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

Re: Use of netCDF in GrADS

Hi John,

> In case you hadn't heard, Mike Fiorino has come around, quite
> thoroughly, to the idea of GrADS supporting netCDF.  In fact, he has
> offered to actually do the work to "make it so".  Good news all
> around!  Of course, he didn't promise *when* he'd do it.....but at
> least we've made the requisite *major* course change.

That's great, I'm glad to hear it.  Thanks for arguing the case for netCDF
so articulately on the GrADS-users mailing list.  Actually, I'm not sure I
should thank you, because eventually this will mean a heavier support load
for us, but it still seems like good news.

> A quick unrelated question (if you have time) - is there an
> estimated release date for V2.4 of netCDF?  And does it really 
> remedy the performance problems on the Cray?

We're working on it, and hope to have a beta release in June, but with
vacations this might easily slip.  The time consuming part is testing the
release on everything from MSDOS and VMS platforms to Crays, fixing the
problems we find, and then retesting.

I've just talked to Jeff Kuehn, who did the Cray optimization work, and he
is sending us his changes today.  He said that he'd managed to get the
netCDF access times within 5% of Fortran binary reads and writes (for
floating-point fields) on the Cray even in bad cases such as the I/O access
patterns of the Community Climate Model.  We're still planning to include
his Cray optimizations in the 2.4 release.

--Russ