-
Notifications
You must be signed in to change notification settings - Fork 0
/
benchmark.py
226 lines (163 loc) · 6.1 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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
import numpy as np
import xarray as xr
import matplotlib.pyplot as plt
import timeit
from helbigwindparam import downscale, laplacian_map, xsi
from helbigwindparam.slope_average import mu_average
from helbigwindparam.subgrid import std_moving_window_map, xsi_numpy, xsi_numexpr, x_sgp_numpy, x_sgp_numexpr
from helbigwindparam.slope import slope_mu_map
from helbigwindparam.downscale import x_dsc_numpy, x_dsc_numexpr
path_to_file = "/home/letoumelinl/bias_correction/Data/1_Raw/DEM/DEM_ALPES_L93_30m.nc"
test = xr.open_dataset(path_to_file)
launch_computation = False
"""
Benchmark
24/10/2022
30_000 x 30_000 px (30 m grid spacing)
No crash observed.
Tensorflow: 13.01s
Numba (run 0): 39.32s
Numpy: 39.84s
Numba (run 1): 23.12s
"""
if launch_computation:
alti = test.alti.values[:30_000, :30_000]
laplacian_map(alti, 30, helbig=True, library_laplacian="tensorflow")
laplacian_map(alti, 30, helbig=True, library_laplacian="numba")
laplacian_map(alti, 30, helbig=True, library_laplacian="numpy")
laplacian_map(alti, 30, helbig=True, library_laplacian="numba")
"""
Benchmark
24/10/2022
3500 x 3500 px (30 m grid spacing)
Code crashes for topographies larger than 3500px by 3500px
Tensorflow: 8.53s
Numba (run 0): 192.53s
Numpy (list comprehension): 164.14s
Numba (run 1): 193s
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
mu_avg_tf = mu_average(alti, 30, x_win=69//2, y_win=79//2, library_mu_avg="tensorflow")
_ = mu_average(alti, 30, x_win=69//2, y_win=79//2, library_mu_avg="numba")
_ = mu_average(alti, 30, x_win=69//2, y_win=79//2, library_mu_avg="numpy")
_ = mu_average(alti, 30, x_win=69//2, y_win=79//2, library_mu_avg="numba")
"""
Benchmark
24/10/2022
3500 x 3500 px (30 m grid spacing)
Code crashes for topographies larger than 3500px by 3500px
Tensorflow: 21.11s
Numba (run 0): 354.8s
Numpy (list comprehension): 352.45s
Numba (run 1): 346.72s
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
std_avg_tf = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="tensorflow")
std_avg_nb = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="numba")
_ = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="numpy")
_ = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="numba")
plt.figure()
plt.imshow(std_avg_tf)
plt.figure()
plt.imshow(std_avg_nb)
"""
Benchmark
24/10/2022
3500 x 3500 px (30 m grid spacing)
There seems to be a scale effect: large arrays are faster with numexpr than with numpy.
Numpy seems to be better for small arrays.
Numpy: 1.30s (100 iterations)
Numexpr: 2.40s (100 iterations)
For an array of shape 40_000px x 40_000px
Numpy: 6.7s
Numexpr: 2.79s
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
mu_avg_tf = mu_average(alti, 30, x_win=69 // 2, y_win=79 // 2, library_mu_avg="tensorflow")
std_avg_tf = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="tensorflow")
timeit.timeit('xsi_numpy(std_avg_tf, mu_avg_tf)', number=100, globals=globals())
timeit.timeit('xsi_numexpr(std_avg_tf, mu_avg_tf)', number=100, globals=globals())
"""
Benchmark
24/10/2022
3500 x 3500 px (30 m grid spacing)
There seems to be a scale effect: large arrays are faster with numexpr than with numpy.
Numpy seems to be better for small arrays.
Numpy: 1.30s (100 iterations)
Numexpr: 2.40s (100 iterations)
For an array of shape 40_000px x 40_000px
Numpy: 6.7s
Numexpr: 2.79s
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
mu_avg_tf = mu_average(alti, 30, x_win=69 // 2, y_win=79 // 2, library_mu_avg="tensorflow")
std_avg_tf = std_moving_window_map(alti, x_win=69 // 2, y_win=79 // 2, library_std_avg="tensorflow")
timeit.timeit('xsi_numpy(std_avg_tf, mu_avg_tf)', number=100, globals=globals())
timeit.timeit('xsi_numexpr(std_avg_tf, mu_avg_tf)', number=100, globals=globals())
test = np.ones((40_000, 40_000))
xsi_numpy(test, test)
xsi_numexpr(test, test)
"""
Benchmark
24/10/2022
3500 x 3500 px (30 m grid spacing)
There seems to be a scale effect: large arrays are faster with numexpr than with numpy.
Numpy seems to be better for small arrays.
Numpy: 0.67s
Numexpr: 0.14s
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
mu_avg_tf = mu_average(alti, 30, x_win=69 // 2, y_win=79 // 2, library_mu_avg="tensorflow")
xsi_ = xsi(alti, mu_avg_tf, x_win=69//2, y_win=79//2, library_std_avg="tensorflow")
a = 3.354688
b = 1.998767
c = 0.20286
d = 5.951
l = 2_000
x_sgp_numpy(mu_avg_tf, xsi_, l, a, b, c, d)
x_sgp_numexpr(mu_avg_tf, xsi_, l, a, b, c, d)
"""
Benchmark
24/10/2022
Domain: full DEM
Numpy: 20.52s
Numexpr: 5.05s
"""
if launch_computation:
alti = test.alti.values
laplacian = laplacian_map(alti, 30, helbig=True)
mu = slope_mu_map(alti, 30)
a = 17.0393
b = 0.737
c = 1.0234
d = 0.3794
e = 1.9821
x_dsc_numpy(laplacian, mu, a, b, c, d, e)
x_dsc_numexpr(laplacian, mu, a, b, c, d, e)
"""
Benchmark
24/10/2022
Domain: 3500 x 3500px
Optimized configuration: 31.04s
config["library_dsc"] = "numexpr" # "numpy" or "numexpr"
config["library_std_avg"] = "tensorflow" # "tensorflow", "numpy", "numba"
config["library_mu_avg"] = "tensorflow" # "tensorflow", "numpy", "numba"
config["library_xsi"] = "numexpr" # "numpy", "numexpr"
config["library_laplacian"] = "tensorflow" # "numba", "numpy"
config["library_x_sgp"] = "numexpr" # "numpy", "numexpr"
Numpy configuration: 524.15s
config["library_dsc"] = "numpy" # "numpy" or "numexpr"
config["library_std_avg"] = "numpy" # "tensorflow", "numpy", "numba"
config["library_mu_avg"] = "numpy" # "tensorflow", "numpy", "numba"
config["library_xsi"] = "numpy" # "numpy", "numexpr"
config["library_laplacian"] = "numpy" # "numba", "numpy"
config["library_x_sgp"] = "numpy" # "numpy", "numexpr"
"""
if launch_computation:
alti = test.alti.values[10_000:13_500, 10_000:13_500]
downscale(alti, dx=30, l=2_000, x_win=69//2, y_win=79//2)