From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Tue Apr 15 2014 - 18:22:54 CDT
please *always* copy the mailing list on your replies. the fact that i
answered doesn't make me your personal consultant.
On Wed, Apr 16, 2014 at 12:50 AM, Hassan Kianinejad <hasan_at_cs.ucla.edu> wrote:
> Actually I am a Computer Science major and trying to Accelerate this application myself. I thought it would be more convenient for me to start from a sequential code in order to be easier to understand it when I read it. However the benchmarks included in NAMD-Lite (like "alanin" and "medwater") are not big enough for simulation purposes. I would like to have 100,000 to 1 million particles to simulate on NAMD-Lite (which I believe would be really slow).
there are two major problems here:
1- it is very dangerous to try to accelerate a piece of software
without understanding what it does. your project is doomed from the
2-- it is rather pointless to try to optimize a code that was meant to
do method development and *not* to be efficient. for such larger
systems it would be utterly useless to try and optimize a non-parallel
version unless you have a "magic spell" that will make it 10000x or
more faster (than regular NAMD).
> I know many users do not care about this problem because most of them are using it rather than caring about the implementation; however I would be really glad if you could guide me through the process please.
what process? you have to begin by taking some classes in in physics
(classical mechanics and statistical thermodynamics for starters) and
come back in a year or two. what you are asking for is the job of an
adviser and i most certainly am not interested to be an adviser on
such an ill-advised project.
> Hassan Kianinejad
> Graduate Student
> VLSI CAD Lab, UCLA
> ----- Original Message -----
> From: "Axel Kohlmeyer" <akohlmey_at_gmail.com>
> To: "Hassan Kianinejad" <hasan_at_cs.ucla.edu>
> Cc: "NAMD list" <namd-l_at_ks.uiuc.edu>
> Sent: Tuesday, April 15, 2014 3:42:32 PM
> Subject: Re: namd-l: NAMD-Lite
> On Wed, Apr 16, 2014 at 12:22 AM, Hassan Kianinejad <hasan_at_cs.ucla.edu> wrote:
>> Dear NAMD users
>> I have a question and I asked this many times, but no one replied.
> asking the same question many times, doesn't make it a better
> question. if you don't get an answer, then there is likely a reason:
> - nobody knows the answer
> - nobody cares
> - people think that you should figure out the answer for yourselfui
> - your question is ill posed
>> I want to use NAMD benchmarks to run on NAMD-lite which is the sequential version; however the configuration files are different and they do not support each other.
> they don't, because they are different entities. from the NAMD-lite home page:
> NAMD-Lite is a rapid prototyping framework for developing new
> simulation methods for biomolecules. It is a sequential, C language
> code base consisting of modular parts that are easy to modify and
> reuse. The intention is to separate the development of the method from
> the additional complication due to parallel implementation, providing
> a simpler way to test and validate new methods before bringing them
> into NAMD.
> that doesn't sound at all like it is a serial version of NAMD. more
> like NAMD-lite is a code for people that know what they are doing
> (which apparently doesn't apply as much to you). what is the reason,
> you simply don't use NAMD as it is?
>> I wondered if anyone could help me figure out a solution.
> a solution to what? you are only asking half a question. how can you
> expect a meaningful answer. i don't see a (NAMD) problem in need of
> solving here.
>> Hassan Kianinejad
>> Graduate Student
>> VLSI CAD Lab, UCLA
> Dr. Axel Kohlmeyer akohlmey_at_gmail.com http://goo.gl/1wk0
> College of Science & Technology, Temple University, Philadelphia PA, USA
> International Centre for Theoretical Physics, Trieste. Italy.
-- Dr. Axel Kohlmeyer akohlmey_at_gmail.com http://goo.gl/1wk0 College of Science & Technology, Temple University, Philadelphia PA, USA International Centre for Theoretical Physics, Trieste. Italy.
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:20:44 CST