People talk about profiling to find bottlenecks, but no one explains how to fix them. Many devs shoot in the dark and when the optimization backfires, they aren't sure how to proceed. This is where you wish for a theory to systematically reason about the optimizations so that

The more oop you write, the more poop you get. Try for yourself: oopoopoop.

TIL: The earliest example of computational complexity is from 1844 when Lamé studied worst case complexity of Euclid's algorithm for computing GCD.

The rivalry between Knuth and Douglas McIlroy is well known. While McIlroy favoured real-world implementation, Knuth was known for mathematical and algorithmic elegance. When asked to implement a program to list the most frequent words in a file, Knuth wrote a 10-page program

RT @BengaluruSys: Bengaluru Systems is proud to partner with @fifthel and @abhi9u @anscombes4tet @Aditi_ahj on this fantastic workshop - if….

My north star for choosing topics is their (relative) uniqueness, and this one is probably the most original piece I've ever written. The idea itself is not new but the angle of holistically analyzing various optimizations is something i've not seen anywhere else.

Errata: Also, increased penalty on cache misses and branch misses increases the CPI (and lowers IPC). I reversed the direction. Again, all of this is correctly explained in the article.

Errata: In the first post I meant to write performance = 1 / (instruction count x cycles per instruction x cycle time). I mistakenly wrote instructions per cycle (IPC) instead of cycles per instruction.

This article is made available to everyone thanks to a sponsorship from .@coderabbitai. CodeRabbit is your AI co-pilot for code reviews. It instantly comments on pull requests, suggests one-click fixes, and catches subtle issues using custom AST Grep rules. Free for all.

Read it here -

RT @mgill25: @MoHFW_INDIA @JPNadda My mom is a retired govt. employee who recently retired, got diagnosed with Cancer. Your useless govt.….

RT @anscombes4tet: Join us at @fifthel next week for a free hands-on paper reading workshop where we’ll dissect a ML-Sys '25 paper from pro….

RT @abhi9u: When optimizing a loop in Python, a performance trick is to create a local alias for global functions or variables used in the….

This article was sponsored by @coderabbitai which allowed me to release it freely for everyone. Very grateful for their support. CodeRabbit is your AI co-pilot for code reviews. It instantly comments on pull requests, suggests one-click fixes, and catches subtle issues using.

Full article:

When optimizing a loop in Python, a performance trick is to create a local alias for global functions or variables used in the loop. E.g.:.l = len.for _ in range(1_000_000):. l(lst). But we still need it given so many perf improvements in Python? I did a benchmark and it appears

Thank you so much, Vipul. I'm very grateful for your sustained support.

RT @bml_khubbard: Here's a link for the eBook (PDF) version of my book. For some reason it is on a different web-page. .

I will follow whatever rabbit holes I want to. You are more than welcome to not read.

RT @abhi9u: I've been running a series on x86 assembly at Confessions of a Code Addict. Recently covered system calls. One interesting aspe….