Summary of How we are making Python 3.11 faster - presented by Mark Shannon

00:00:00 - 00:25:00

In this video, Mark Shannon discusses how Python 3.11 is made faster by optimizing code and avoiding memory accesses where possible. The interpreter has a specialized adaptive form that can optimize certain bytecodes, resulting in faster execution. These optimizations come with a risk of introducing new bugs, but Shannon feels that the increased performance is worth it.

• 00:00:00 The speaker discusses changes made to Python 3.11 in order to make it faster. The changes focus on improving performance of memory accesses, and avoiding the use of memory accesses where possible. The speaker also discusses the concept of "before object oriented programming," which is still relevant to low-level system programming.
• 00:05:00 In this video, Mark Shannon discusses how Python 3.11 is making code more efficient by avoiding dependent memory loads and memory allocations. The stack-based implementation of frames in Python 3.11 is faster than the older, cache-based implementation.
• 00:10:00 Python 3.11 is faster due to improvements to frame objects and zero cost exceptions.
• 00:15:00 The Python 3.11 interpreter has a specialized adaptive interpreter that can optimize certain bytecodes. This will result in faster execution of these bytecodes.
• 00:20:00 The video explains how Python 3.11 is made faster by specializing code. The different forms of specialization are load attribute instance value, load attribute class value, and de-optimization. The quickening form is when the basic straightforward form is changed to the adaptive form, which has a counter that warm-up. The next time the warm-up counter hits zero, the code enters the specialized form.
• 00:25:00 In this talk, Mark Shannon describes various optimizations that they have made to Python 3.11, which ultimately results in faster performance. They also mention that there is a risk of introducing new bugs, but feel that the increased performance is worth it.