Floating point exception of VASP with gfortran (or g95) at Linux AlphaServer

Questions regarding the compilation of VASP on various platforms: hardware, compilers and libraries, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
Brane
Newbie
Newbie
Posts: 14
Joined: Sat Feb 05, 2005 3:53 am
License Nr.: 866
Location: U.S.

Floating point exception of VASP with gfortran (or g95) at Linux AlphaServer

#1 Post by Brane » Thu Mar 22, 2007 9:26 am

Hi, all

I am trying to compile vasp using gfortran (4.1.1) at Alpha Linux boxes. I can compile vasp sucessfully (thanks for lahaye's previous message). However, when I execute vasp, it thows out a Floating point exception to me after running tens of seconds. To check out what's happening, I look at the OURCAR file and find vasp stopped at the beginning of electronic relaxaion.

Code: Select all

$tail OUTCAR

    METAGGA=      F    non-selfconsistent MetaGGA calc.

 Electronic Relaxation 1
   ENCUT  =  300.0 eV  22.05 Ry    4.70 a.u.   5.45  5.45  5.45*2*pi/ulx,y,z
   ENINI  =  300.0     initial cutoff
   ENAUG  =  241.9 eV  augmentation charge cutoff
   NELM   =     60;   NELMIN=  2; NELMDL= -5     # of ELM steps
   EDIFF  = 0.1E-03   stopping-criterion for ELM
   LREAL  =      F    real-space projection
   LCOMPAT=      F    compatible to vasp.4.4

I turned on the FFLAG -mieee when compiling GotoBLAS, vasp.4.lib, and vasp.4.6. (I have also tried using the FFLAG -mfp-trap-mode=su, but the FPE error also exists.)

Also, if I use g95 to compile vasp, the floating point exception error haunts. But vasp went further

Code: Select all

$tail OUTCAR
 initial charge density was supplied:
 charge density of overlapping atoms calculated
 number of electron    8.0000000 magnetization
 keeping initial charge density in first step


----------------------------------------------------------------------------------------------------
 Maximum index for augmentation-charges  26563 (set IRDMAX)


Any sugestions and comments are appreciated.

Thanks
Brane
Last edited by Brane on Thu Mar 22, 2007 9:26 am, edited 1 time in total.

lahaye
Jr. Member
Jr. Member
Posts: 98
Joined: Fri Apr 14, 2006 5:08 am
Location: Suwon - Korea

Floating point exception of VASP with gfortran (or g95) at Linux AlphaServer

#2 Post by lahaye » Tue Mar 27, 2007 5:50 am

Hi Brane,

I'm glad my previous info was useful to you. Apparently I'm not
the only one who's crunching numbers on an old Alpha system :).

I'm running Linux/CentOS 4.4 on my Alpha system. What is your OS?
I am using the GCC suite 4.1.3 from SVN. I don't know if that
makes a difference with your 4.1.1 concerning the FPE problem.

I can send you my Makefiles, if you wish.
Let me know: (MyName) @ skku.edu

Lahaye.
Last edited by lahaye on Tue Mar 27, 2007 5:50 am, edited 1 time in total.

Brane
Newbie
Newbie
Posts: 14
Joined: Sat Feb 05, 2005 3:53 am
License Nr.: 866
Location: U.S.

Floating point exception of VASP with gfortran (or g95) at Linux AlphaServer

#3 Post by Brane » Sat Apr 14, 2007 1:53 pm

Let me draw a conclusion and close this thread.

In summary, I failed to compile vasp using gfortran and g95 in both Gentoo 2006.1 and Centos 4.3 Alpha Linux. Luckily enough, a friend in local HP was consulted and we got a CFAL under the EE license. However, the libcxml BLAS was useless since it was compiled with the older GLIBC. Alternatively, the blas-3.0.5.alpha.rpm was used. Now VASP runs happily in Gentoo Alpha Linux--kernal 2.6.16 at the Alpha boxes. (I discarded Gentos 4.3 Alpha Linux since the kernel (2.6.9.34EL.smp) has a bug --couldnot allocate 8 bytes percpu.)

Finally, I would like to thank Lahaye for his kind comments and private communications.

Regards,
Brane
Last edited by Brane on Sat Apr 14, 2007 1:53 pm, edited 1 time in total.

Post Reply