Global Interpreter Lock: Episode I - Break the Seal

Speaker

Tzung-Bi Shih /Mandarin

Tags

multi-threading, performance, parallel computing

Abstract

In order to achieve high throughput, we usually execute independent tasks simultaneously in a process through leveraging multi-threaded technology. The tasks are truly parallel because multi-processor and hyper-threading technologies are quite mature in modern x86 CPU. While thing happens within CPython, the behavior will not be what we think. The talk addresses on CPython limitation of parallelism and enumerates some common practices to evade the seal.

About Speaker


Tzung-Bi Shih was born in Xi-Hu, Taiwan, in 1988. He received Bachelor and M.S. degree in Computer Science from National Chiao Tung University in 2010 and 2011, respectively. He was a R&D engineer in Trend Micro during 2011-2014. He has been a Python programmer for 2 years. He is a beginning researcher and he is interested in operating system, computer architecture, especially, in parallel computing.

Organization/Company

Independent