【Codon】 When a process that consumes a large amount of memory is executed, it is killed and the process is stopped.

[icst005]

I was not sure if there was a report applicable to the same case, so I am reporting it just in case.

When a process that consumes a large amount of memory (test04.py) is executed, it is killed and the process stops.
When the same code is executed in Python, the process completes normally.

[Execution Result]
-test04.py Execution result with Codon
ic@DESKTOP-C83DAGF:/mnt/c/Users/ic$ codon run -release /mnt/z/koko/ps3/cd/test04_codon.py

kaishi
test04_codon
from python import numpy as np
Killed
ic@DESKTOP-C83DAGF:/mnt/c/Users/ic$

-test04.py Execution result with Python

kaishi
test04_codon
from python import numpy as np
0
1
2
3

　・・・

9997
9998
2367.8824689388275
test04_python
import numpy as np
owari

[Execution Procedure]
Execute the following from Windows PowerShell Terminal on Windows in the following order.

wsl -d Ubuntu-22.04
codon run -release /mnt/z/koko/ps3/cd/test04.py
（Python, with Anaconda, running on JupyterNotebook）

[Execution environment]
Codon Ver: 0.18
Python Ver: 3.10.8
Windows 10 Pro 22H2(WSL Ubuntu 22.04)
Windows Feature Experience Pack 1000.19055.1000.0
Core(TM) i7-7700HQ
Memory 64GB

[Sampl Code]
-test04.py
from python import numpy as np #------------①
from python import pandas as pd
from python import csv
import time
from python import random

print("kaishi")
print("test04_codon")
print("from python import numpy as np")

t00 = time.time()

for aa0 in range(9999):

print(aa0)

m0 = 9000000
np1 = np.zeros((m0, 1), dtype=np.int64)

np.random.seed()
np1[0: m0, 0] = np.random.randint(0, 99999, m0)

np1 = None

t01 = time.time()
t02 = t01 - t00

print(t02)
print("test04_codon")
print("from python import numpy as np")
print("owari")

[ghuls]

I assume this is because the python objects still exsit and each time your assign a numpy expression to np1, a different python object gets bound to np1 (which is only a variable in codon and not in python).

Putting this code in a python function, does not blow up memory.

from python import numpy as np #------------①
from python import pandas as pd
from python import csv
import time
from python import random

@python
def random_int(aa0):
    import numpy as np
    print(aa0)

    m0 = 9000000
    np1 = np.zeros((m0, 1), dtype=np.int64)

    np.random.seed()
    np1[0: m0, 0] = np.random.randint(0, 99999, m0)

    np1 = None

print("kaishi")
print("test04_codon")
print("from python import numpy as np")

t00 = time.time()

for aa0 in range(9999):
    random_int(aa0)

t01 = time.time()
t02 = t01 - t00

print(t02)
print("test04_codon")
print("from python import numpy as np")

On a side note, calling del on the object seems not supported:

from python import numpy as np #------------①
from python import pandas as pd
from python import csv
import time
from python import random

print("kaishi")
print("test04_codon")
print("from python import numpy as np")

t00 = time.time()

for aa0 in range(9999):
    print(aa0)

    m0 = 9000000
    np1 = np.zeros((m0, 1), dtype=np.int64)

    np.random.seed()
    np1[0: m0, 0] = np.random.randint(0, 99999, m0)

    del np1
    #np1 = None

t01 = time.time()
t02 = t01 - t00

print(t02)
print("test04_codon")
print("from python import numpy as np")
print("owari")

$ codon run --release test_memory_codon.py
test04.py:22:5-12: error: 'pyobj' object has no method '__init__' with arguments (pyobj)
Command exited with non-zero status 1