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
|
#!/usr/bin/env python
import sys, math
import argparse
from volk_test_funcs import *
try:
import matplotlib
import matplotlib.pyplot as plt
except ImportError:
sys.stderr.write("Could not import Matplotlib (http://matplotlib.sourceforge.net/)\n")
sys.exit(1)
def main():
desc='Plot Volk performance results from a SQLite database. ' + \
'Run one of the volk tests first (e.g, volk_math.py)'
parser = argparse.ArgumentParser(description=desc)
parser.add_argument('-D', '--database', type=str,
default='volk_results.db',
help='Database file to read data from [default: %(default)s]')
parser.add_argument('-E', '--errorbars',
action='store_true', default=False,
help='Show error bars (1 standard dev.)')
parser.add_argument('-P', '--plot', type=str,
choices=['mean', 'min', 'max'],
default='mean',
help='Set the type of plot to produce [default: %(default)s]')
parser.add_argument('-%', '--percent', type=str,
default=None, metavar="table",
help='Show percent difference to the given type [default: %(default)s]')
args = parser.parse_args()
# Set up global plotting properties
matplotlib.rcParams['figure.subplot.bottom'] = 0.2
matplotlib.rcParams['figure.subplot.top'] = 0.95
matplotlib.rcParams['figure.subplot.right'] = 0.98
matplotlib.rcParams['ytick.labelsize'] = 16
matplotlib.rcParams['xtick.labelsize'] = 16
matplotlib.rcParams['legend.fontsize'] = 18
# Get list of tables to compare
conn = create_connection(args.database)
tables = list_tables(conn)
M = len(tables)
# Colors to distinguish each table in the bar graph
# More than 5 tables will wrap around to the start.
colors = ['b', 'r', 'g', 'm', 'k']
# Set up figure for plotting
f0 = plt.figure(0, facecolor='w', figsize=(14,10))
s0 = f0.add_subplot(1,1,1)
# Create a register of names that exist in all tables
tmp_regs = []
for table in tables:
# Get results from the next table
res = get_results(conn, table[0])
tmp_regs.append(list())
for r in res:
try:
tmp_regs[-1].index(r['kernel'])
except ValueError:
tmp_regs[-1].append(r['kernel'])
# Get only those names that are common in all tables
name_reg = tmp_regs[0]
for t in tmp_regs[1:]:
name_reg = list(set(name_reg) & set(t))
name_reg.sort()
# Pull the data out for each table into a dictionary
# we can ref the table by it's name and the data associated
# with a given kernel in name_reg by it's name.
# This ensures there is no sorting issue with the data in the
# dictionary, so the kernels are plotted against each other.
table_data = dict()
for i,table in enumerate(tables):
# Get results from the next table
res = get_results(conn, table[0])
data = dict()
for r in res:
data[r['kernel']] = r
table_data[table[0]] = data
if args.percent is not None:
for i,t in enumerate(table_data):
if args.percent == t:
norm_data = []
for name in name_reg:
if(args.plot == 'max'):
norm_data.append(table_data[t][name]['max'])
elif(args.plot == 'min'):
norm_data.append(table_data[t][name]['min'])
elif(args.plot == 'mean'):
norm_data.append(table_data[t][name]['avg'])
# Plot the results
x0 = xrange(len(name_reg))
i = 0
for t in (table_data):
ydata = []
stds = []
for name in name_reg:
stds.append(math.sqrt(table_data[t][name]['var']))
if(args.plot == 'max'):
ydata.append(table_data[t][name]['max'])
elif(args.plot == 'min'):
ydata.append(table_data[t][name]['min'])
elif(args.plot == 'mean'):
ydata.append(table_data[t][name]['avg'])
if args.percent is not None:
ydata = [-100*(y-n)/y for y,n in zip(ydata,norm_data)]
if(args.percent != t):
# makes x values for this data set placement
# width of bars depends on number of comparisons
wdth = 0.80/(M-1)
x1 = [x + i*wdth for x in x0]
i += 1
s0.bar(x1, ydata, width=wdth,
color=colors[(i-1)%M], label=t,
edgecolor='k', linewidth=2)
else:
# makes x values for this data set placement
# width of bars depends on number of comparisons
wdth = 0.80/M
x1 = [x + i*wdth for x in x0]
i += 1
if(args.errorbars is False):
s0.bar(x1, ydata, width=wdth,
color=colors[(i-1)%M], label=t,
edgecolor='k', linewidth=2)
else:
s0.bar(x1, ydata, width=wdth,
yerr=stds,
color=colors[i%M], label=t,
edgecolor='k', linewidth=2,
error_kw={"ecolor": 'k', "capsize":5,
"linewidth":2})
nitems = res[0]['nitems']
if args.percent is None:
s0.set_ylabel("Processing time (sec) [{0:G} items]".format(nitems),
fontsize=22, fontweight='bold',
horizontalalignment='center')
else:
s0.set_ylabel("% Improvement over {0} [{1:G} items]".format(
args.percent, nitems),
fontsize=22, fontweight='bold')
s0.legend()
s0.set_xticks(x0)
s0.set_xticklabels(name_reg)
for label in s0.xaxis.get_ticklabels():
label.set_rotation(45)
label.set_fontsize(16)
plt.show()
if __name__ == "__main__":
main()
|
