Page 1 of 1

Non-linear optics dataset splitting

Posted: Thu Aug 05, 2010 1:24 pm
by schumann
Hi,

I'm trying to do some nonlinear response calculation. The test file:
/abinit-6.2.1/tests/tutorespfn/Input/tnlo_2.in
works just fine for me.

But if I deal with my real problem, it always stopped till now in the 5th dataset without an (usable) error message.

So I want to know, if there is any possibility to do only this last dataset. I can change all the get* in ird*, but as I understood there is no ird1den variable which could replace get1den. Is there some workaround for this?

Thanks in advance for your help.

Silvia

Re: Non-linear optics dataset splitting

Posted: Sat Sep 04, 2010 10:32 am
by mverstra
we need to know where it stops (output and log) and to get the unusable error message, which always helps. Also, platform, abinit version etc...

Matthieu

Re: Non-linear optics dataset splitting  [SOLVED]

Posted: Thu Sep 23, 2010 8:20 am
by schumann
Hi,

since output and log file are too big to attach, I try to copy the main information here.

The end of the output file shows no error, just stops writing this:

getshell : finite difference formula of Marzari and Vanderbilt
(see Marzari and Vanderbilt, PRB 56, 12847 (1997), Appendix B)

number of first neighbours : 2
weight : 1411.3440000

number of second neighbours : 6
weight : 282.2688000



This is the end of the log file:

-P-0000 == DATASET 5 ==================================================================
-P-0000
[...]
mkfilename : getwfk/=0, take file _WFK from output of DATASET 2.

mkfilename : get1wf/=0, take file _1WF from output of DATASET 4.

mkfilename : getden/=0, take file _DEN from output of DATASET 1.

mkfilename : get1den/=0, take file _DEN from output of DATASET 4.

[...]
getshell : finite difference formula of Marzari and Vanderbilt
(see Marzari and Vanderbilt, PRB 56, 12847 (1997), Appendix B)

number of first neighbours : 2
weight : 1411.3440000

number of second neighbours : 6
weight : 282.2688000
list of nband
isppol, ikpt, nband_k= 1 1 16
isppol, ikpt, nband_k= 1 2 16
[...]
-P-0000 leave_test : synchronization done...
kpgio: loop on k-points done in parallel
-P-0000
-P-0000 leave_new : decision taken to exit ...

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

=== Build Information ===
Version : 6.2.1
Build target : x86_64_linux_intel11.0
Build date : 20100726

=== Compiler Suite ===
C compiler : gnu
CFLAGS : -O3 -mp1 -ip -xW
C++ compiler : gnu11.0
CXXFLAGS : -m64 -g -O2
Fortran compiler : intel11.0
FCFLAGS : -O3 -mp1 -ip -xW
FC_LDFLAGS : -static-libgcc -static-intel

=== Optimizations ===
Debug level : yes
Optimization level : standard
Architecture : unknown_unknown

=== MPI ===
Parallel build : yes
Parallel I/O : no

=== Linear algebra ===
Library flavor : @linalg_flavor@
Use ScaLAPACK : yes

=== Plug-ins ===
BigDFT : yes
ETSF I/O : yes
LibXC : yes
FoX : no
NetCDF : yes
Wannier90 : yes

=== Experimental features ===
Bindings : no
Exports : no
GW double-precision : no
Macroave build : yes

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Job /licsoft/lsf/6.2/linux2.6-glibc2.3-x86_64/bin/openmpi_wrapper abinit

TID HOST_NAME COMMAND_LINE STATUS TERMINATION_TIME
===== ========== ================ ======================= ===================
00000 p2s086 abinit Exit (1) 08/24/2010 10:50:17
00001 p2s099 abinit Exit (1) 08/24/2010 10:50:17
00002 p2s112 abinit Exit (1) 08/24/2010 10:50:17
00003 p2s117 abinit Exit (1) 08/24/2010 10:50:17
[...]



Hope, this is all the information you need. If not, just ask.

Thanks, Silvia

P.S. And I try to configure my account to get an email notification if some answer comes in... next reply should be more in time...

Re: Non-linear optics dataset splitting

Posted: Mon Oct 11, 2010 12:40 pm
by mverstra
ird1den is implicit: if a den is needed the dataset will look for it on disk. If you are just running 1 dataset without anything and have not specified get1den then the expected file name is

tnlxi_DS5_1DENxx

or something like that, according to the name you chose in the files file.

If your job finishes with
=== Build Information ===

you probably have an error message before this. Otherwise, you'll have to run abinit from within a debugger (does the input work in sequential?). Once on the machine, either
gdb PATHTOABINIT/abinit
gdb> run < t.files

or if abinit is already running get its process id (pid) and run:
gdb PATHTOABINIT/abinit PID
gdb> cont
...

you should be able to tell where it is stopping (routine, line...) and maybe why. You have no other out/err files for your run?

Matthieu