Installation problem with VASP.5.2
Posted: Wed Nov 17, 2010 5:30 pm
Dear VASP users:
The installation of VASP.5.2 has troubled us for several weeks.
We've compiled VASP.4.6 on Suse10 plateform successfully using Intel 11 compiler and have submitted some testing jobs downloaded from the VASP workshop for verification. It runs without any problem even for other advanced systems. When it came to VASP.5.2, the same installation procedure was valid; however, the testing jobs showed different result comparing to which published on the Handson pdf. For example, the calculation of work function for pure nickel (3_6_Ni111clean_400eV) gave totally nonsense in local potential profile, and the calculated LDOS result (3_7_COonNi111 LDOS) could not match with the exhibited plot.
After many tries, we found that the correct result could be expected to come out only when the system is cubic (where NGX=NGY=NGZ). Another critical problem is that the job would fail without any error information when it went to the next ionic step for large systems (more than about 40 atoms):
----------------------------------------- Iteration 1( 1) ---------------------------------------
----------------------------------------- Iteration 1( 2) ---------------------------------------
... …
----------------------------------------- Iteration 1( 18) ---------------------------------------
--------------------------------------------------------------------------------------------------
POTLOK: cpu time 0.09: real time 0.09
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
WAVPRE: cpu time 0.02: real time 0.02
FEWALD: cpu time 0.00: real time 0.00
GENKIN: cpu time 0.02: real time 0.02
(JOB ENDED HERE, NO INFORMATION BELLOW…)
I went over the old threads to check if someone encountered the same problem. It seems like the stack size limit may be the cause. Then I saw this:
--
a) set KMP_STACKSIZE explicitely
(setenv KMP_STACKSIZE=2m would e.g. set it to 2 Megabytes)
please ask your sys admin about the maximum available stack size and/or to increase the stack size maximum..
b) check the stack limit of your shell (“ulimit-s"); if it is limited, please set “ulimit -s unlimited" for your shell (in the .bashrc file for example)
--(by admin 2/26/2008)
I tried, but still in vain.
Hope to hear some suggestions from you.
Thanks!
Hsin
The installation of VASP.5.2 has troubled us for several weeks.
We've compiled VASP.4.6 on Suse10 plateform successfully using Intel 11 compiler and have submitted some testing jobs downloaded from the VASP workshop for verification. It runs without any problem even for other advanced systems. When it came to VASP.5.2, the same installation procedure was valid; however, the testing jobs showed different result comparing to which published on the Handson pdf. For example, the calculation of work function for pure nickel (3_6_Ni111clean_400eV) gave totally nonsense in local potential profile, and the calculated LDOS result (3_7_COonNi111 LDOS) could not match with the exhibited plot.
After many tries, we found that the correct result could be expected to come out only when the system is cubic (where NGX=NGY=NGZ). Another critical problem is that the job would fail without any error information when it went to the next ionic step for large systems (more than about 40 atoms):
----------------------------------------- Iteration 1( 1) ---------------------------------------
----------------------------------------- Iteration 1( 2) ---------------------------------------
... …
----------------------------------------- Iteration 1( 18) ---------------------------------------
--------------------------------------------------------------------------------------------------
POTLOK: cpu time 0.09: real time 0.09
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
WAVPRE: cpu time 0.02: real time 0.02
FEWALD: cpu time 0.00: real time 0.00
GENKIN: cpu time 0.02: real time 0.02
(JOB ENDED HERE, NO INFORMATION BELLOW…)
I went over the old threads to check if someone encountered the same problem. It seems like the stack size limit may be the cause. Then I saw this:
--
a) set KMP_STACKSIZE explicitely
(setenv KMP_STACKSIZE=2m would e.g. set it to 2 Megabytes)
please ask your sys admin about the maximum available stack size and/or to increase the stack size maximum..
b) check the stack limit of your shell (“ulimit-s"); if it is limited, please set “ulimit -s unlimited" for your shell (in the .bashrc file for example)
--(by admin 2/26/2008)
I tried, but still in vain.
Hope to hear some suggestions from you.
Thanks!
Hsin