Re: super computing python

From: Davide Del Vento <ddvento_at_nyahnyahspammersnyahnyah>
Date: Fri Dec 02 2011 - 10:03:45 MST

Hi Jim and all,

I am the mentor for that internship. What Mary says is correct - what
she says is state of the art. But science (including computer science)
continually evolves.

Despite being (relatively) young, I remember when in this industry
people preferred using Assembler instead of Fortran/C because it was
faster (I did write some programs in Assembler myself for this very
reason). Things changed when CPUs got more complex, scientific models
much more sophisticated and Fortran/C compilers got smarter. Now, I'm
not aware of any scientist using Assembler anymore (are you?).

I believe that we might be at the edge of a similar revolution. CPUs are
getting even more complex (but not faster, at least not from the
single-threaded point of view), don't even mention the increase of
complexity in scientific models, and yes compilers are improving too,
but most of the gains from this side are coming in a different way (from
JITs). On top of this, performance increase is expected to come from the
model themselves, even making them "better" or porting them to some
"very strange" and very hard to use architectures, like a GPU.

Maybe another level of abstraction will be added, and in a decade people
might be writing scientific models in python, looking at current Fortran
models like we look today at those cryptic Assembler programs. Or maybe
not, and that's why it's a good research project :-)

Since Mary mentions, I am also the chair of the SEA, and in particular
the chair of the SEA seminar series. If you have a nice talk you think
will be good for a seminar, please let me know privately.
Moreover, SEA is organising a conference. If your talk is of higher
profile, you should submit if for the conference, instead of just a
But please hurry, the deadline is approaching (you need just the title
and abstract for now, so you can do it in 10 minutes)

Final note: I'm very busy these days, so I'm not monitoring the mailing
list, please send me emails if you do expect an answer. Or - better yet
- if you'll be at AGU next week, stop by the NCAR booth, and meet me in

Davide Del Vento,
NCAR Computational & Information Services Laboratory
Consulting Services Software Engineer
SEA Chair
office: Mesa Lab, Room 55G

>> > <>
>> > <>
>> > Date: Thu, 1 Dec 2011 09:39:44 -0700
>> > <>
>> > Subject: Re: super computing python
>> >
>> > Hi Jim,
>> >
>> > Thanks for pointing this out!
>> >
>> > This position is for a summer student intern in our SIParCS (Summer
>> > Internships in Parallel Computational Science) program. The progra m
>> > runs for about 10-11 weeks and usually involves researching some topic
>> > in parallel computing.
>> >
>> > There are certainly pockets of Python users at NCAR, but not a wide
>> > adoption. We have a Python email list, and it hardly gets any traffic.
>> >
>> > The big models are usually written in compiled languages, and then
>> > people use individual tools of their own choice for post-processing:
>> > NCL, Python, Grads, IDL, Matlab, R, and others.
>> >
>> > This reminds me: we have monthly software engineering seminars, if
>> > anybody is interested in coming to NCAR and giving a talk on how they
>> > use Python in the climate/atmospheric sciences. I believe they can pay
>> > for travel and a small honorarium. I think this topic would be of
>> > great interest at NCAR, especially if you can appeal to both the
>> > scientists and software engineers.
>> >
>> > If anybody is interested in looking into this, send me an email and
>> > I'll put you in touch with the seminar folks.
>> >
>> > --Mary
>> >
>> > On Nov 30, 2011, at 2:22 PM, Boyle, Jim wrote:
>> >
>> > > Does this job indicate an increased interest in python at NCAR?
>> > >
>> > > <>
>> > >
>> > > --Jim Boyle
>> > > _______________________________________________
>> > > pyngl-talk mailing list
>> > > List instructions, subscriber options, unsubscribe:
>> > >
>> >
>> > _______________________________________________
>> > pyngl-talk mailing list
>> > List instructions, subscriber options, unsubscribe:
>> >
pyngl-talk mailing list
List instructions, subscriber options, unsubscribe:
Received on Fri Dec 2 10:03:50 2011

This archive was generated by hypermail 2.1.8 : Wed Jan 04 2012 - 10:54:01 MST