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
|
# Copyright 2015 ARM Limited
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import math
import statistics
import math
from functools import reduce
import utils_print
stats_diff_headers = ['mean1', 'stdev1 (% of mean1)', 'mean2',
'stdev2 (% of mean2)', '(mean2 - mean1) / mean1 * 100']
stats_diff_formats = ['.3f', '.3f', '.3f', '.3f', '.3f']
stats_headers = ['min', 'max', 'mean', 'stdev', 'stdev (% of mean)']
stats_formats = ['.3f', '.3f', '.3f', '.3f', '.3f']
def CalcGeomean(nums):
assert len(nums) != 0
return math.exp((sum(map(lambda x: math.log(x), nums))) / len(nums))
def CalcGeomeanErrorImpl(nums, stdevs, geomean, length):
assert len(nums) == len(stdevs)
assert length != 0
return math.sqrt(sum(map(lambda x, y: (x / y) ** 2, stdevs, nums))) * geomean / length
def CalcGeomeanError(nums, stdevs, geomean):
return CalcGeomeanErrorImpl(nums, stdevs, geomean, len(nums))
def CalcGeomeanRelationError(nums1, nums2, stdevs1, stdevs2, geomean):
assert len(nums1) == len(nums2) == len(stdevs1) == len(stdevs2)
nums = nums1 + nums2
stdevs = stdevs1 + stdevs2
return CalcGeomeanErrorImpl(nums, stdevs, geomean, len(nums1))
def GetRelativeDiff(x1, x2):
return (x2 - x1) / x1 * 100 if x1 else float("inf")
def ComputeStats(nums):
m = min(nums)
M = max(nums)
ave = statistics.mean(nums)
d = statistics.pstdev(nums, ave)
dp = d / ave * 100 if ave != 0 else float("inf")
return m, M, ave, d, dp
def GetSuiteName(benchmark):
return benchmark.split("/", 2)[1]
def ComputeGeomean(dict_results):
if not dict_results: return
stats_dict = {}
for benchmark in dict_results:
suite_name = GetSuiteName(benchmark)
if (suite_name not in stats_dict):
stats_dict[suite_name] = {}
stats_dict[suite_name][benchmark] = dict_results[benchmark]
# Overall and per suite geomean calculations.
mean_list = []
stdev_list = []
results = []
for suite_name in stats_dict:
suite_mean_list = []
suite_stdev_list = []
for benchmark in stats_dict[suite_name]:
m, M, mean, stdev, dp = ComputeStats(stats_dict[suite_name][benchmark])
suite_mean_list.append(mean)
suite_stdev_list.append(stdev)
mean_list.append(mean)
stdev_list.append(stdev)
suite_geomean = CalcGeomean(suite_mean_list)
suite_geomean_err = CalcGeomeanError(suite_mean_list, suite_stdev_list, suite_geomean)
results.append([suite_name, suite_geomean, suite_geomean_err,
suite_geomean_err / suite_geomean * 100])
geomean = CalcGeomean(mean_list)
geomean_err = CalcGeomeanError(mean_list, stdev_list, geomean)
results.append(['OVERALL', geomean, geomean_err, geomean_err / geomean * 100])
return results
def ComputeAndPrintGeomean(dict_results):
if not dict_results: return
results = ComputeGeomean(dict_results)
print("GEOMEANS:")
headers = ['suite', 'geomean', 'error', 'error (% of geomean)']
utils_print.PrintTable(headers, ['.3f'] * len(headers), results)
# Print a table showing the difference between two runs of benchmarks.
def PrintDiff(res_1, res_2, title = ''):
# Only print results for benchmarks present in both sets of results.
# Pay attention to maintain the order of the keys.
benchmarks = [b for b in res_1.keys() if b in res_2.keys()]
if not benchmarks: return
headers = [title] + stats_diff_headers
results = []
stats_dict = {}
# collecting data
for bench in benchmarks:
suite_name = GetSuiteName(bench)
if (suite_name not in stats_dict):
stats_dict[suite_name] = {}
stats_dict[suite_name][bench] = []
data1 = m1, M1, ave1, d1, dp1 = ComputeStats(res_1[bench])
data2 = m2, M2, ave2, d2, dp2 = ComputeStats(res_2[bench])
stats_dict[suite_name][bench].append(data1)
stats_dict[suite_name][bench].append(data2)
diff = GetRelativeDiff(ave1, ave2)
results.append([bench, ave1, dp1, ave2, dp2, diff])
utils_print.PrintTable(headers, ['.3f'] + stats_diff_formats, results)
# overall and per suite geomeans calculations
print("\nGEOMEANS:")
mean_list1 = []
mean_list2 = []
stdev_list1 = []
stdev_list2 = []
headers = ['suite', 'geomean', 'error', 'error (% of geomean)']
results = []
for suite_name in stats_dict:
suite_mean_list1 = []
suite_mean_list2 = []
suite_stdev_list1 = []
suite_stdev_list2 = []
for benchmark in stats_dict[suite_name]:
bench_mean1 = stats_dict[suite_name][benchmark][0][2]
bench_mean2 = stats_dict[suite_name][benchmark][1][2]
bench_stdev1 = stats_dict[suite_name][benchmark][0][3]
bench_stdev2 = stats_dict[suite_name][benchmark][1][3]
suite_mean_list1.append(bench_mean1)
suite_mean_list2.append(bench_mean2)
suite_stdev_list1.append(bench_stdev1)
suite_stdev_list2.append(bench_stdev2)
mean_list1.append(bench_mean1)
mean_list2.append(bench_mean2)
stdev_list1.append(bench_stdev1)
stdev_list2.append(bench_stdev2)
suite_geomean = CalcGeomean(suite_mean_list2) / CalcGeomean(suite_mean_list1)
suite_geomean_err = CalcGeomeanRelationError(suite_mean_list1, suite_mean_list2,
suite_stdev_list1, suite_stdev_list2, suite_geomean)
results.append([suite_name, suite_geomean, suite_geomean_err,
suite_geomean_err / suite_geomean * 100])
geomean = CalcGeomean(mean_list2) / CalcGeomean(mean_list1)
geomean_err = CalcGeomeanRelationError(mean_list1, mean_list2,
stdev_list1, stdev_list2, geomean)
results.append(['OVERALL', geomean, geomean_err, geomean_err / geomean * 100])
utils_print.PrintTable(headers, ['.3f'] * len(headers), results)
|
