Mercurial > hg > cc > cirrus_work
comparison bin/_ex1.sh @ 2:b4801f5696b2
compute node workers, see cirrus_home/bin repo for login node masters
author | Henry S. Thompson <ht@inf.ed.ac.uk> |
---|---|
date | Mon, 18 Jul 2022 19:22:42 +0100 |
parents | |
children | 668579197bec |
comparison
equal
deleted
inserted
replaced
1:d5b6748f29a9 | 2:b4801f5696b2 |
---|---|
1 #!/bin/bash | |
2 # This runs on the compute nodes... | |
3 # count top 21 solitary languages in parallel, taking input directly from /work/dc007/dc007/hst/results/$1/cdx_counts/xxx.tsv | |
4 | |
5 N=$SLURM_JOB_NUM_NODES | |
6 n=$SLURM_NTASKS | |
7 c=$SLURM_CPUS_PER_TASK | |
8 nodename=$SLURMD_NODENAME | |
9 local=$SLURM_LOCALID | |
10 node=$SLURM_NODEID | |
11 task=$SLURM_PROCID | |
12 | |
13 threadsPerTask=2 | |
14 pjobs=$((c / $threadsPerTask)) | |
15 | |
16 cc=$1 | |
17 resdir=$W/$USER/results/$cc/$2 | |
18 srcdir=$W/hst/results/$cc/cdx_counts | |
19 langs=$3 | |
20 s1=$4 | |
21 sn=$5 | |
22 | |
23 echo $(date) task $n.$task on $nodename:$N.$node start | |
24 | |
25 mkdir -p $resdir | |
26 | |
27 doit () { | |
28 echo $(date) start $1 $task $PARALLEL_SEQ | |
29 fgrep ' w ' $srcdir/$1.tsv | awk 'BEGIN { | uniq -c | \ | |
30 $W/shared/bin/uniq_merge.py > $resdir/${langs}_$1.tsv | |
31 echo $(date) end $1 $task $PARALLEL_SEQ | |
32 } | |
33 | |
34 export -f doit | |
35 export srcdir resdir task | |
36 | |
37 seq $s1 $sn | while read i | |
38 do if [ $((i % $n)) -eq $task ] | |
39 then printf '%03g\n' $i | |
40 fi | |
41 done | \ | |
42 parallel -j $pjobs doit '{}' | |
43 | |
44 echo $(date) task $n.$task on $nodename:$N.$node end | |
45 |