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
|
import os
import string
import numpy as np
import matplotlib.pyplot as plt
def parse_file(filename):
timeslist = []
meanlist = []
stddevlist = []
with open(filename) as f:
for line in f:
linelist = line.split()
if linelist[0].split('/')[-1]=='manual_time':
timeslist.append(int(linelist[2]))
if linelist[0].split('/')[-1]=='manual_time_mean':
meanlist.append(int(linelist[2]))
if linelist[0].split('/')[-1]=='manual_time_stddev':
stddevlist.append(int(linelist[2]))
return np.array(timeslist), np.array(meanlist), np.array(stddevlist)
def prep_arrays(serial_times, parallel_times, index_s, index_p, N):
times = [x for x in parallel_times[index_p::N]]
times.insert(0, serial_times[index_s])
# print(times)
return np.array(times)
if __name__ == "__main__":
import sys
# Input arguments:
# File names
serial_times, serial_means, serial_stddevs = parse_file(sys.argv[1])
parallel_times, parallel_means, parallel_stddevs = parse_file(sys.argv[2])
# benchmarking parameters
n1 = int(sys.argv[3]) # number of image sizes
n01 = int(sys.argv[4]) # index of first image size
n2 = int(sys.argv[5]) # number of visibilities sizes
n02 = int(sys.argv[6]) # index of first visibility size
n3 = int(sys.argv[7]) # number of kernel sizes
n03 = int(sys.argv[8]) # index of first kernel size
# number of benchmarked functions (eg. [ctor, direct, adjoint]=3)
nb = int(sys.argv[9])
# number of benchmarked algorithms (eg. for the measurement operator, [distr, MPI]=2)
na = int(sys.argv[10])
# number of benchmarked ncore configurations, serial included
nconf = int(sys.argv[11])
ylabel = "Time (ms or us)"
# Parameter lists etc
im_sizes = [128, 256, 512, 1024, 2048, 4096]
vis_sizes = ['1M', '10M']
k_sizes = [2, 4, 8, 16]
x = range(nconf)
x_name = ['serial',
'1',
'2',
'3',
'4',
'8',
'12'
]
colors = ['black', 'red', 'green', 'blue']
markers = ['o', '^', 's', 'x']
bm_name = ['Constructor', 'Apply Direct', 'Apply Adjoint']
al_name = ['Distributed', 'MPI']
npar = 0
par_name = []
for im in range(n01,n01+n1):
for vis in range(n02,n02+n2):
for k in range(n03,n03+n3):
if k_sizes[k]*k_sizes[k]<im_sizes[im]:
npar = npar+1
par_name.append('im='+str(im_sizes[im])+', vis='+vis_sizes[vis]+', k='+str(k_sizes[k]))
N = npar*nb
for bm in range(0,nb):
fig = plt.figure()
fig.suptitle(bm_name[bm], fontsize=16)
for al in range(1,na+1):
print(al,na)
ax = fig.add_subplot(1,na,al)
for par in range(0,npar):
index = bm*npar + (al-1)*N + par
y = prep_arrays(serial_means, parallel_means, index-(al-1)*N, index, na*N)
err = prep_arrays(serial_stddevs, parallel_stddevs, index-(al-1)*N, index, na*N)
print(x, y, err)
group = par_name[par]
ax.errorbar(x, y, yerr=err, c=colors[par%4], fmt='o', label=group)
plt.title(al_name[al-1])
plt.setp(ax.get_xticklabels(), visible=True)
plt.xticks(range(nconf), x_name)
plt.xlabel("# nodes (with 16 threads per node)")
plt.ylabel(ylabel)
plt.legend()
plt.show()
|
