HuffmanDecoder Class to help reduce lag and other various performance optimizations - #285
HuffmanDecoder Class to help reduce lag and other various performance optimizations#285Vizonex wants to merge 3 commits into
Conversation
…to prevent performance regressions
|
I used ruff format src tests Hope you don't mind that at all. |
Sorry, something went wrong.
|
Please keep the style changes and other refactoring out of this PR - it makes reviewing the main changes difficult for me. Please explain the changes around HuffmanDecoder and the usage of array. I don't yet understand how the previous behaviour should have not been thread-safe, or what errors you encountered. As mentioned previously, the hpack library needs external safe-guards in any case. Please build a reproducible benchmark test and show results that your changes are better. |
Sorry, something went wrong.
|
@Kriechi I'm gonna shut down this pr and restart from scratch. Maybe we could add pytest-codspeed first and then start optimizing so that we get a clear idea of what we need to improve? |
Sorry, something went wrong.
Sorry, something went wrong.
These are my ideas on optimizations to make to make hpack run a bit smoother since it's pure python and I know of a few techniques that could help the code get moving and possibly help remove some lag if any are present.
There's some exciting plans happening including aiohttp getting this dependency hence my contributions.
fixes #284
This is still a draft but hopefully someone could help me review my changes :) I should be pretty responsive as my other goals in my agenda are just waiting on people to respond so I have plenty of time on my hands on this end.
Edit: It's not a draft anymore it's ready. :)