 that will support 64-bit offsets
for large files, based on changes contributed by Greg Sjaardema at
Sandia.  And we're trying to make sure we don't diverge too far from a
Parallel netCDF effort by a group at Argonne/Northwestern that is
undergoing testing on multiple supercomputer platforms.  So you can
see the problem of integrating changes may be a little
over-constrained :-).
                 
We can't promise to integrate a large set of changes that we don't
understand, but changes for portability or performance that are
clearly written, maintainable, and have platform-specific aspects
well-isolated would be ideal ...

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden          http://www.unidata.ucar.edu/staff/russ