[Wien] Problem in x-optic
Peeyush kumar kamlesh
peeyush.physik.rku at gmail.com
Sat Feb 15 19:22:03 CET 2020
Dear Gavin Sir,
opticpara_lapw file is as follows:
#!/bin/tcsh -f
#
# Run optic in parallel mode
#
# $Author: M.Lee $
#
touch .lock_
foreach i (.lock_*)
rm $i
end
onintr exit
set name = $0
set bin = $name:h #default directory for WIEN-executables
if !(-d $bin) set bin = .
set name = $name:t
unalias rm
alias testinput 'if (! -e \!:1 || -z \!:1) goto \!:2'
alias testerror 'if (! -z \!:1.error) goto error'
set t = time
set cmplx
set log = :parallel
set defmach = `hostname`
set updn # spinpolarization switch
set dnup = 'dn' # spinpolarization switch
set sc # semicore-switch
set hf # hybrid-switch
set so # spinorbit-switch
set remote = ssh
set tmp_dir = /tmp
set init = init:
set res = residue:
set taskset0
set taskset='no'
set scratch =
if ( $?SCRATCH && $SCRATCH != '') then
set scratch=`echo $SCRATCH | sed -e 's/\/$//'` # we are afraid
# different settings in different
# computing centers
#use global variable for scratch if set
endif
############################################################################
# In this section use 0 to turn of an option, 1 to turn it on,
# respectively choose a value
set useremote = 1 # using remote shell to launch processes
setenv DELAY 0.1 # delay launching of processes by n
seconds
set debug = 0 # set verbosity of debugging output
############################################################################
# and now we look if we should override the defaults
if (-e $bin/parallel_options) then
source $bin/parallel_options
endif
if ( $?TASKSET ) then
set taskset="$TASKSET"
endif
if ( $?USE_REMOTE ) then
set useremote = $USE_REMOTE
endif
############################################################################
set tmp = $tmp_dir/opticpara.$user.$$
set tmp2 = $tmp_dir/opticpara.$user.$$_2
if ($#argv < 1) then
echo usage: $0 deffile
exit
endif
while ($#argv)
switch ($1)
case -h:
case -H:
set help
shift; breaksw
case -c:
set cmplx = c
shift; breaksw
case -up:
set updn = 'up'
set dnup = 'dn'
shift; breaksw
case -dn:
set updn = 'dn'
set dnup = 'up'
shift; breaksw
case -hf:
set hf = 'hf'
shift; breaksw
case -so:
set so = 'so'
shift; breaksw
case -sc:
set sc = 's'
shift; breaksw
default:
set def = $1:r
shift; breaksw
endsw
end
set exe = $bin/optic$cmplx
set exe = optic$cmplx
#are we running parallel?
testinput .processes single
echo "running OPTIC in parallel mode"
echo "RUNNING" >.opticpara
#before we start, we wipe away all parallel error files
if ( -e optic_1.error ) rm *optic_*.error
if ( -e .timeop_1) rm .timeop_*
if (-e .machines.help) rm .machines.help
grep -v $init .processes|grep : | grep -v $res >$tmp2
set mist = `wc $tmp2 `
set maxproc = $mist[1]
#set machine = `grep $init .processes |cut -f2 -d: | xargs`
set machine = `grep -v $init .processes |grep : | grep -v $res | cut -f2
-d: | xargs`
set lockfile = `cut -f2 -d: $tmp2 | awk '{print $1 NR}'|xargs`
set residue = `grep $res .processes|cut -f2 -d:`
if ($residue == "") unset residue
if ($debug > 0) echo machines: $machine
echo "** " Error in Parallel OPTIC >$def.error
#bounding cpus
set p_cpu_bound = ($machine)
set i=1
set cpu=0
set old=old
while ($i <= $#p_cpu_bound)
if($old != $p_cpu_bound[$i]) then
set cpu=0
endif
set old=$p_cpu_bound[$i]
set p_cpu_bound[$i] = $cpu
@ cpu ++
@ i ++
end
#echo $machine
#echo $p_cpu_bound
#get name of case
setenv PWD `pwd|sed "s/tmp_mnt\///"`
#echo $PWD
setenv PWD $cwd
set case = $PWD
set case = $case:t
if ($case == "") then
echo "ERROR: cannot detect working directory $cwd -> exit"
exit 1
endif
####set case = $case:r #head of file-names
if ($debug > 0) echo Setting up case $case for parallel execution
if ($debug > 0) echo of OPTIC
if ($debug > 0) echo " "
#
#creating def files
if ($debug > 0) echo " "
if ($debug > 0) echo -n "creating "$def"_*.def: "
set i = 1
while ($i <= $maxproc)
if ($debug > 0) echo -n "$i "
cp $def.def .tmp
#subsituting in files:
cat <<theend >.script
s/vector$hf$so$dnup/&_$i/w .mist
s/vector$hf$so$updn/&_$i/w .mist
s/outputop/&_$i/w .mist
s/symmat/&_$i/w .mist
s/symma1/&_$i/w .mist
s/symma2/&_$i/w .mist
s/mommat2/&_$i/w .mist
s/mat_diag/&_$i/w .mist
s/mme/&_$i/w .mist
s/symop/&_$i/w .mist
theend
sed -f .script .tmp > .tmp1
sed "s/vector_${i}_$i\&dn/vectordn_$i/" .tmp1>.tmp2
sed "s/vector_${i}_$i\&up/vectorup_$i/" .tmp2>.tmp1
sed "s/vector_${i}dn_$i/vectordn_$i/" .tmp1>.tmp2
sed "s/vector_${i}_$i/vector_$i/" .tmp2>.tmp1
sed "s/vectorhf_${i}dn_$i/vectorhfdn_$i/" .tmp1> .tmp2
sed "s/vectorso_${i}dn_$i/vectorsodn_$i/" .tmp2> "$def"_$i.def
#similar fix for SO necessary
@ i ++
end
if ($debug > 0) echo " "
#starting processes
if ($debug > 0) echo " "
if ($debug > 0) echo "starting process: "
echo "-> "starting parallel optic at `date` >>$log
set loop = 0
set endloop = 0
set runmach = ""
echo "files:$maxproc" >.processes2
# change working dir because of problems with automounted directories
# cd $cwd
while ($loop < $maxproc)
set p = 1
if ($?residue && $?resok) set p = 2
while ($p <= $#machine)
if ($loop < $maxproc) then
if !(-e .lock_$lockfile[$p]) then
@ loop ++
echo "${loop}:${maxproc} : $p_cpu_bound[$p]" >.processes2
if ($debug > 0) echo prepare $loop on $machine[$p]
set runmach = ($runmach $machine[$p])
echo $runmach >>.processes2
if ($debug > 1) echo " > $exe ${def}_${loop}.def on $machine[$p]"
if ($debug > 1) echo " > $exe ${def}_${loop}.def on
$machine[$p]">>$log
if ($useremote == 1) then
if ($debug > 1) echo use remote
touch .lock_$lockfile[$p]
echo -n "$runmach[$loop] ">.timeop_$loop
if("$taskset" != 'no') set taskset0="$taskset
$p_cpu_bound[$p]"
($remote $machine[$p] "cd $PWD;$t $taskset0 $exe ${def}_${loop}.def;rm -f
.lock_$lockfile[$p]") >>.timeop_$loop &
else
if ($debug > 1) echo not using remote shell
touch .lock_$lockfile[$p]
echo -n "$runmach[$loop] " >.timeop_$loop
(cd $PWD;$t $exe ${def}_${loop}.def;rm -f .lock_$lockfile[$p])
>>.timeop_$loop &
endif
endif
if ($debug > 1) echo sleeping for $DELAY seconds
sleep $DELAY
jobs -l >.optic${cmplx}para.$$.`hostname`
endif
@ p ++
end
end
#wait for execution to be completed
if ($debug > 0) echo " "
if ($debug > 0) echo "waiting for processes: "
wait
set i = 1
while ($i <= $maxproc)
testerror "$def"_$i
@ i ++
end
#cpu summary:
set i = 1
while ($i <= $maxproc)
# echo " "`cat .timeop_$i`
#fix for bash timing
bashtime2csh.pl_lapw .timeop_$i > .time_tmp
mv .time_tmp .timeop_$i
echo " "`cat .timeop_$i` >>$log
@ i ++
end
# postanalysis
echo " Summary of opticpara:" >$tmp
set p = 1
while ($p <= $#machine)
set m = $runmach[$p]
cat .timeop_* | grep $m | tr "():" " " | \
awk '{u += $2; cl += 60*$4+$5} \
END {print " '$m'\t user=" u "\t wallclock=" cl}' >>$tmp
@ p ++
end
uniq < $tmp |tee -a $log
echo "<- "done at `date` >>$log
# concatenating the case.symmat files and case.mommat files
if (-e $case.symop) rm $case.symop
if (-e ${scratch}$case.mme$updn) rm ${scratch}$case.mme$updn
if (-e ${scratch}$case.symmat$updn) rm ${scratch}$case.symmat$updn
if (-e ${scratch}$case.mommat2$updn) rm ${scratch}$case.mommat2$updn
if (-e ${scratch}$case.mat_diag$updn) rm ${scratch}$case.mat_diag$updn
touch ${scratch}$case.symmat$updn
touch ${scratch}$case.mommat2$updn
touch ${scratch}$case.mat_diag$updn
touch ${scratch}$case.mme$updn
mv $case.symop_1 $case.symop
rm $case.symop_*
if( "$hf" != "hf" ) then
opticcopy_lapw
endif
set i = 1
while ($i <= $maxproc)
if ( $i == 1 ) then
# testinput $case.symmat_$i$updn scratchwarning
cat ${scratch}$case.symmat_$i$updn > ${scratch}$case.symmat$updn
cat ${scratch}$case.symma1_${i}$updn > ${scratch}$case.symma1$updn
cat ${scratch}$case.symma2_${i}$updn > ${scratch}$case.symma2$updn
if (-e ${scratch}$case.mommat2_$i$updn) then
cat ${scratch}$case.mommat2_$i$updn > ${scratch}$case.mommat2$updn
endif
cat ${scratch}$case.mat_diag_$i$updn > ${scratch}$case.mat_diag$updn
cat ${scratch}$case.mme_$i$updn > ${scratch}$case.mme$updn
else
tail -n +2 ${scratch}$case.symmat_$i$updn >>
${scratch}$case.symmat$updn
tail -n +2 ${scratch}$case.symma1_${i}$updn >>
${scratch}$case.symma1$updn
tail -n +2 ${scratch}$case.symma2_${i}$updn >>
${scratch}$case.symma2$updn
if (-e ${scratch}$case.mommat2_$i$updn) then
tail -n +2 ${scratch}$case.mommat2_$i$updn >>
${scratch}$case.mommat2$updn
endif
tail -n +2 ${scratch}$case.mat_diag_$i$updn >>
${scratch}$case.mat_diag$updn
tail -n +2 ${scratch}$case.mme_$i$updn >> ${scratch}$case.mme$updn
endif
rm ${scratch}$case.symmat_$i$updn
rm -f ${scratch}$case.mommat2_$i$updn
rm ${scratch}$case.mat_diag_$i$updn
rm ${scratch}$case.mme_$i$updn
@ i ++
end
echo "<- "done at `date` >>$log
echo
"-----------------------------------------------------------------">>$log
rm $def.error
#rm .in.tmp .in.tmp1
touch $def.error
rm $tmp* >&/dev/null
rm .optic${cmplx}para.$$.`hostname` >&/dev/null
echo "DONE" >.opticpara
exit 0
single:
echo "running in single mode"
$exe $def.def
rm $tmp* >&/dev/null
rm .optic${cmplx}para.$$.`hostname` >&/dev/null
exit 0
scratchwarning:
echo "Could not find $case.symmat_$i$updn , which is probably because you
used a scratch directory $SCRATCH"
echo "Copy these files from the remote machines and concatenate them
yourself"
echo "with commands like (for all your parallel calculations i):"
echo " cat $case.symmat_"'$i'"$updn >> $case.symmat$updn when $i
=1a "
echo " tail -n +2 $case.symmat_"'$i'"$updn >> $case.symmat$updn else"
exit 1
error:
echo "** " OPTIC crashed!
echo "** " OPTIC crashed at `date`>>$log
echo "** " check ERROR FILES! >>$log
echo
"-----------------------------------------------------------------">>$log
echo "** " Error in Parallel OPTIC >>$def.error
rm $tmp* >&/dev/null
rm .optic${cmplx}para.$$.`hostname` >&/dev/null
echo "ERROR" >.opticpara
exit 1
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20200215/80e3bae9/attachment-0001.html>
More information about the Wien
mailing list