[Wien] lapw1c ERROR on Opteron

[Sorin Lazar]

Dear Prof. Blaha

I think that I found the reason for the crash of lapw1c!  Unfortunately I 
don't think that I will be able to solve it without your help.
What I did was to take two cases one which requires complex calculation and 
one which doesn't.
I run the calculation (only the first SCF cycle) on a machine where I never 
had problems (DEC Alpha ) and I did the same on the new Opteron machine.
For the case which doesn't require complex calculation the results were 
exactly the same (checking various file like case. vsp, .output0,etc) on 
both machines.
For the case which require complex calculation I found tremendous 
differences. The problem why lapw1c crashes is that in case.vsp (on the 
Opteron machine) the numbers are aberrant (3 to 10 orders of magnitude 
higher then on the Alpha machine).
Maybe you have an idea why is that.
What makes the difference for lapw0 when calculates a case which is non 
complex and a case which requires complex calculation.
If you have any idea please let me know!

With many thanks,
Sorin

Example:

ALPHA machine:

    VLM(R) FOR L  0   M= 0

    -0.620662580230E+02-0.620685468463E+02-0.620697665523E+02-0.620707254729E+02
    -0.620718779445E+02-0.620729258446E+02-0.620740029915E+02-0.620750494961E+02

Opteron machine:

  VLM(R) FOR L  0   M= 0

    -0.236375390974E+05-0.239375327240E+05-0.242413425879E+05-0.245490180716E+05
    -0.248606088216E+05-0.251761643580E+05-0.254957352668E+05-0.258193725823E+05




At 10:29 2003-11-12 +0100, you wrote:

> > I managed to compile successfully the whole package with no errors using
> > two different libraries , lapack/blas from the distribution( thanks to your
> > tip! ) and using 'acml' which is an Opteron optimized library from AMD.
> > I run some tests with a simple structure (TiC) which doesn't require
> > complex calculation and everything works fine.
> > When I tried to do some calculation on a simple structure which require
> > complex calculation everything (struct. gen, initialization) works fine
> > till 'scf'. Lapw1c gives the same error message:
> > "
> > Cholesky INFO =             1
> >   'SECLR4' - POTRF (Scalapack/LAPACK) failed.
> > "
>
>One guess: When installing "manually", i.e. with      make
>you get only the "real" version. You must use         make complex
>to compile and generate    lapw1c
>
>Otherwise, I have no idea.
>
>
>                                       P.Blaha
>--------------------------------------------------------------------------
>Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
>Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
>Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
>--------------------------------------------------------------------------
>
>_______________________________________________
>Wien mailing list
>Wien at zeus.theochem.tuwien.ac.at
>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>
>
>---
>Incoming mail is certified Virus Free.
>Checked by AVG anti-virus system (http://www.grisoft.com).
>Version: 6.0.528 / Virus Database: 324 - Release Date: 2003-10-16
-------------- next part --------------

---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.528 / Virus Database: 324 - Release Date: 2003-10-16