The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Dear Kristopher,

I have not tried your inputs (and I do not have DEC platform here),
but I might have a suggestion beforehand.
Which kind of makefile_macros was used for your compilation of
v.4.2.4/v.4.3.3 ? If you adopt high optimization (e.g. -O3),
how about reducing the optimization level (or non-optimization flag, 
"-O0" ?).
for Src_3xc routines ? We can set F90 compiler flags for each 
Src_directory,
FFLAGS_Src_3xc = ... etc.
(We can find examples in ~ABINIT/Machine_dept_files/*/* )

Best wishes,
Masayoshi

P.S. We are glad to see your post follows our bug_report format. ;-)
http://www.abinit.org/ABINIT/bug_report.html
I have noticed that the following suggestion could be inserted in 4):

"If you compile your executables by yourself,
which kind of compiler options are adopted in your makefile_macros ?"

Thank you for your post, which let us notice this suggestion.

On 2004/09/02, at 2:29, Kristopher Andersen wrote:

> I'm getting a floating point excepting in rhohxc.f. It occurs on line
> 230. I've written out the first few entries of m_norm, and apparently
> the problem is caused because m_norm=0.0 (?!). Any ideas on what's
> going on here? It may be related to vector magnitzation, since I only
> see the problem with nspden = 4. (Only a guess.)
>
> Here's the standard bug report form. My input files are attached.
>
> 1) version: Reproducible with both 4.2.4 and 4.3.3.
>
> 2) machine: DEC Alpha (Tru64)
>
> 3) operation system: uname reports "OSF1 V5.1 2650 alpha"
>
> 4) fortran compiler: I compiled Abinit 4.2.4 with f95, but also saw
>    the problem with the precompiled binaries available in
>    compaq_ev67_seq_4.2.4.tar.gz and compaq_ev67_seq_4.3.3.tar.gz
>
> 5) description: All I see is the floating point exception. Maybe
>    something more sinister is happening earlier on?
>
> 6) error message: Here's the last few lines of log file.
>
>   getcut: wavevector=  0.0000  0.0000  0.0000  ngfft=  60  60  60
>          ecut(hartree)=      5.000   => boxcut(ratio)=   2.10286
>
>   mkdenpos : WARNING -
>    Density went < 0 at    164166 points
>    and was set to  1.00E-14.  Lowest was  -0.66E-12.
>    Likely due to too low boxcut or too low ecut for pseudopotential
>    core charge.
>    forrtl: error (75): floating point exception
>    0: __FINI_00_remove_gp_range [0x3ff81a21488]
>    1: __FINI_00_remove_gp_range [0x3ff81a2a4d0]
>    2: __FINI_00_remove_gp_range [0x3ff800d5d10]
>    3: rhohxc_ [rhohxc_cpp.f: 230, 0x1204c8ea8]
>    4: setvtr_ [setvtr_cpp.f: 267, 0x12032ba10]
>    5: scfcv_ [scfcv_cpp.f: 630, 0x1200fe1b4]
>    6: gstate_ [gstate_cpp.f: 681, 0x1200b6cc0]
>    7: driver_ [driver_cpp.f: 1172, 0x120079be4]
>    8: abinit_ [abinis_cpp.f: 504, 0x12005e858]
>    9: main [for_main.c: 203, 0x1200609dc]
>   10: __start [0x120059828]
>
> 7) wait time: Errors occurs within a minute.
>
> --
> Kristopher Andersen               http://yclept.ucdavis.edu/~andersen
> Department of Physics             mailto:andersen@physics.ucdavis.edu
> University of California at Davis                   +1 (530) 
> 754-9764<rhohxc_bug.tar.gz>