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
|
#!/usr/bin/env python3
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
import gc
import io
import json
import os
import sys
import time
from timeit import timeit
import numpy as np
import psutil
from memory_profiler import memory_usage
from tabulate import tabulate
import orjson
os.sched_setaffinity(os.getpid(), {0, 1})
kind = sys.argv[1] if len(sys.argv) >= 1 else ""
if kind == "float16":
dtype = np.float16
array = np.random.random(size=(50000, 100)).astype(dtype)
elif kind == "float32":
dtype = np.float32
array = np.random.random(size=(50000, 100)).astype(dtype)
elif kind == "float64":
dtype = np.float64
array = np.random.random(size=(50000, 100))
assert array.dtype == np.float64
elif kind == "bool":
dtype = np.bool_
array = np.random.choice((True, False), size=(100000, 200))
elif kind == "int8":
dtype = np.int8
array = np.random.randint(((2**7) - 1), size=(100000, 100), dtype=dtype)
elif kind == "int16":
dtype = np.int16
array = np.random.randint(((2**15) - 1), size=(100000, 100), dtype=dtype)
elif kind == "int32":
dtype = np.int32
array = np.random.randint(((2**31) - 1), size=(100000, 100), dtype=dtype)
elif kind == "uint8":
dtype = np.uint8
array = np.random.randint(((2**8) - 1), size=(100000, 100), dtype=dtype)
elif kind == "uint16":
dtype = np.uint16
array = np.random.randint(((2**16) - 1), size=(100000, 100), dtype=dtype)
elif kind == "uint32":
dtype = np.uint32
array = np.random.randint(((2**31) - 1), size=(100000, 100), dtype=dtype)
else:
print(
"usage: pynumpy (bool|int16|int32|float16|float32|float64|int8|uint8|uint16|uint32)",
)
sys.exit(1)
proc = psutil.Process()
def default(__obj):
if isinstance(__obj, np.ndarray):
return __obj.tolist()
raise TypeError
headers = ("Library", "Latency (ms)", "RSS diff (MiB)", "vs. orjson")
LIBRARIES = ("orjson", "json")
ITERATIONS = 10
def orjson_dumps():
return orjson.dumps(array, option=orjson.OPT_SERIALIZE_NUMPY)
def json_dumps():
return json.dumps(array, default=default).encode("utf-8")
output_in_mib = len(orjson_dumps()) / 1024 / 1024
print(f"{output_in_mib:,.1f}MiB {kind} output (orjson)")
gc.collect()
mem_before = proc.memory_full_info().rss / 1024 / 1024
def per_iter_latency(val):
if val is None:
return None
return (val * 1000) / ITERATIONS
def test_correctness(func):
return np.array_equal(array, np.array(orjson.loads(func()), dtype=dtype))
table = []
for lib_name in LIBRARIES:
gc.collect()
print(f"{lib_name}...")
func = locals()[f"{lib_name}_dumps"]
if func is None:
total_latency = None
latency = None
mem = None
correct = False
else:
total_latency = timeit(
func,
number=ITERATIONS,
)
latency = per_iter_latency(total_latency)
time.sleep(1)
mem = max(memory_usage((func,), interval=0.001, timeout=latency * 2))
correct = test_correctness(func)
if lib_name == "orjson":
compared_to_orjson = 1
orjson_latency = latency
elif latency:
compared_to_orjson = latency / orjson_latency
else:
compared_to_orjson = None
if not correct:
latency = None
mem = 0
mem_diff = mem - mem_before
table.append(
(
lib_name,
f"{latency:,.0f}" if latency else "",
f"{mem_diff:,.0f}" if mem else "",
f"{compared_to_orjson:,.1f}" if (latency and compared_to_orjson) else "",
),
)
buf = io.StringIO()
buf.write(tabulate(table, headers, tablefmt="github"))
buf.write("\n")
print(buf.getvalue())
|
