The ABINIT Users Mailing List ( CLOSED )

[haiou haiou <haiouwyj@gmail.com>]

Anglade,
   It's so nice of you to give me such valuable advices. As for the segmentation fault, i have searched it with google and found some helpful information. There are many possibilities that may cause this problem. It happens whenever the stacksize is not large enough while running the calculation. On the linux platform, we can increase the stack size using "ulimit -s unlimited", which in my opinion can only solve the problem temporarily. And we can use the method you provide to find exactly where in the abinit program this error occurring.
   And i got another problem. I calculated the imaginary spectrum and Reflectance spectrum of Al. It was not smooth enough with several peaks as could be seen in the picture. The reflectivity calculated differs insufferably from the values measured experimently. Can you have a look at my input file and tell me if any variable is not set reasonable? Your favorable assistance will be highly appreciated.



ndtset 6

#First dataset : SC run with kpoints in the IBZ
   iscf1  3
  nband1  3     !
 nstep1 1000   
 kptopt1 1
#nbdbuf1 0
 prtden1 1   getden1 0   getwfk1 0    ! Usual file handling data

#Second dataset : NSC run with large number of bands, and points in the IBZ
    iscf2 -2  
   nband2 60  !!!!!! This number of bands might be too low for non-linear optics and real part of linear optics
   nstep2 1000
  kptopt2  1
  getwfk2  1   getden2 1   ! Usual file handling data

#Third dataset : NSC run with large number of bands, and points in the the full BZ
    iscf3 -2  
   nband3 60  !!!!!! This number of bands might be too low for non-linear optics and real part of linear optics 
   nstep3 1000
  kptopt3  1
  getwfk3  2   getden3 1   ! Usual file handling data


#Fourth dataset : ddk response function along axis 1
   iscf4 -3
  nband4 60   !!!!!! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep4  1  nline4  0
 kptopt4  1

   nqpt4  1  qpt4  0.0d0 0.0d0 0.0d0 
  rfdir4  1 0 0
 rfelfd4  2
 getwfk4  3

#Fifth dataset : ddk response function along axis 2
   iscf5 -3
  nband5 60   !!!!!! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep5  1  nline5  0
 kptopt5  1

   nqpt5  1  qpt5  0.0d0 0.0d0 0.0d0
  rfdir5  0 1 0
 rfelfd5  2
 getwfk5  3

#Sixth dataset : ddk response function along axis 3
   iscf6 -3
  nband6 60   !!!!!! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep6  1  nline6  0
 kptopt6  1

   nqpt6  1  qpt6  0.0d0 0.0d0 0.0d0
  rfdir6  0 0 1
 rfelfd6  2
 getwfk6  3


#Data common to all datasets
 nshiftk 4                !
 shiftk  0.5 0.5 0.5
         0.5 0.0 0.0
         0.0 0.5 0.0
         0.0 0.0 0.5
 ngkpt  60 60 60        !!!!!! This is much too low : should be at least 24x24x24

 acell   3*7.60         
 ecut 12.00             !!!!!! This is also too low
 occopt 3
 iscf 3
 ixc 3
 natom  1  nbdbuf 2
 ntypat  1
 rprim   0 .5 .5  .5 0 .5  .5 .5 0
 xred 3*0.00d0
 tnons 144*0.0
 typat  1  tolwfr  1.e-20
 znucl  13
 
############################parall##########

# paral_kgb 1
 npband 1 npfft 1 npkpt 2
 timopt -1
 fftalg 401 wfoptalg 4 nloalg 4 fft_opt_lob 2 
 iprcch 0 intxc 0

2009/12/17 Anglade Pierre-Matthieu <anglade@gmail.com>

Hi,

Unfortunately, I may not be able to be very helpful.
Since your calculation works with ngkpt 60 60 60, it looks like the
segfault you observe is likely linked to the number of kpt used. So
far I guess we can not say much more than that.
 Knowing exactly what is going on inside the code at the segfault
moment would likely give you (and us) more information on what shall
be corrected. If you have the possibility to do so, could you compile
the code again but with debug flags enabled ? And then run your
calculation monitoring each process with a debugger ? Finally when a
process fail, you would be able to backtrace the events and find what
exactly has caused the segmentation violation.

For instance, you would do the following :
configure --enable-debug
make
mpirun -np xxx abinit ....
top or ps -u login to get the PID of each parallel process launched
(on each used node of your cluster)
gdb abinit PID (on each node again)
bt

regards

PMA

On Wed, Dec 16, 2009 at 4:38 PM, haiou <haiouwyj@gmail.com> wrote:
> Can anybody help me ?
>



--
Pierre-Matthieu Anglade

Attachment: Reflectance spectrum.jpg
Description: JPEG image