文件名称：Python的全局锁问题-python cookbook(第3版)高清中文完整版
文件大小：4.84MB
文件格式：PDF
更新时间：2024-06-29 23:06:43
python cookbook 第3版 高清 中文完整版
12.9 Python的全局锁问题 问题 You’ve heard about the Global Interpreter Lock (GIL), and are worried that it might be affecting the performance of your multithreaded program. 解决方案 Although Python fully supports thread programming, parts of the C implementation of the interpreter are not entirely thread safe to a level of allowing fully concurrent execution. In fact, the interpreter is protected by a so-called Global Interpreter Lock (GIL) that only allows one Python thread to execute at any given time. The most no‐ ticeable effect of the GIL is that multithreaded Python programs are not able to fully take advantage of multiple CPU cores (e.g., a computationally intensive application using more than one thread only runs on a single CPU). Before discussing common GIL workarounds, it is important to emphasize that the GIL tends to only affect programs that are heavily CPU bound (i.e., dominated by compu‐ tation). If your program is mostly doing I/O, such as network communication, threads are often a sensible choice because they’re mostly going to spend their time sitting around waiting. In fact, you can create thousands of Python threads with barely a con‐ cern. Modern operating systems have no trouble running with that many threads, so it’s simply not something you should worry much about. For CPU-bound programs, you really need to study the nature of the computation being performed. For instance, careful choice of the underlying algorithm may produce a far greater speedup than trying to parallelize an unoptimal algorithm with threads. Simi‐ larly, given that Python is interpreted, you might get a far greater speedup simply by moving performance-critical code into a C extension module. Extensions such as NumPy are also highly effective at speeding up certain kinds of calculations involving array data. Last, but not least, you might investigate alternative implementations, such as PyPy, which features optimizations such as a JIT compiler