/
benchmark.py
executable file
·204 lines (167 loc) · 7.47 KB
/
benchmark.py
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
#!/usr/bin/python
import subprocess
import re # regex
import argparse
all_benchmarks = {
'normal': [
[
{ 'n': 256, 't': 12, 'p': 0.25 },
{ 'n': 512, 't': 12, 'p': 0.25 },
{ 'n': 1024, 't': 12, 'p': 0.25 },
{ 'n': 1024+512, 't': 12, 'p': 0.25 },
],
[
{ 'n': 256, 't': 12, 'p': 0.5 },
{ 'n': 512, 't': 12, 'p': 0.5 },
{ 'n': 1024, 't': 12, 'p': 0.5 },
{ 'n': 1024+512, 't': 12, 'p': 0.5 },
],
[
{ 'n': 256, 't': 12, 'p': 0.75 },
{ 'n': 512, 't': 12, 'p': 0.75 },
{ 'n': 1024, 't': 12, 'p': 0.75 },
{ 'n': 1024+512, 't': 12, 'p': 0.75 },
],
[
{ 'n': 256, 't': 12, 'p': 1.0 },
{ 'n': 512, 't': 12, 'p': 1.0 },
{ 'n': 1024, 't': 12, 'p': 1.0 },
{ 'n': 1024+512, 't': 12, 'p': 1.0 },
]
],
'brief': [
[
{ 'n': 1024, 't': 12, 'p': 0.5 },
{ 'n': 2048, 't': 12, 'p': 0.5 },
]
],
'thread_scale': [
[
{ 'n': 1024, 't': 1, 'p': 0.5 },
{ 'n': 1024, 't': 2, 'p': 0.5 },
{ 'n': 1024, 't': 3, 'p': 0.5 },
{ 'n': 1024, 't': 4, 'p': 0.5 },
{ 'n': 1024, 't': 5, 'p': 0.5 },
{ 'n': 1024, 't': 6, 'p': 0.5 },
{ 'n': 1024, 't': 7, 'p': 0.5 },
{ 'n': 1024, 't': 8, 'p': 0.5 },
{ 'n': 1024, 't': 9, 'p': 0.5 },
{ 'n': 1024, 't': 10, 'p': 0.5 },
{ 'n': 1024, 't': 11, 'p': 0.5 },
{ 'n': 1024, 't': 12, 'p': 0.5 },
]
],
'serious': [
[
{ 'n': 1024, 't': 12, 'p': 0.5 },
{ 'n': 2048, 't': 12, 'p': 0.5 },
{ 'n': 4096, 't': 12, 'p': 0.5 },
{ 'n': 8192, 't': 12, 'p': 0.5 },
]
]
}
DEFAULT_BENCH = 'normal'
DEFAULT_BLOCK_SIZE = 8
parser = argparse.ArgumentParser()
parser.add_argument('-a', '--algorithm', choices=['f', 'j'], required=True,
help='Algorithm to benchmark')
parser.add_argument('-s', '--seed', default=42,
help='Seed for graph generation')
parser.add_argument('-d', '--block_size', default=DEFAULT_BLOCK_SIZE,
help='The block size of the graph for Floyd-Warshall')
parser.add_argument('-b', '--benchmark', choices=all_benchmarks.keys(), default=DEFAULT_BENCH,
help='The name of the benchmark to run')
parser.add_argument('-v', '--verbose', action='store_true',
help='Print commands as they run')
parser.add_argument('-g', '--cuda', action='store_true', help='Run CUDA version')
parser.add_argument('-r', '--compare', action='store_true', help='Compare different parallel schemes. Recommended to be used with "-b serious"')
args = parser.parse_args()
def create_cmd(params):
cmd = []
for attr, value in params.iteritems():
cmd += ['-' + attr, str(value)]
return cmd
def run_cmd(command, verbose):
if verbose:
print 'Running command ' + ' '.join(command)
p = subprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
return p.communicate()
def extract_time(stdout):
return float(re.search(r'(\d*\.?\d*)ms', stdout).group(1))
def run_bench(bench_list, algorithm, seed, block_size, verbose, cuda, caching_seq=True, seq_cache={}):
print ''
print ' {0:-^52} '.format('')
print '|{0:^52}|'.format(' Benchmark for {0}\'s Algorithm '
.format('Floyd-Warshall' if algorithm is 'f' else 'Johnson'))
print '|{0:^52}|'.format('seed = {0}{1}'.format(seed, ', block size = {0}'.format(block_size) if algorithm is 'f' else ''))
print ' {0:-^52} '.format('')
print '| {0:<4} | {1:<5} | {2:<2} | {3:<8} | {4:<8} | {5:<8} |'.format('p', 'n', 't', 'seq (ms)',
'par (ms)', 'speedup')
for bench in bench_list:
print ' {0:-^52} '.format('')
for param_obj in bench:
param_obj['a'] = algorithm
param_obj['s'] = seed
param_obj['d'] = block_size
params = create_cmd(param_obj)
cache_key = str(param_obj['p']) + str(param_obj['n'])
if not caching_seq or cache_key not in seq_cache:
stdout, stderr = run_cmd(['./apsp-seq'] + params, verbose)
if len(stderr):
print 'Sequential Error: ' + stderr
return
seq_cache[cache_key] = extract_time(stdout)
seq_time = seq_cache[cache_key]
if cuda: stdout, stderr = run_cmd(['./apsp-cuda'] + params,verbose)
else: stdout, stderr = run_cmd(['./apsp-omp'] + params, verbose)
if len(stderr):
print 'Parallel Error: ' + stderr
return
par_time = extract_time(stdout)
print '| {p:>4.2f} | {n:>5} | {t:>2} | {0:>8.1f} | {1:>8.1f} | {2:>7.1f}x |'.format(seq_time, par_time,
seq_time / par_time,
**param_obj)
print ' {0:-^52} '.format('')
print ''
def run_par_bench(bench_list, algorithm, seed, block_size, verbose, caching_seq=True, seq_cache={}):
print ''
print ' {0:-^54} '.format('')
print '|{0:^54}|'.format(' Benchmark for {0}\'s Algorithm '
.format('Floyd-Warshall' if algorithm is 'f' else 'Johnson'))
print '|{0:^54}|'.format('seed = {0}{1}'.format(seed, ', block size = {0}'.format(block_size) if algorithm is 'f' else ''))
print ' {0:-^54} '.format('')
print '| {0:<4} | {1:<5} | {2:<2} | {3:<8} | {4:<8} | {5:<8} |'.format('p', 'n', 't', 'ISPC (ms)',
'OMP (ms)', 'CUDA (ms)')
for bench in bench_list:
print ' {0:-^54} '.format('')
for param_obj in bench:
param_obj['a'] = algorithm
param_obj['s'] = seed
param_obj['d'] = block_size
params = create_cmd(param_obj)
stdout, stderr = run_cmd(['./apsp-omp'] + params, verbose)
if len(stderr):
print 'OMP Error: ' + stderr
return
omp_time = extract_time(stdout)
stdout, stderr = run_cmd(['./apsp-omp-ispc'] + params, verbose)
if len(stderr):
print 'OMP ISPC Error: ' + stderr
return
omp_ispc_time = extract_time(stdout)
stdout, stderr = run_cmd(['./apsp-cuda'] + params,verbose)
if len(stderr):
print 'CUDA Error: ' + stderr
return
cuda_time = extract_time(stdout)
print '| {p:>4.2f} | {n:>5} | {t:>2} | {0:>9.1f} | {1:>8.1f} | {2:>9.1f} |'.format(omp_ispc_time, omp_time,
cuda_time,
**param_obj)
print ' {0:-^52} '.format('')
print ''
def choose_benchmark():
if (args.compare):
run_par_bench(all_benchmarks[args.benchmark], args.algorithm, args.seed, args.block_size, args.verbose)
else:
run_bench(all_benchmarks[args.benchmark], args.algorithm, args.seed, args.block_size, args.verbose, args.cuda)
choose_benchmark()