forked from PyCOMPLETE/CoupledBunchAnalysis
-
Notifications
You must be signed in to change notification settings - Fork 0
/
011_speedup_from_pyparislog.py
201 lines (161 loc) · 6.59 KB
/
011_speedup_from_pyparislog.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
import time
import numpy as np
import parse_pyparislog as ppl
import matplotlib.pyplot as plt
import mystyle as ms
na = np.array
groups = [
{
'list_folders' : [
'../test2_on_HPC_1slot_1cores/004_multibunch_with_ecloud',
'../test2_on_HPC_1slot_8cores/004_multibunch_with_ecloud',
'../test2_on_HPC_1slot/004_multibunch_with_ecloud',
'../test2_on_HPC_2slot/004_multibunch_with_ecloud',
#'../test2_on_HPC_4slot_240cores/004_multibunch_with_ecloud',
'../test2_on_HPC_4slot/004_multibunch_with_ecloud',
'../test2_on_HPC_8slot/004_multibunch_with_ecloud',
], 'fact_HT' : 1., 'b_spac_ns':20, 'tag':'Hyper threading OFF',
'plt_vs_nreal':True,
'comment': "Run on BE-short, scan in bunch slot size, HT OFF, no glob synch"
},
{
'list_folders' : [
'../test4_on_HPC_1slot_1cores/004_multibunch_with_ecloud',
'../test4_on_HPC_1slot_8cores/004_multibunch_with_ecloud',
'../test4_on_HPC_1slot/004_multibunch_with_ecloud',
'../test4_on_HPC_2slot/004_multibunch_with_ecloud',
'../test4_on_HPC_4slot/004_multibunch_with_ecloud',
'../test4_on_HPC_8slot/004_multibunch_with_ecloud',
], 'fact_HT' : 2., 'b_spac_ns':20, 'tag':'Hyper threading ON',
'plt_vs_nreal':False,
'comment': "Run on BE-short, scan in bunch slot size, HT ON"
},
{
'list_folders' : [
'../test6_on_HPC_1slot_1cores/004_multibunch_with_ecloud',
'../test6_on_HPC_1slot_8cores/004_multibunch_with_ecloud',
'../test6_on_HPC_1slot/004_multibunch_with_ecloud',
'../test6_on_HPC_2slot/004_multibunch_with_ecloud',
'../test6_on_HPC_4slot/004_multibunch_with_ecloud',
'../test6_on_HPC_8slot/004_multibunch_with_ecloud',
], 'fact_HT' : 2., 'b_spac_ns':20, 'tag':'HT ON, no glob synch',
'plt_vs_nreal':False,
'comment': "Run on BE-short, scan in bunch slot size, HT OFF, no glob synch"
},
]
# groups = [
# {
# 'list_folders' : [
# '../test5_on_HPC_8bunches/004_multibunch_with_ecloud/',
# '../test5_on_HPC_16bunches/004_multibunch_with_ecloud/',
# '../test5_on_HPC_32bunches/004_multibunch_with_ecloud/',
# '../test5_on_HPC_64bunches/004_multibunch_with_ecloud/',
# ],'fact_HT' :1., 'b_spac_ns':20., 'tag':'nbun_scan', 'plt_vs_nreal':False,
# },
# ]
# groups = [
# {
# 'list_folders' : [
# '../test9bis_on_HPC_25ns_correct_be_long/004_multibunch_with_ecloud/',
# '../test10_onHPC_144b/004_multibunch_with_ecloud/',
# '../test12_onHPC_288b/004_multibunch_with_ecloud/',
# ],'fact_HT' :1., 'b_spac_ns':25., 'tag':'nbun_scan', 'plt_vs_nreal':False,
# },
# ]
mode_timeplot = 'loglog'
plt.close('all')
ms.mystyle_arial(fontsz=16, dist_tick_lab=5)
fig1 = plt.figure(1, figsize=(8,6*1.3))
fig1.set_facecolor('white')
ax1 = fig1.add_subplot(2,1,1)
ax2 = fig1.add_subplot(2,1,2, sharex=ax1)
fig10 = plt.figure(10)
fig10.set_facecolor('white')
ax10 = fig10.add_subplot(1,1,1)
fig20 = plt.figure(20)
fig20.set_facecolor('white')
ax20 = fig20.add_subplot(1,1,1)
fig30 = plt.figure(30, figsize=(8,6*1.3))
fig30.set_facecolor('white')
ax30 = fig30.add_subplot(2,1,1)
ax31 = fig30.add_subplot(2,1,2, sharex=ax30)
for gg in groups:
list_folders = gg['list_folders']
fact_HT = gg['fact_HT']
tag = gg['tag']
#n_bunches = gg['n_bunches']
b_spac_ns = gg['b_spac_ns']
plt_vs_nreal = gg['plt_vs_nreal']
n_cores_list = []
avgt_turn_steps_list = []
n_slots_list = []
n_bunches_list = []
for sim_folder in list_folders:
dict_config, ibun_arr, t_arr, iturn_arr, iter_turn_steps, \
iturn_steps, tturn_steps, n_turns_steps, avgt_turn_steps = ppl.parse_pyparislog(sim_folder+'/pyparislog.txt')
# identify slot size
with open(sim_folder + '/Simulation.py', 'r') as fid:
lns = fid.readlines()
found = True
for ln in lns:
if ln.startswith('b_spac_s'):
slot_size_s = eval(ln.split('=')[-1])
n_slots = np.max(ibun_arr)+1
n_bun = n_slots*slot_size_s/(b_spac_ns*1e-9)
n_cores_list.append(dict_config['N_cores'])
avgt_turn_steps_list.append(avgt_turn_steps[1])
n_slots_list.append(n_slots)
n_bunches_list.append(n_bun)
n_cores_list = np.array(n_cores_list)
if plt_vs_nreal:
n_cores_list[1:] = n_cores_list[1:]/fact_HT
# # I got convinced that these are not needed
# if fact_HT>1 and plt_vs_nreal:
# # I don't manage to simulate single core with hyperthreading
# avgt_turn_steps_list[0]*=2
speedup = 1/(np.array(avgt_turn_steps_list)/avgt_turn_steps_list[0])
gg['avgt_turn'] = na(avgt_turn_steps_list)
ax1.plot(n_cores_list, speedup, '.-', label=tag, linewidth=2, markersize=10)
ax1.set_ylabel('Speed-up')
ax1.grid(True)
ax2.plot(n_cores_list, na(n_slots_list)/na(n_bunches_list),'.-', label=tag, linewidth=2, markersize=10)
ax2.grid(True)
ax2.set_ylabel('N slots per bunch passage')
ax10.plot(n_cores_list, np.array(avgt_turn_steps_list)/3600, '.-', label=tag, linewidth=2, markersize=10)
ax10.set_ylabel('Time per turn [h]')
ax10.grid(True)
ax20.plot(n_cores_list, 1000*np.array(avgt_turn_steps_list)/3600/24, '.-', label=tag, linewidth=2, markersize=10)
ax20.set_ylabel('Time per 1000 turns [days]')
ax20.grid(True)
ax30.plot(n_bunches_list, 1000*np.array(avgt_turn_steps_list)/3600/24, '.-', label=tag, linewidth=2, markersize=10)
ax31.plot(n_bunches_list, n_cores_list, '.-', label=tag, linewidth=2, markersize=10)
ax1.plot([1,max(n_cores_list)], [1,max(n_cores_list)], 'k')
for ax in [ax2, ax10, ax20]:
ax.set_xlabel('N. CPU cores')
ax1.legend(loc='lower right', prop={'size':16}).draggable()
ax10.legend(loc='upper right', prop={'size':16}).draggable()
ax20.legend(loc='upper right', prop={'size':16}).draggable()
ax30.set_ylim(bottom=0.)
ax30.set_ylabel('Time per 1000 turns [days]')
ax31.set_xlabel('N. bunches')
ax31.set_ylabel('N. CPU cores')
ax31.set_xlim(left=0.)
ax31.set_ylim(bottom=0.)
ax30.grid(True)
ax31.grid(True)
if mode_timeplot == 'loglog':
ax20.set_xscale("log")
ax20.set_yscale("log")
from matplotlib.ticker import ScalarFormatter
for axis in [ax20.xaxis, ax20.yaxis]:
axis.set_major_formatter(ScalarFormatter())
if len(groups)>1:
figcomp = plt.figure(100)
axcomp = figcomp.add_subplot(1,1,1)
figcomp.set_facecolor('w')
axcomp.plot(n_cores_list[1:], (groups[1]['avgt_turn']/groups[0]['avgt_turn'])[1:], '.-', linewidth=2, markersize=10)
axcomp.grid(True)
axcomp.set_ylim(1, 2)
axcomp.set_xlabel('N. CPU cores')
axcomp.set_ylabel('Performance ratio')
plt.show()