[CPMD-list] code problem with TISTACK(NTIMX)

[Juerg Hutter]

Hi Christian

this type of error is related to unmatched
start and stop calls for the timer routine.
(TISTART - TIHALT)
One known such case has been introduced in version 3.7.2
in fnonloc.F for the case of Vanderbilt and/or Goedecker PP.
The easy bug fix is at line 526 of fnonloc.F

      IF (TIVAN.OR.TISGP) THEN
        CALL FNONLOC(C2,F,NSTATE,AUXC,DDIA,IKIND,ISPIN)
        CALL TIHALT('   FNONLOC',ISUB)
        RETURN
      ENDIF

However, it's possible that you found another such case.
Let us know if this doesn't help.

regards

Juerg Hutter

----------------------------------------------------------
Juerg Hutter                   Phone : ++41 1 635 4491
Physical Chemistry Institute   FAX   : ++41 1 635 6838
University of Zurich           E-mail: hutter at pci.unizh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
----------------------------------------------------------


On Tue, 18 May 2004, Christian Tuma wrote:

> Dear CPMD developers,
>
> I run a number of geometry optimisations that always crashed after
> geometry step 198 or 199. Upon closer investigation I found the reason
> for that behaviour:
> In "rlbfgs.F" line 306
>
>             CALL DCOPY (NDIM, DXPAR, 1, WLBFGS(IGPNT+NPT_L+1), 1)
>
> the variable IGPNT is suddenly changed to an unreasonable high value
> leading to a crash of DCOPY.
> The reason for that sudden change of IGPNT is an overflow of the index
> of the vector TISTACK ! In "time.inc" I found the following declarations:
>
>       PARAMETER      (NTIMX=200)
> ...
>       CHARACTER*(LENTIM) TISTACK(NTIMX)
>
> Later on in "timer.F" lines 44,45
>
>       ITSTACK=ITSTACK+1
>       TISTACK(ITSTACK)=SNAME
>
> the variable ITSTACK exceeds the value of NTIMX, i.e. it gets bigger
> than 200.
> I tried to set NTIMX to 250 instead of 200, but this only postponed
> the crash to geometry step 247. From that I conlcude that the number
> of geometry steps cannot exceed the value of NTIMX. Since that looks
> like a bug to me I would like to ask the developers of CPMD to have
> a look at this problem since I have no idea about the concept and
> realisation of the timing routines etc..
>
> (The platform I used was a Linux-PC, Redhat 9, ifc 7 and ifort 8,
> CPMD 3.7.2.)
>
> Thank you!
>
> cheers,
> Christian.
>
> --
> Christian Tuma             Humboldt-Universitaet Berlin
> ct at chemie.hu-berlin.de     Arbeitsgruppe Quantenchemie (Prof. Sauer)
> phone: +49-30-20937140     Brook-Taylor-Str. 2, 12489 Berlin, GERMANY
> fax:   +49-30-20937136     http://www.chemie.hu-berlin.de/ag_sauer
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list
>