[Octopus-users] octopus-1.4 with ifort 8.1 on SGI Altix 350
Fa-gung.Fan at xeroxlabs.com
Tue May 17 05:34:37 WEST 2005
Dear Professor Marques,
Thank you very much. After adding "LCAOStart = no" to the bench-gs input file, the case ran.
I also did few more tests with the case (bench-gs) and found that:
1) If parameter Spacing was set to 0.25 (or a larger value), the case ran even without adding "LCAOStart = no" to the input file.
2) If parameter Spacing was set to 0.15 (or a smaller value), a Segmentation fault terminated the job (even when "LCAOStart = no").
Somehow, fine mesh caused the Segmentation fault (i.e., when the number of grid points exceeded certain value). It looks like the problem was related to failed memory allocation or array size exceeded while assigning values to the array.
I did not use netCDF. Can that be the reason?
From: Miguel Marques [mailto:marques at teor.fis.uc.pt]
Sent: Saturday, May 14, 2005 8:31 AM
To: Fan, Fa-gung
Cc: octopus-users at tddft.org
Subject: Re: [Octopus-users] octopus-1.4 with ifort 8.1 on SGI Altix 350
Dear Fa-Gung Fan
It is indeed a strange behavior. I have no idea why it happens. In any
case, you can skip the LCAO step, by just adding "LCAOStart = no" to the
input file. The code will still converge although slower.
Sex, 2005-05-13 às 12:47 -0400, Fan, Fa-gung escreveu:
> Dear users and developers:
> I compiled octopus 1.4 on an SGI Altix 350 system (running Linux) with
> Intel Fortran compiler 8.1 (ifort). fftw-2.1.5 was used, and the
> BLAS/Lapack in lscs library were used. The code seemed to compile OK.
> When tested with /share/octopus/samples/Na2, it worked. However, with
> "bench-gs" in the same directory, it terminated with a Segmentation
> fault when it got to "call X(lalg_gemm)(np*dim, nset, norbs, R_TOTYPE
> (M_ONE), ...)" in subroutine "lcao_wf".
> Also tested with C2H2 and SiH4, it worked for these two cases as
> Any clue why this happened?
> Fa-Gung Fan
> Octopus-users mailing list
> Octopus-users at tddft.org
More information about the Octopus-users