Python advanced #001 GIL (2020)

-


 

1. What is GIL?

The GIL (global interpreter lock) is used by CPython interpreter which written in C programming language. There are other python interpreters that are written in other languages, such as JPython written in Java.

CPython uses this GIL lock to make sure that only one thread executes python bytecode at a time. The other thread must wait for its release.

Using this GIL, CPython ensures that when a thread is executing python bytecode, any python objects will not be accessed by another thread. It provides a safe threading way to achieve multitasking.

Its downside is more obvious in modern computers which always feature multi-core cpus. Because only one thread could execute code at a time, CPython will only occupy one cpu core at a time leaving other cpu cores idle. 

1.1 Why Guido doesn't remove it

Actually, Guido (creator of Python) did that several years ago. But eventually, he quit and make a statement that removing the GIL is not an easy task; by removing GIL could suffer the performance of the python programs in a single-processor case and overcoming this performance issue is believed to make this implementation much more complicated and costly to maintain.

1.2 Does multi-threads program run faster than a single thread one because of GIL?

The answer is "yes" or "no".

When you are doing IO tasks using multi-threads, your program definately run faster than a single-tread program.

IO tasks include networking, read/write files etc. For instance:

client_socket.recv(1024)

When we want to receive data from a client socket (refer to networking topics in this blog), the program will block here and wait for incoming data. After 5 minutes, data comes and the program no longer block and continue to execute the next code and then handle the next incoming data.

If there are three requests at the same time, our single-thread program could only handle one at a time which will take 5 minutes. So after all, it will cost 15 minutes to handle all those tasks.

However, if we use multi-threads, when the first task is waiting for the incoming data in one thread, our GIL lock will be released and other threads could execute its code and start to receive data. So it only takes 5 minutes to receive all those three requests.





Comments

Post a Comment

Popular posts from this blog

How to create a memory puzzle game with Python and Pygame (#005)

-
First look at the following code: import time # create Font object text_font = pygame . font . SysFont( 'arial' , 50 ) # create a list that stores two pos that going to compare to each other pos_to_compare_list = [] # create a list that stores all the pos that matched pos_matched = [] # set backgroud color to white while True : for event in pygame . event . get(): if event . type == QUIT: pygame . quit() sys . exit() elif event . type == MOUSEBUTTONDOWN: (x,y) = pygame . mouse . get_pos() pos = check_click(x,y, square_pos) if pos: text_surface = text_font . render(board[pos[ 0 ]][pos[ 1 ]], True , WHITE) if len (pos_to_compare_list) < 2 : if pos not in pos_to_compare_list: pos_to_compare_list . append(pos) ...
Read more

How to write a slide puzzle game with Python and Pygame (2020 tutorial)

-
Image
Table of Contents Preparation Introduction Import modules and define contants Create the main game screen Creae the data structure of the board Let's draw the tiles The game logic Define the missing functions Preparation In this tutorial, we are gonna use Python 3.8 and Pygame v1.96.  For the installation of Python and Pygame, please follow the documents found on each webiste. Python: https://www.python.org/ Pygame: https://www.pygame.org/ Introduction As you can see from the picture above, we create a board with 15 green tiles and one white tile which represent a empty slot. Our gaol is to place all the 15 green tiles back into its sorted order, namely: A B C D E F G H I J K L  M N O If we click on the tile "F", the tile will slide down south. And if all the tiles are back into their sorted order, you won the game and quit the game. That's pretty much the gist of the game, later you could add other features to the game. Import modules and define contatnts impor...
Read more

Introduction to multitasking with Python #003 gevent (2020 tutorial)

-
Image
  1 . Review of Iterator In python, we could achieve multitasking through another way by using gevent module. Howerver before introducing gevent, we have to review the concept of iterator. You may have used python for loop millions of times: for i in Iterable:     print(i) We have learned several iterable objects, str, list, tuple, dict and set. Those iterable objects or sequences could be used with the for loop. And we take that for granted. But look at the following example: >>> for i in 100: print(i) Traceback (most recent call last):   File "<pyshell#2>", line 1, in <module>     for i in 100: TypeError: 'int' object is not iterable >>>   Why cannot int type be used with for loop in Python and how could you make an object iterable and be used with for loop? Now, let's make an iterable object that could be used with for loop. We could use isinstance() built-in method to check whether an object is iterable or not. ...
Read more