About Python's multithreading and GIL

Overview

RELEASE THE GIL

===========================================================================

Python is a great language 😃 , but the GIL (Global Interpreter Lock) is a huge bottleneck 😢 . It's a lock that is held by the interpreter while executing any Python code. This means that if you have a bunch of Python threads running, they all will have to wait for the GIL to be released. This can be a problem, because the GIL is held for a long time, which can lead to many performance issues.

The GIL is released whenever a thread is ready to run, and it's held back until all threads are ready to run. This means that if you have a bunch of Python threads running, they'll all be waiting for the GIL to be released but they'll all be blocked still after release. So, even though the threads will run concurrently, only one thread will be allowed to run at the exact same time.

As you see, there are no way to get rid of the GIL from Python Interpreters.

You may ask, why Python intepreter has this GIL?

Ans: The main Python implementation(CPython) is written in C, and many C codes are not thread safe. Everything are objects in Python, garbage collection are done using reference counting. The big problem is if one thread(let's say first) that is tracing the reference count of an object, if any other thread(let's say second) deletes that same object, the first thread will have no idea that same object is deleted by other thread. That's because reference counting is always done by per thread basis. And also, there may be some situation if one thread may wait for some other threads which already finished its work, that's called deadlock. If all threads are accessing the same object and changing at the same time, it may lead to race condtions. To prevent deadlock or race conditions, there's some locks needed per thread basis which can protect their resources.

Note that, in Java or C#, garbage collection is done by tracing garbage collection, which are aware always of the threads, and provides atomicity.

What is the benefit of the GIL?

Ans: The following reasons are:

  1. It's easy to write C extensions in Python and provides better C API rather than many popular languages like Java or C#.
  2. GIL makes single threaded programs faster.
  3. CPython Interpreter loading time is faster than JVM loading time because of GIL.
  4. Developers don't have to worry about thread-safety, because GIL always provides thread safety.

But, the big question is: How can we get rid of the GIL for getting performance?

Ans: As a Python coder, you should always be aware of the performance you're getting from code, and the GIL. Although, there are multiple ways we can get rid of this. Those are:

  1. Using numpy arrays for fast computation or some other good libraries which provides C API.
  2. Using Cython or Numba, although Cython is preferred.
  3. Using OpenMP API for Python.
  4. Using Pybind11 if you are good enough to write C++ code.
  5. If you're not satisfied with all I've mentioned, you can write raw C code provided Python C API.

In this repo, you will see how to release the GIL using Cython.

Note that, Python functions are made in such a way, that those can release the GIL when it's I/O bound. Such as:

  1. While reading from a file, or writing to a file.
  2. While reading from a socket, or writing to a socket.
  3. While reading from a database, or writing to a database.
  4. While getting data from internet, or sending to a intenet.
  5. While sleeping for some time.

In those cases, for I/O bound tasks you can use threading module. And, for long running I/O bound tasks, it is recommended to use async/await(coroutines) syntax for better performance.

If don't need threads to run in parallel, you can use multiprocessing module. Running tasks in separate processes won't be any problem for the GIL because every process has their own interpreter, which are again locked by the GIL.

Which Python interpreter don't have the GIL?

Ans: Jython and IronPython don't have the GIL because they are implemented to run in JVM and .NET environment respectively. PyPy has the GIL but PyPy-STM(Software Transactional Memory) doesn't have the GIL.

References

You can know more about this topic which I mentioned below:

  1. Python's Infamous GIL - Larry Hastings
  2. The Gilectomy project - Larry Hastings
  3. Inside the Python GIL - David Beazley

Happy Coding with Python! 😊

Owner
Souvik Ghosh
A self-taught programmer 👨‍💻
Souvik Ghosh
Discover and load entry points from installed packages

Entry points are a way for Python packages to advertise objects with some common interface. The most common examples are console_scripts entry points,

Thomas Kluyver 69 Jul 05, 2022
This project recreates the R-based RCy3 Cytoscape Automation library as a Python package.

Python library for calling Cytoscape Automation via CyREST

Cytoscape Consortium 40 Dec 22, 2022
Writeup and scripts for the 2021 malwarebytes crackme

Malwarebytes Crackme 2021 Tools and environment setup We will be doing this analysis in a Windows 10 VM with the flare-vm tools installed. Most of the

Jerome Leow 9 Dec 02, 2022
Python wrapper around Apple App Store Api

App Store Connect Api This is a Python wrapper around the Apple App Store Api : https://developer.apple.com/documentation/appstoreconnectapi So far, i

123 Jan 06, 2023
Shared utility scripts for AI for Earth projects and team members

Overview Shared utilities developed by the Microsoft AI for Earth team The general convention in this repo is that users who want to consume these uti

Microsoft 38 Dec 30, 2022
51AC8 is a stack based golfing / esolang that I am trying to make.

51AC8 is a stack based golfing / esolang that I am trying to make.

7 May 22, 2022
ARK sõidueksami Matrixi bot

ARK Sõidueksami bot Küsib ARK-i lehelt uusimad eksami ajad ja saadab sõnumi Matrixi kanali Dev setup Linux python3 -m venv venv source venv/bin/activa

Arti Zirk 3 Jun 15, 2021
Automation of VASP DFT workflows with ASE - application scripts

This repo contains a library that aims at automatizing some Density Functional Theory (DFT) workflows in VASP by using the ASE toolkit.

Frank Niessen 5 Sep 06, 2022
This is a practice on Airflow, which is building virtual env, installing Airflow and constructing data pipeline (DAGs)

airflow-test This is a practice on Airflow, which is Builing virtualbox env and setting Airflow on that env Installing Airflow using python virtual en

Jaeyoung 1 Nov 01, 2021
MeepoBenchmark - This project aims at providing the scripts, logs, and analytic results for Meepo Blockchain

MeepoBenchmark - This project aims at providing the scripts, logs, and analytic results for Meepo Blockchain

Peilin Zheng 3 Aug 16, 2022
A Way to Use Python, Easier.

PyTools A Way to Use Python, Easier. How to Install Just copy this code, then make a new file in your project directory called PyTools.py, then paste

Kamran 2 Aug 15, 2022
A modern Python build backend

trampolim A modern Python build backend. Features Task system, allowing to run arbitrary Python code during the build process (Planned) Easy to use CL

Filipe Laíns 39 Nov 08, 2022
JARVIS PC Assistant is an assisting program to make your computer easier to use

JARVIS-PC-Assistant JARVIS PC Assistant is an assisting program to make your computer easier to use Welcome to the J.A.R.V.I.S. PC Assistant help file

Dasun Nethsara 2 Dec 02, 2022
The Official Jaseci Code Repository

Jaseci Release Notes Version 1.2.2 Updates Added new built-ins for nodes and edges (context, info, and details) Fixed dot output Added reset command t

136 Dec 20, 2022
inverted pendulum fuzzy control python code (python 2.7.18)

inverted-pendulum-fuzzy-control- inverted pendulum fuzzy control python code (python 2.7.18) We have 3 general functions for 3 main steps: fuzzificati

arian mottaghi 4 May 23, 2022
A simple countdown timer in eazy code to show timer with python

Countdown_Timer The simple CLI countdown timer in eazy code to show timer How Work First you fill the input by int-- (Enter the time in Seconds:) for

Yasin Rezvani 3 Nov 15, 2022
Basic repository showing how to use Hydra + Hydra launchers on SLURM cluster

Slurm-Hydra-Submitit This repository is a minimal working example on how to: setup Hydra setup batch of slurm jobs on top of Hydra via submitit-launch

Raphael Meudec 2 Jul 25, 2022
Fonts used to be an install-and-forget thing, but many of are now updated regularly.

Your font manager. Fonts used to be an install-and-forget thing, but many of are now updated regularly. fontman helps you keep track of the fonts you

Nico Schlömer 20 Feb 07, 2022
The ROS package for Airbotics.

airbotics The ROS package for Airbotics: Developed for ROS 1 and Python 3.8. The package has not been officially released on ROS yet so manual install

Airbotics 19 Dec 25, 2022
A collection of Python library code for building Python applications.

Abseil Python Common Libraries This repository is a collection of Python library code for building Python applications. The code is collected from Goo

Abseil 2k Jan 07, 2023