How to limit CPU usage to a single core with freud.Box.wrap?

[zhangyemayapro]

Description

I am using freud with the following code:

import freud
import numpy as np
import threading
import time

fr_box = freud.Box.from_box([10, 10, 0])
a = np.random.random((100, 3))

while True:
    fr_box.wrap(a)
    if int(time.time()) % 2 == 0:
        print(f"Active threads: {threading.active_count()}")


The output of the code repeatedly shows the following:


Active threads: 8
Active threads: 8
Active threads: 8
Active threads: 8
Active threads: 8

Whe I run the program in the foreground, the program consumes 100% CPU across all available cores, even though I expect it to run on a single core.
![image](https://github.com/user-attachments/assets/b1bd9fa3-b355-479a-b626-609ad62ece36)


## I have tried:
```python 

import freud
freud.parallel.set_num_threads(1)



## version
freud  3.1.0   


Thanks for your help~

[joaander]

I am not able to reproduce the behavior you describe.

1. According to https://docs.python.org/3/library/threading.html#threading.active_count, threading.active_count should

   Return the number of Thread objects currently alive.

   freud uses TBB at the C++ level, not Thread objects. When I run your unmodified script, I see Active threads: 1 as expected -- whether I call set_num_threads or not. Some other import must be creating active Thread objects.

2. set_num_threads is working as expected for me. Running your script without the set_num_threads call, I run ps in another terminal and get:

   $ ps -p $(pgrep python) -o %cpu
   %CPU
   880.7
   

When I run the the following script:

import freud
import numpy as np
import threading
import time

freud.parallel.set_num_threads(1)

fr_box = freud.Box.from_box([10, 10, 0])
a = np.random.random((100, 3))

while True:
    fr_box.wrap(a)
    if int(time.time()) % 2 == 0:
        print(f"Active threads: {threading.active_count()}")

I get the expected 100% CPU usage:

$ ps -p $(pgrep python) -o %cpu
%CPU
99.1

Your image doesn't show which processes are contributing to that CPU usage. It is possible that another program is running in the background consuming the other 7 CPUs.

[zhangyemayapro]

Thanks for your help.🤩
I find this issue can be reproduce by the following process:
Initially, 37% of the total CPUs are occupied:

After I run with freud.parallel.set_num_threads(1)

Things goes well.

But，if without freud.parallel.set_num_threads(1), all cpus are occupied.

If i add freud.parallel.set_num_threads(1) without restarting jupyter, all the cpus are still occupied.

Most amazingly, even after completing the jupyter cell,all the cpus are still occupied. Unless I kill the jupyter Notebook.

[joaander]

Your images still don't show which process is burning CPU usage. It highly likely that this is the Jupyter process that drives the web server. It could also be Jupyter code running in the background of your kernel process (that could be the source of your 8 active threads). These may be I/O threads or processes delivering the massive number of print() outputs to the server.

Please try running this script on the command line where you will observe the results I reported.

[zhangyemayapro]

Thanks.
I run the code on the terminnal with "python test.py"

import freud
import numpy as np
import threading
import time



fr_box = freud.Box.from_box([10, 10, 0])
a = np.random.random((100, 3))
s=time.time()


print("start 1")
while time.time()-s<10:
    fr_box.wrap(a)
print("end 1")

freud.parallel.set_num_threads(1)

while True:
    fr_box.wrap(a)
    
    
print("end 2")

It shows:

[joaander]

You failed to run the exact script I provided, so you should not expect the same results. freud.parallel.set_num_threads(1) must be the first function call that you make to the freud module.

[zhangyemayapro]

Thanks for your help👍😄🥳