[Wien] Fermi error again

[Torsten Andersen]

So it works for small systems...

then I still think you have to check for compiler errors on the SGI. Try 
to set the optimization level to zero (-O0) when you compile, and check 
if it still behaves like you mentioned below. I suspect that an 
optimization limits the size of certain arrays that will be fatal for 
"larger" systems. I have seen this on an Opteron, where everything works 
fine for the test case, but as soon as one starts a more complicated 
case, it doesn't work...

If the zero optimization also doesn't work, try to use blas_lapw instead 
of SGI's BLAS.

Best regards,
Torsten Andersen.

Nandan Tandon wrote:
> 
> Thanks for your reply.
> Yes, i get same results for other 32 atom systems and smaller systems 
> which i have been working on, and as i mentioned the version is also the 
> same.
> 
> Regards,
> 
> Nandan.
> 
> On Tue, 14 Feb 2006, Torsten Andersen wrote:
> 
>> Dear Nandan,
>>
>> I would suspect a compiler error on the SGI. Does it work for other 
>> cases?
>>
>> If so, do you get the same results for any case on both machines?
>>
>> Best regards,
>> Torsten Andersen.
>>
>> Nandan Tandon wrote:
>>
>>>
>>> I had sent the following querry sometime back without getting any 
>>> response. now, i have even tried different option in the in2c file 
>>> such as GAUSS and TEMP but i still get only 2 energy eigen values as 
>>> mentioned below and the program stops with "error in lapw2" (on the 
>>> SGI machine).
>>> Please give me some suggestions.
>>>
>>> Regards,
>>>
>>> Nandan.
>>>
>>>
>>>
>>> I am performing a 72 atom system calculation on two machines:
>>> 1. SGI Altix (itanium 2) with shared memory (32Gb).
>>>    NUME=5000 and NMATMAX=10000
>>> 2. PIV 1GB RAM, 2.0GHz
>>>    NUME=2000 and NMATMAX=10000
>>>
>>> I am trying to get the self consistancy over single k-point, with 
>>> TETRA option in case.in2c file. The first 3 lines of the case.in2c on 
>>> both the above machines are:
>>>
>>> TOT             (TOT,FOR,QTL,EFG,FERMI)
>>>       -9.0     650.0 0.50 0.05                EMIN, NE, ESEPERMIN, 
>>> ESEPER0
>>> TETRA    0.000          (GAUSS,ROOT,TEMP,TETRA,ALL      eval)
>>>
>>> The total number of eigen values evaluated are:
>>> Machine 1---> 2                   Machine 2--->678
>>>
>>>
>>> On the machine 1 the SCF cycle ends with error in lapw2c
>>>
>>> 'FERMI' - EFERMI OUT OF ENERGY RANGE
>>>
>>> The energy range in the case.in1c file is identical too.
>>> My question is, why does the same 72 atom system give 2 different 
>>> results on different machines? Is there some other parameter which 
>>> has to be changed on the machine 1 for successful SCF completion?
>>>
>>> I am using the same version of WIEN2K on both the machines (VERSION 
>>> WIEN2k_04.10 (Release 8/10/2004), which i am in the process of 
>>> updating to the latest version).
>>>
>>>
>>
>>
> 

-- 
Dr. Torsten Andersen        TA-web: http://deep.at/myspace/
AG Hübner, Department of Physics, Kaiserslautern University
http://cmt.physik.uni-kl.de    http://www.physik.uni-kl.de/