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
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
|
#!/usr/bin/env python
import sys
import time
import optparse
from os.path import basename
from os import listdir
from tempfile import NamedTemporaryFile as Tmp
from plot import decode, Plot
from gnuplot import curve
SCHEDULERS = ['MC', 'MC-MERGE', 'MC-MERGE-REDIR']
COLS = {'plugin': 0,
'overhead': 1,
'n_tasks': 2,
'samples': 3,
'filtered': 4,
'max': 5,
'avg': 6,
'min': 7,
'med': 8,
'std': 9,
'var': 10,
'iqr_max': 11,
'iqr_min': 12}
def parse_args():
p = optparse.OptionParser(description='make graphs')
p.add_option('--paper', dest='paper', action='store_true', default=False)
return p.parse_args()
def usage(msg):
print >>sys.stderr, msg
sys.exit(1)
def get_csv_files(directory):
a = []
for f in listdir(directory):
bn = basename(f)
if f.startswith('scheduler') and f.endswith('.csv'):
a.append(f)
return a
def gnuplot_col(col_name):
# gnuplot is 1-indexed
return 1 + COLS[col_name]
def get_sched_title(sched):
SCHEDULERS = {'MC': 'Basic',
'MC-MERGE': 'IM + TM',
'MC-MERGE-REDIR': 'IM + TM + WR'}
return SCHEDULERS[sched]
def get_overhead_title(ov):
OV = {'LVLA-SCHED': '(A)',
'LVLB-SCHED': '(B)',
'LVLC-SCHED': '(C)',
'LVLA-RELEASE': '(A)',
'LVLB-RELEASE': '(B)',
'LVLC-RELEASE': '(C)',
}
return OV[ov]
def set_plot_opts(opts, p):
p.rounded_caps = True
p.font = 'Helvetica'
p.default_style = 'linespoints lw 2.5'
p.default_style += ' smooth bezier'
p.key = 'off'
p.monochrome = False
p.dashed_lines = True
p.xrange = (18, 122)
p.yrange = (0, '')
if opts.paper:
ext = 'pdf'
p.size = ('8.5cm', '5.25cm')
line_width = 1.5
point_size = 0.3
p.font_size = '5pt'
else:
ext = 'png'
p.size = (1024, 768)
line_width = 3.0
point_size = 1.5
p.font_size = 'font "/usr/share/fonts/truetype/ttf-dejavu/DejaVuSans.ttf" 18'
p.output = '{0}.{1}'.format(p.output, ext)
p.format = ext
for i, c in enumerate(p.curves):
c.style = "linespoints ls %d" % (i + 1)
# try:
# # don't use yellow if we have this curve
# p.curves[5].style = "linespoints ls 7"
# except IndexError:
# pass
p.line_styles = [
(1, 'lt 1 pt 1 lw {0} ps {1} lc rgbcolor "#ff0000"'.format(line_width, point_size)),
(2, 'lt 1 pt 4 lw {0} ps {1} lc rgbcolor "#00ff00"'.format(line_width, point_size)),
(3, 'lt 1 pt 7 lw {0} ps {1} lc rgbcolor "#0000ff"'.format(line_width, point_size)),
(4, 'lt 2 pt 1 lw {0} ps {1} lc rgbcolor "#ff0000"'.format(line_width, point_size)),
(5, 'lt 2 pt 4 lw {0} ps {1} lc rgbcolor "#00ff00"'.format(line_width, point_size)),
(6, 'lt 2 pt 7 lw {0} ps {1} lc rgbcolor "#0000ff"'.format(line_width, point_size)),
(7, 'lt 3 pt 1 lw {0} ps {1} lc rgbcolor "#ff0000"'.format(line_width, point_size)),
(8, 'lt 3 pt 4 lw {0} ps {1} lc rgbcolor "#00ff00"'.format(line_width, point_size)),
(9, 'lt 3 pt 7 lw {0} ps {1} lc rgbcolor "#0000ff"'.format(line_width, point_size)),
# (1, 'lw {0} ps {1} lc rgbcolor "#ff0000"'.format(line_width, point_size)),
# (2, "lw {0} ps {1} lt 2".format(line_width, point_size)),
# (3, "lw {0} ps {1} lt3 ".format(line_width, point_size)),
# (4, "lw {0} ps {1}".format(line_width, point_size)),
# (5, 'pt 6 lw {0} ps {1} lc rgbcolor "#ff910d"'.format(line_width, point_size)),
# (6, "pt 7 lw {0} ps {1}".format(line_width, point_size)),
# (7, 'lw {0} ps {1} lc rgbcolor "#000000"'.format(line_width, point_size)),
# (8, "lw {0} ps {1}".format(line_width, point_size)),
# (9, "lw {0} ps {1}".format(line_width, point_size)),
]
def get_data_matrix(fname):
ret = []
with open(fname, 'r') as f:
for row in f:
new_row = []
for field in row.split(','):
s_field = field.strip()
try:
conv = float(s_field)
except ValueError:
conv = s_field
new_row.append(conv)
ret.append(new_row)
return ret
def include_level_a_releases(data_dir, o_type, scheduler, ycol):
fmt_f = '{0}/scheduler={1}_overhead={2}.csv'
a_fname = fmt_f.format(data_dir, scheduler, 'LVLA-RELEASE')
b_c_fname = fmt_f.format(data_dir, scheduler, 'RELEASE')
a_data = get_data_matrix(a_fname)
b_c_data = get_data_matrix(b_c_fname)
new_data = []
for i, a_row in enumerate(a_data):
b_c_row = b_c_data[i]
# make a copy of a_row
new_row = list(a_row)
if gnuplot_col('max') == ycol:
idx = COLS['max']
new_row[idx] = max(a_row[idx], b_c_row[idx])
elif gnuplot_col('avg') == ycol:
avg_col = COLS['avg']
a_samples = a_row[COLS['samples']]
b_c_samples = b_c_row[COLS['samples']]
a_filtered = a_row[COLS['filtered']]
b_c_filtered = b_c_row[COLS['filtered']]
a_avg = a_row[avg_col]
b_c_avg = b_c_row[avg_col]
n_a = a_samples - a_filtered
n_b_c = b_c_samples - b_c_filtered
new_avg = (a_avg * n_a + b_c_avg * n_b_c) / (n_a + n_b_c)
new_row[avg_col] = new_avg
else:
raise RuntimeError("Don't know how to merge this column.")
new_data.append(new_row)
f = Tmp()
for row in new_data:
f.write(', '.join(map(str, row)))
f.write('\n')
f.file.flush()
return (f.name, f)
def plot_release(opts, data_dir, ycol, title, fname):
p = Plot()
p.output = '{0}/{1}'.format(data_dir, fname)
refs = [] # need to save reference to file handle so it is not deleted
# for o_type in ['LVLA-RELEASE','LVLB-RELEASE', 'LVLC-RELEASE']:
for o_type in ['LVLA-RELEASE','LVLC-RELEASE']:
for sched in SCHEDULERS:
# if o_type == 'RELEASE':
# # we have to make the regular release include the level-A
# # releases
# fname, ref = include_level_a_releases(data_dir, o_type,
# sched, ycol)
# refs.append(ref)
# else:
fname = '{0}/scheduler={1}_overhead={2}.csv'.format(data_dir, sched, o_type)
ti = '{0} {1}'.format(get_sched_title(sched), get_overhead_title(o_type))
p.curves += [curve(fname=fname, xcol=gnuplot_col('n_tasks'),
ycol=ycol, title=ti)]
p.xlabel = 'number of tasks'
p.ylabel = 'overhead (microseconds)'
set_plot_opts(opts, p)
if gnuplot_col('avg') == ycol:
# make this graph's y-scale match the scale on the average-case
# scheduling overhead graph because they are next to each other
p.yrange = (0, 16)
# it gets the key
p.key = 'top left'
p.gnuplot_exec()
def plot_release_jim(opts, data_dir, ycol, title, fname):
"""Jim just wanted level-A without any techniques and with all our techniques."""
p = Plot()
p.output = '{0}/{1}'.format(data_dir, fname)
refs = [] # need to save reference to file handle so it is not deleted
for sched in ['MC', 'MC-MERGE-REDIR']:
o_type = 'LVLA-RELEASE'
fname = '{0}/scheduler={1}_overhead={2}.csv'.format(data_dir, sched, o_type)
ti = '{0} {1}'.format(get_sched_title(sched), get_overhead_title(o_type))
p.curves += [curve(fname=fname, xcol=gnuplot_col('n_tasks'), ycol=ycol, title=ti)]
p.xlabel = 'number of tasks'
p.ylabel = 'overhead (microseconds)'
set_plot_opts(opts, p)
p.key = 'top left'
p.title = 'level-A worst-case release overhead'
p.gnuplot_exec()
def plot_sched(opts, data_dir, ycol, title, fname):
p = Plot()
p.output = '{0}/{1}'.format(data_dir, fname)
# for o_type in ['LVLA-SCHED','LVLB-SCHED','LVLC-SCHED']:
for o_type in ['LVLA-SCHED','LVLC-SCHED']:
for sched in SCHEDULERS:
fname = '{0}/scheduler={1}_overhead={2}.csv'.format(data_dir, sched, o_type)
ti = '{0} {1}'.format(get_sched_title(sched), get_overhead_title(o_type))
p.curves += [curve(fname=fname, xcol=gnuplot_col('n_tasks'),
ycol=ycol, title=ti)]
p.xlabel = 'number of tasks'
p.ylabel = 'overhead (microseconds)'
set_plot_opts(opts, p)
p.gnuplot_exec()
def main():
opts, extra = parse_args()
if len(extra) < 1:
usage('missing args')
data_dir = extra[0]
plot_sched(opts, data_dir, gnuplot_col('max'), 'worst-case scheduling overhead', 'overhead=SCHED_type=MAX')
plot_release(opts, data_dir, gnuplot_col('max'), 'worst-case release overhead', 'overhead=RELEASE_type=MAX')
plot_sched(opts, data_dir, gnuplot_col('avg'), 'average-case scheduling overhead', 'overhead=SCHED_type=AVG')
plot_release(opts, data_dir, gnuplot_col('avg'), 'average-case release overhead', 'overhead=RELEASE_type=AVG')
plot_release_jim(opts, data_dir, gnuplot_col('max'), 'worst-case release overhead', 'overhead=RELEASE_type=MAX_for-jim=1')
if __name__ == '__main__':
main()
|
