[Wien] big QTL-B when excutables compiled with ifort10.1 and mkl10.0

[Xing Gao]

Thank Dr. Blaha,

I will try to compare the calculation step by step.

Regarding the NEVL=1000, I was misled by the compilation with ifort10.1 because even though I increased it to NEVL=2000, I could not get through this problem until I changed the optimization -O3 to -O2 on our machine. And with lower optimization, even though I did not change it to 2000, it worked for cases of more than 1000 number of eigenvalues. Therefore I thought it was problem of reallocate of arrays.

P.S. I do have case of less than 100 atoms in one unit cell and because of the f electrons and semi-core d electrons, the total number of electrons is bigger than 1000.

Best,

Gao

--- On Thu, 12/4/08, Peter Blaha <pblaha at theochem.tuwien.ac.at> wrote:
From: Peter Blaha <pblaha at theochem.tuwien.ac.at>
Subject: Re: [Wien] big QTL-B when excutables compiled with ifort10.1 and mkl10.0
To: "A Mailing list for WIEN2k users" <wien at zeus.theochem.tuwien.ac.at>
Date: Thursday, December 4, 2008, 1:26 AM

> Recently I am trying to run wien2k_08.2 on  an x86_64 machine with Intel 
> Xeon processors. There is no problem with the compilation (ifort 10.1 
> and mkl 10.0) and it works fine for several calculations. However, I did 
> get one problem in a system containing rare-earth element Yb. The 
> program stopped at the very first iteration with big QTL-B values. I am 
> sure the structure file is correct. I have the same RMT values for all 
> atoms, the Kmax is not very big, and the Fermi level is about 0.43. 
> Normally I should not expect this error. Actually, I did this 
> calculation with WIEN2k_06.3 compiled with PGI compiler and ACML library 
> on another machine with AMD processor before. There is no problem at 
> all. Therefore, I do not think that the ghost state is real here. My 
> question is if there is something different in version 06_3 and 08_2 (I 
> checked the update info, it seems this is not the case, I do not use 
> iterative diagonalization and run the program with k-point 
> parallelization only) or it is due to the compilation?

When you have a working installation, compare the two calculations step 
by step, starting with the initialization,....

> I also have another trouble related to spaghetti. There is no problem to 
> run spaghetti with total number of eigenvalues less than 1000, but will 
> get segmentation fault with total number of eigenvalues more than 1000 
> on both machines mentioned above. When I debug the program and the error 
> is "Subscript out of range for array k_name (spag.f: 162)". Has
anybody 
> met this kind of problem before?

edit spag.f and change
       NEVL=1000
to a larger value.
PS: 1000 electrons means typically 100 atoms and usually a 
"bandstructure" for such large unit cells is not really very 
interesting.....
_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20081204/87d00e05/attachment.html