[Pw_forum] How to using gamma point calculation with high efficiency

[vega lew]

Dear Axel,

> that just means you are not using your MPI library correctly.
> QE assumes that consecutive MPI task ids are on the same node.
> most MPI libraries decide this based on the host/machine file,
> you have to read your MPI documentation how to arrange it the
> way it should would. you've seen from my timings, that it should
> work quite well. i am using OpenMPI and that package has on top
> of using hosts in the hostfile order the option to explicitely
> follow a round-robin or a by-node scheme when assigning MPI tasks.

I'm using MPICH2. So before using it, I should  type command ' mpdboot -n 5 -f mpd.mashines'.
Maybe MPICH2 is not arrange the MPI task IDs in the way that Q-E assumes. So I'll try OPENMPI.
By the way, what vision does you recommend me to install?

> whether or not a gamma point calculation is sufficient is not
> determined by what other people do, but by the physics of the
> problem. 

I have realized that. Late time maybe I doesn't say it in a right way. I'm using gamma point calculation
is not all because other people do so, but also many scientific papers of high impact and high quality, 
such as paper of American Physical Society and American Chemical Society said using 2X2X1 mesh of k points
result in no large difference in energy of geometry from the gamma point only calculation. It's accepted 
by most of scientists do the similar job for many years. I think it's not necessary to validate it by my self
because of the weakness of our calculation condition.

>that reminds me, you still owe us an explanation about
> the difference between 'commercial' and 'academic' calculations.
> i am _very_ curious about this.

I don't have enough authority to illustrate the difference beteen the two. I only saying something from my
own brain. I think Q-E is more suitable for scientific research, because of its free and open-source. And 
another very important thing is Q-E also accepted by many resarchers who used Q-E and doing excellent
calculation. Also many pieces of work published on high quality and high impact Journals, like Nature Materials,
Journal of the American Chemical Society et al... But for comercial software, it cost a lot of money and the
source code is not open for their copyright. So it's much more like a black box. We can't know any detail that
how the results comes out. Maybe it's more accurate, but we can't believe it until we see the source code.
But if we know the source code, their secret is out. As far as I know, commercial software is more suitable for 
industry researcher such as China National Petroleum Corporation or China Petrochemical Corporation et al.
The industry resarchers is often unwilling to say whether the resarch work using commercial simulation software 
is accurate or not. We only see the product of the industry, but can't see what's the role of the software
in the product exploitation. Over all, people especially the scientists maybe couldn't trust the commercial software 
until the see the source. Most scientist can't bear the big uncertain factor that the software brings to their resarching.

> you cannot. as you should have already seen from the numbers
> i sent you. g-space parallelization doesn't scale across gigabit
> ethernet using TCP/IP networking. just run the job within one
> node and then you can run 5 jobs at the same time.

Do you think the Q-E can achieve ~80 atoms gamma point calculation in one computer with single quad-core CPU
and 8Gmemory in less than 1 month running? Maybe it's too unbelievable task. 

I'll try OPENMPI later, thank you for helping.


Vega Lew
PH.D Candidate in Chemical Engineering
State Key Laboratory of Materials-oriented Chemical Engineering
College of Chemistry and Chemical Engineering
Nanjing University of Technology, 210009, Nanjing, Jiangsu, China



_________________________________________________________________
News, entertainment and everything you care about at Live.com. Get it now!
http://www.live.com/getstarted.aspx
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20080629/d1e242e3/attachment-0001.htm