Memory usage¶
Base sizes for each event¶
Note: values below are from a 64-bit system using Python 3.7
Every callback event is a tuple of 6 entries. A base tuple object is 56 byte.
Object references in Python are 8 byte each, so each tuple is 56 + 6 * 8 = 104 byte in total.
The objects that each of the entries refers to are the following:
time : integer
priority : integer
event_id : integer
callback : method or function
args : tuple
kwargs : dict
Integers are variable in size. For 0, the size is 24 byte. For relatively small numbers (e.g., 1) the size is 28 byte. Python maintains a pool to prevent excessive allocation of commonly used immutable objects.
An empty tuple is 56 byte and a N-item tuples is 56 + N * 8 byte.
Some preliminary measurements indicate that a function is 144 byte and a method is 72 byte.
An empty or small number of key-values dict is 248 byte. The more key-values are added, the larger it becomes.
Example case analysis¶
Below is an analysis of the memory size it allocates for every event assuming the callback has 2 positional integer arguments:
Origin Size (byte)
main tuple 56 + 6 * 8 (always)
time (int) 28 (assuming time is not in integer pool)
priority (int) 0 (assuming priority is 0, so in the integer pool)
event_id (int) 28 (assuming event_id is not in integer pool)
callback 0 (function or method is generally re-used)
args (tuple) 56 + 2 * 8 (two positional arguments)
args[0] (int) 28 (assuming not in integer pool)
args[1] (int) 28 (assuming not in integer pool)
kwargs (dict) 248 (always (minimum))
---
Total: 536
It can be verified with the following test script (requires python3 -m pip install guppy3):
import unittest
import random
from guppy import hpy
from discrevpy import simulator
class TestMemory(unittest.TestCase):
def test_memory_usage(self):
def x(a, b):
pass
simulator.ready()
random.seed(29106970)
for i in range(1000000):
v1 = random.randint(0, 1000000000)
v2 = random.randint(0, 1000000000)
t = random.randint(0, 1000000000)
simulator.schedule(t, x, v1, v2)
print(hpy().heap())
simulator.run()
simulator.reset()
… and then running it using: python3 -m pytest -s -k test_memory_usage