[Wien] save_lapw problem with W2WEB

2023-09-05 Thread Fecher, Gerhard
Dear Peter,
I have an issue with save_lapw (Wien2k 23.2) when calling it from w2web after a 
parallel calculation.
using -d savedir (by checking the box: Save calculation in a directory as 
specified) results in
save_lapw -d -eels savedir
this does not happen for a non-parallel calculation.
It seems that the p or -p from the parallel execution is transferred to the 
"savelapw.pl" script
and eels is executed because of "$cmdline .= " -eels" if ($p)" in line 50
maybe its better to associate eels with a different character than p.

further, from W2Web, when saving with old to a directory, the command is
 save_lapw -o -d savedir
and the -d overwrites -o such that the new saving scheme is used and not the 
old one

the same appears for save_lapw -o -f or save_lapw -o -s, or other combinations 
where one of -d, -f, -s appears after -o.
reason is the 'set new' in the save_lapw script in the parts
  case -[D|d]:
  case -[F|f]:
  case -[S|s]:

I think 'set new' is not necessary to be repeated for -d, -f, -s, because it is 
already done before the while statements,
or is there a special need to have it always in such cases and to supress the 
old saving scheme?
(unset new appears only in -o and -nodel)

In my tests, -d -eels savedir or -d -o savedir did not work at all, also from 
command line.

To use -o -d savedir, I also changed the following (mainly just copied from the 
new scheme) in save_lapw:
(but this should be cross-checked to be correct or complete)

# old save scheme
if ($?new) goto new
echo 'Fallback to compatibility mode with "old" save_lapw'
foreach i (clmsum clmup clmdn dmatup dmatdn dmatud vorbup vorbdn vorbud struct 
vrespsum vrespup vrespdn eeceup eecedn r2v r2vdn r2v2 r2v2dn r2v_nonloc 
r2v_nonlocdn r2v_half r2v_halfdn tausum tauup taudn )
#old#if (! -z $file.$i && -e $file.$i )  cp -p $file.$i $savefile.$i
if ($savefile == "") then
set to = $i
else
set to = $savefile.$i:e
endif
#if (-e $savedir/$to)  then
#if !($?force) goto exists
#endif
if !($?silent) echo "   $i -> $savedir/$to"
if (! -z $i && -e $i) cp -p $i $savedir/$to
end

set hftest=`grep "HYBRID EIGENVALUES" $file.scf`
if($#hftest != 0) then
foreach i ( vectorhf vectorhfup vectorhfdn vectorhfso vectorhfsoup 
vectorhfsodn energyhf energyhfup energyhfdn energyhfso energyhfsoup 
energyhfsodn energyhf_rbz energyhfup_rbz energyhfdn_rbz weighthf weighthfup 
weighthfdn weighthf_rbz weighthfup_rbz weighthfdn_rbz weighthfnoso 
weighthfnosoup weighthfnosodn corewf corewfup corewfdn)
#old#if (! -z $file.$i && -e $file.$i ) cp -p $file.$i $savefile.$i
if ($savefile == "") then
   set to = $i
else
   set to = $savefile.$i:e
endif
if !($?silent) echo "   $i -> $savedir/$to"
if (! -z $i && -e $i) cp -p $i $savedir/$to
end
echo "HF calculation detected and corresponding vectorhf, energyhf and 
weighthf files are also saved (provided they are NOT on '$SCRATCH'"
endif

if (-e $file.scf )  mv $file.scf $savefile.scf
if(! $?nodel) then
if (-e $file.broyd1)  rm $file.broyd*
echo 'broyden files deleted, clm*, dmat*, vorb*, vresp*, tau*,  eece*, scf 
and struct files saved under' $savefile
endif
exit (0)

# new save scheme, includes input files etc.




Optional remark:
It would also be nice if the hidden .machinesXXX as well as other numbered 
(XXX) files are removed with clean_lapw and/or save_lapw.
For example, I guess the informations of all the case.output1up_XXX files is 
also contained in case.output1up
and similar for output1dn, eenergyup, energydn, and maybe others.
When using many cores, the things become sometimes a little unclear and after a 
while searching/inspecting becomes slow even though harddisc space doesn't 
matter much to me.

In version 21.1 I used the following (without removing empty files), but this 
might not be complete for 23.2:

echo "cleaning `pwd ` "
(ls $file.vec* $file.help* $file.vrespva* $file.vrespco* $file.clmsc0* 
$file.clmsc1* $file.clmscup0* $file.clmscup1* $file.clmscdn0* $file.clmscdn1* 
$file.clmval_* $file.clmvalup_* $file.clmvaldn_* $file.output0[1-9]0[0-9] 
$file.output00[1-9][0-9] $file.output0[1-9][1-9][0-9] $file.output1_*_proc* 
$file.output1up_*_proc* $file.output1dn_*_proc* $file.output2_* 
$file.output2up_* $file.output2upeece_* $file.output2dn_* $file.output2dneece_* 
$file.recprlist $file.scfdm_* $file.scfdmup_* $file.scfdmdn_* $file.dmat_[0-9]* 
$file.dmatup_[0-9]* $file.dmatdn_[0-9]* $file.dmatud_[0-9]* > cleanfilelist1 
)>& /dev/null
set a=`cat cleanfilelist1`
(ls $file.storeHinv* $file.nsh* $file.vint* $file.storeHinv* $file.nval* 
$file.broy* $file.*_old *~ $file.corew* fort.* ftn* *.error > cleanfilelist2 
)>& /dev/null
set a1=`cat cleanfilelist2`
(ls .command* .running* .lapw?para .time?_* .tmp? :parallel* This_file* 
.processes .script .mist* .time_* :log STDOUT $file.klist_[0-9]* $file.weight* 
.in.tmp* *.def *.scf1*_* *.scf2*_* *.tmp* > 

Re: [Wien] Problem with libiomp5.so and error messages that disappear and do not stop the calculation

2023-09-05 Thread Peter Blaha

In addition to what Jan already answered:

I am running WIEN2k version 23.2 on a machine with Ubuntu 22.04.3, 
6.2.0-31-generic Linux Kernel, OpenMP and gfortran+openblas because 
according to the user guide (section 11.1.1) FFTW3, ELPA and LIBXC will 
most probably work only with gfortran and not with ifort.


This statement concerns the libraries which come with Ubuntu. All those 
libraries are compiled using gcc and gfortran and thus not compatible 
with ifort.
When you want to use ifort, you need to compile these libraries 
yourself, which is not terrible complicated (see UG, Installation).


By the way, when I made the volume optimization (constant a:b:c ratio) 
at the start it appeared the message “Error in LAPW0” in lapw0.error 
file but that did not stop the calculation, at the end the message had 
disappeared and there were not more hints about an error in 
case.dayfileor STDOUTfiles. In fact I could even plot the curve of 
energy versus volume.


This is the "normal" behavior in WIEN2k. When a program is started, an 
error message is written into program.error and when te programs 
finishes regularely, the message will be deleted and empy error files 
remain. So nothing to worry.

--

--
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300
Email: peter.bl...@tuwien.ac.atWIEN2k: http://www.wien2k.at
WWW:   http://www.imc.tuwien.ac.at
-
___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html