-
Notifications
You must be signed in to change notification settings - Fork 1
/
otf2_metric_phase_parser.py
167 lines (148 loc) · 6.52 KB
/
otf2_metric_phase_parser.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
import otf2
from otf2.events import *
import numpy as np
import pandas as pd
def get_metric_events(trace_name):
with otf2.reader.open(trace_name) as trace:
metric_events = []
for metric_members in trace.definitions.metric_members:
metric_events.append(metric_members.name)
return metric_events
def open_trace(trace_name):
with otf2.reader.open(trace_name) as trace:
for location,event in trace.events:
yield event
def get_count_phase_num(trace_name,phase_name):
event = open_trace(trace_name)
count = 0
for i in event:
if isinstance(i, Enter):
if(i.region.name == phase_name):
count +=1
elif isinstance(i, Leave):
if(i.region.name == phase_name):
count += 1;
return count/2
def get_papi_values(trace_name, papi_events, num_phase_iter, num_processes):
event = open_trace(trace_name)
values_list = [0]*len(papi_events)
count = 0
temp_count = 0
# print(num_processes)
# print(num_phase_iter)
if(num_processes <= (num_phase_iter/num_processes)):
print("First loop is being executed")
for event_ in event:
if isinstance(event_, Metric):
if(len(event_.values) == len(papi_events)):
count +=1
if(count > num_phase_iter*2 - int(num_processes)):
temp_count += 1
# print(values_list)
for i in range(0, len(papi_events)):
values_list[i] += event_.values[i]
print(values_list)
else:
print("Second loop is being executed")
for event_ in event:
if isinstance(event_, Metric):
if(len(event_.values) == len(papi_events)):
print(event_)
count +=1
if(count > num_processes):
temp_count +=1
for i in range(0, len(papi_events)):
values_list[i] += event_.values[i]
print(values_list)
print(temp_count)
values_list = [i/num_processes for i in values_list]
temp_value = num_phase_iter/num_processes
values_list = [i/temp_value for i in values_list]
return values_list
def get_papi_values_w_time_stamps(time_stamps, papi_events, trace_name):
event = open_trace(trace_name)
values_list = [0]*len(papi_events)
count = 0
time_stamps.sort(key=int)
# print(len(time_stamps))
for event_ in event:
if isinstance(event_, Metric):
if(count <= len(time_stamps) - 1 and time_stamps[count] == event_.time):
count += 1
for i in range(0, len(papi_events)):
values_list[i] += event_.values[i]
# print(values_list)
values_list = [i/len(time_stamps) for i in values_list]
return values_list
def get_time_stamps(trace_name, phase_region, num_processes, num_phase_iter):
event = open_trace(trace_name)
time_stamps =[]
count = 0
for event_ in event:
if isinstance(event_, Enter) or isinstance(event_, Leave):
if (event_.region.name == phase_region):
count +=1
# print(count)
if(count > num_phase_iter*2 - int(num_processes)):
time_stamps.append(event_.time)
# print(time_stamps)
return time_stamps
def get_energy_values(trace_name, other_events):
event = open_trace(trace_name)
metric_values = np.zeros(len(other_events))
metric_events_counts = np.zeros(len(other_events))
time_list = []
for event_ in event:
if isinstance(event_, Metric):
for i in range(0, len(other_events)):
if(len(event_.values) == 1):
if(event_.metric.metric_class.members[0].name == other_events[i]):
metric_values[i] += event_.values[0]
metric_events_counts[i] += 1
time_list.append(event_.time)
for i in range(0, len(other_events)):
metric_values[i] /= metric_events_counts[i]
return metric_values, time_list
def read_trace(trace_name, phase_region, name, num_processes):
count = 0;
time_list = []
metric_events = get_metric_events(trace_name)
num_phase_iter = get_count_phase_num(trace_name, phase_region)
papi_events = [i for i in metric_events if "PAPI" in i]
other_events = [i for i in metric_events if i not in papi_events]
print(num_phase_iter/float(num_processes))
values_list_papi = get_papi_values(trace_name, papi_events, num_phase_iter, float(num_processes))
values_list_hdeem, time_list = get_energy_values(trace_name,other_events)
with otf2.reader.open(trace_name) as trace:
global_offset = trace.definitions.clock_properties.global_offset
resolution = trace.timer_resolution
time_list.sort(key=int)
time_end = time_list[len(time_list) -1]
time_start = time_list[0]
time_end = (time_end - global_offset)/resolution
time_start = (time_start - global_offset)/resolution
time = time_end - time_start
convert_2_csv(papi_events, other_events,values_list_papi,values_list_hdeem,name,time, num_phase_iter/float(num_processes))
def convert_2_csv(papi_events, other_events, papi_values, metric_values, name, time, num_phase_iter):
data = list(papi_values) + list(metric_values)
print(data)
columns = []
for i in range(0, len(papi_events)):
columns.append(papi_events[i])
for i in range(0, len(other_events)):
columns.append((other_events[i]))
data_dict = {columns[i]:data[i] for i in range(0, len(data))}
data_dict.update({'time':time})
data_dict.update({'Number of Phase Iterations':num_phase_iter})
print(data_dict)
df = pd.DataFrame(data = data_dict, index = [0])
df.to_csv(name + ".csv", sep='\t', header=True)
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser(description = 'This script parsers the metrics from an OTF2 trace file and converts it to a csv')
parser.add_argument("-i","--input", help="Trace file with path", required=True)
parser.add_argument("-p", "--phase_region", help="Name of phase region", required=True)
parser.add_argument("-n", "--name", help="Name of the output csv file", required=True)
parser.add_argument("-np","--num_processes", help="Number of MPI processes", required=True)
args = parser.parse_args()
read_trace(args.input, args.phase_region, args.name, args.num_processes)