[CPMD-list] Spectra Issue on BG/L

[Matteo Guglielmi]

Hello everybody,

I'm having troubles running spectra analysis on BG/L.
(water solvated system = 318 atoms)

#### INPUT FILE ####

&CPMD
  RESTART WAVEFUNCTION COORDINATES LATEST
  ELECTRONIC SPECTRA
  LANCZOS DIAGONALIZATION
  MAXSTEP
  10000
  CONVERGENCE ORBITAL
  1.0D-6
  LANCZOS PARAMETER
  3000 8 -1 1.0D-8
  MIRROR
&END

&TDDFT
  TAMM-DANCOFF
  STATES SINGLET
  6
  DAVIDSON PARAMETER
  800 1.D-8 50
  DAVIDSON RDIIS
  20 3 1.D-9
&END

&DFT
  NEWCODE
  FUNCTIONAL PBE
&END

&LINRES
  CONVERGENCE
  1.0D-5
  MAXSTEP
  1000
  HTHRS
  0.5
  OPTIMIZER PCG
&END

&SYSTEM
  POISSON SOLVER TUCKERMAN
  ANGSTROM
  CHARGE
  0.0
  SYMMETRY
  8
  CELL ABSOLUTE
  15.0 16.0 12.0 0.0 0.0 0.0
  CUTOFF
  70.0
&END

##################

#### ERROR MESSAGE ######

          GRADIENT CORRECTION ENERGY =           -8.73816660 A.U.
 
 NFI      GEMAX       CNORM           ETOT        DETOT      TCPU
   1  4.785E-07   1.354E-08   -1798.337215    0.000E+00      4.16

 RESTART INFORMATION WRITTEN ON FILE                  ./RESTART.1

 TIME FOR MINIMUM STRUCTURE :                     117.134 SECONDS
 ***   SPECTRA|  THE TOP OF THE STACK IS           114 MBytes ***
 
 ================================================================
 ==                END OF REFERENCE CALCULATION                ==
 ==               GENERATE INITIAL GUESS VECTORS               ==
 ================================================================

 ***   SPECTRA|  THE TOP OF THE STACK IS           114 MBytes ***
 

 <<<<<<<<<<<<<<<<<<< LANCZOS DIAGONALIZATION <<<<<<<<<<<<<<<<<<<<
 >> TIME FOR INITIAL SUBSPACE DIAGONALIZATION:               5.55
 >> CYCLE     NCONV        B2MAX        B2MIN     #HPSI      TIME
 UNIV| THE LEADING MINOR OF ORDER    1 IS NOT POSITIVE DEFINITE,
 UNIV| THE LEADING MINOR OF ORDER    1 IS NOT POSITIVE DEFINITE,
....

######################


The error message does not occur running the same job on a different
architecture.


Any help is greatly appreciated.