Hi Bergy,
On 11/1/12 10:09 PM, "BERGY" <nucleic81.gmail.com> wrote:
>Dear All,
> I had been request by a third party to check a petascale machine. I am
>new
>to this new concept.
>1. On what does a petascale simulation depend on?. i guess it the number
>of
>atoms to simulate!.
>2. Is there similar benchmarking done for amber12 and published before.
>3. if not how to go about.
This is an extremely non-trivial question and would require substantially
more information about what you actually want to do and what you want to
achieve than what you provide here.
Firstly it would be useful to know which petascale machine you are
referring to. At the last count (
http://www.top500.org/list/2012/06/100)
there were only 20 of them on the planet so it is a fairly small club.
Second the concept of a petascale simulation is a very woolly concept,
really it is one grounded more in politics rather than science. It would
be better to define what the scientific question is you want to address
and then design the simulation to answer this. There are many different
types of simulation you could envision running on a large scale machine
and the approach you take very much depends on the scientific question you
are asking.
I would start by defining exactly what it is you want to test. If you want
to find out if a 100,000 core machine will run AMBER simulations stably
then I suggest running 100,000 independent simulations and seeing which
ones crash. If instead you want to run a single calculation across 100,000
cores then your best approach is to use a replica exchange approach
(ideally with multiple dimensions and a large number of replicas) but
again the simulation you want to run will depend on the scientific
question you want to ask and not just 'I want to run a petascale
simulation'.
If you can provide us with more information about what you want to
accomplish we can try to help you. But I warn you now that such
simulations are definitely not for the faint hearted.
All the best
Ross
/\
\/
|\oss Walker
---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
|
http://www.rosswalker.co.uk |
http://www.wmd-lab.org |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Nov 01 2012 - 22:30:03 PDT
