[Wien] Questions about calculation consuming time and memory usage (parallel running)

[Yi yoo soo]

Dear, all users, 

 

I'm almost new in this field and I have many troubles to perform..

 

I crushed on the problems using Wien2k package on the k-point parallel
calculation. Recently I tested CASTEP in MS4 to see the some features from
EELS spectrum and also starts learning Wien2k to use TELNES module. Anyway,
in a simple structure of MgO (with making 2x2x2 super cell about 60 atoms in
the whole system) the calculation consuming time of ELLS spectrum on CASTEP
is about 15,000 seconds. However, using Wien2k with no parallelization
option, the single SCF cycle consumes about almost 10 hrs. I think that
there is a some mistake on my settings of dimension parameters, etc. It
takes to long time, even the MgO is very simple system. I summarize the
processes which I did to calculated ELNES spectrum on Wien2k. Please help me
to solve this problem. 

 

  _____  

(1)     Used structural information: MgO structure with 2x2x2 super cell
generated with x_supercell

(2)     Details of setting (case.in* files): I just followed the setting
from the *.pdf files uploaded in web page.

init_lapw (with 1000 k-point, RKmax=6.5)

edit case.in1 (increase the energy maximum values to calculate higher energy
region in TELNES from default to 4.5 eV)

edit case.in2 (add one valence electron)

run_lapw -it (use iterative diagonalization, convergence criteria is set to
charge 0.0001)

telnes (for Mg K-edge spectrum, setting edge onset, etc., region from 0 eV
to 35.0 eV)

I did not set the broadening variables (just see the possibility of
calculating spectra)

(almost other variables are used with default values)

 

(3)     Results

With no k-point parallelization option, with no super cell, the calculation
result is almost same as previously reported data. However, if I check the
parallelization option, the calculation result of spectrum is cut below
about 12 eV region. And if I used parallel option in calculating ELNES
spectrum, the results is same as above (the data is not appeared below 12 eV
region, seems to be not calculated). Using super cell structure, I could not
test the calculation result without parallel option yet. 

  _____  

 

 

The above descriptions are the calculation results and settings, then below
is the .machines files and lapw1 dimension parameters.

  _____  

I use the system using two CPUs (each of CPU is quad intel Xeon) with RAM 24
Gb (i.e., DELL Precision work station). 

OS: RHE 5.3

Wien2k is compiled with ifort 11 and mkl10.4. 

While compiling the program, in site_configuration step, I set the shared
memory architecture used option. 

 

(1)     Dimension parameter

NMATMAX = 50,000

NUME = 5,000

I decided these values from the size of memory. However, when I type 'top
-c' in the terminal, lapw2 or other programs use only 2~3 % RAM. Is it OK?

 

(2)     .machines file

1:localhost (total 8 lines of local host, because the total thread is 8)

.

1:localhost

Granularity:1

Extrafine:1

  _____  

 

What I have to modify the files or setting to calculate the ELNES spectrum
correctly? 

Small attentions and opinions would be very helpful to me.

 

Sincerely,

Yoo Soo

 

 

School of Earth and Environmental Science, Seoul National University, Korea

Yi yoo soo

e-mail: yys2064 at snu.ac.kr  OR  yiyoosoo at gmail.com

office: +82-2-877-3072

mobile: +82-10-2655-2064

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20091016/da4d80d8/attachment.htm>