1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
|
#!/bin/sh
# This shell runs a bunch of benchmarks on some specific files
# available at Unidata. If you want to run this shell, you need these
# data files.
# This script gets and benchmarks against some 2D radar data.
# $Id: run_bm_radar_2D.sh,v 1.9 2008/01/16 13:27:01 ed Exp $
set -e
# Radar 2D file. Make sure we have a local disk copy. Not much point
# in benchmarking read and write times over NFS!
TMP=/shecky/data
d1=20070803-2300
file_num=0
for t in 1 2 4
do
file=${d1}_tile${t}-2d.nc3
in_file[$file_num]=$file
let file_num=$file_num+1
if ! test -f $TMP/$file; then
echo "getting file: $file"
cp -f /upc/share/testdata/nssl/mosaic2d_nc/tile${t}/$d1.netcdf.gz $TMP
gunzip -f $TMP/$d1.netcdf.gz
cp $d1.netcdf $TMP/$file
fi
done
num_in_files=${#in_file[@]}
# Copy the 2D rarar file into a netCDF-4 version, with various
# CHUNKING settings.
out1=radar_2d_chunking.csv
out2=radar_2d_chunking_2.csv
rm -rf $out1 $out2
# Turn on header (for the first run of bm_file).
h=-h
# Turn off compression and shuffle filters.
s=0
d=-1
# file_num=0
# for c0 in 251 1001 1501
# do
# for c1 in 251 1001 2001
# do
# # Build the command including chunk sizes for all 13 vars.
# cmd="./bm_file $h -f 4 -o $TMP/$d1-2d_${c0}x${c1}.nc4 -c 0:${d}:${s}:${c0}:${c1}"
# for ((v=1; v < 12; v++))
# do
# cmd="$cmd,${v}:${d}:${s}:${c0}:${c1}"
# done
# cmd="$cmd $TMP/${in_file[${file_num}]}"
# echo "cmd=$cmd"
# if ! ($cmd); then
# exit 1;
# fi
# h=
# # Switch to the next input file of three.
# let file_num=$file_num+1
# test $file_num -eq $num_in_files && file_num=0
# done
# done
file_num=0
for c0 in 251 1001 1501
do
for c1 in 251 1001 2001
do
for try in 0 1 2 3 4 5 6 7 8 9
do
# Confuse the disk buffering by copying the file each time, so
# always reading a new file.
# cp $TMP/${in_file[${file_num}]} $TMP/cp_${in_file[${file_num}]}
# Build the command including chunk sizes for all 13 vars.
cmd="./bm_file $h -f 4 -c 0:${d}:${s}:${c0}:${c1}"
for ((v=1; v < 12; v++))
do
cmd="$cmd,${v}:${d}:${s}:${c0}:${c1}"
done
cmd="$cmd $TMP/${in_file[${file_num}]}"
echo "cmd=$cmd"
sudo bash ./clear_cache.sh
if ! ($cmd >> $out1); then
exit 1;
fi
cmd="./bm_file $h -f 3 -c 0:${d}:${s}:${c0}:${c1}"
for ((v=1; v < 12; v++))
do
cmd="$cmd,${v}:${d}:${s}:${c0}:${c1}"
done
cmd="$cmd $TMP/$d1-2d_${c0}x${c1}.nc4"
echo "cmd=$cmd"
sudo bash ./clear_cache.sh
if ! ($cmd >> $out2); then
exit 1;
fi
# Remove the copy. Next read will a "new" file.
# rm $TMP/cp_${in_file[${file_num}]}
# Turn off header next time around.
h=
# Switch to the next input file of three.
let file_num=$file_num+1
test $file_num -eq $num_in_files && file_num=0
done
done
done
exit 0
|
