PyTorch Internals: Ezyang's Blog
This post is a long form essay version of a talk about PyTorch internals, that I gave at the PyTorch NYC meetup on May 14, 2019.
The latest blog posts to take off on Hacker News. 14,540 posts.
This post is a long form essay version of a talk about PyTorch internals, that I gave at the PyTorch NYC meetup on May 14, 2019.
Many command line utilities support short form options (-f) and long form options (--force). Short form is for interactive usage. In scripts, use the long form.
Software engineer by day, software engineer by night also.
To denote the subsequence of natural numbers 2, 3, ..., 12 without the pernicious three dots, four conventions are open to us
A security research blog.
Ever wondered how AI agents actually work behind the scenes?
Blindspots in LLMs I’ve noticed while AI coding. Sonnet family emphasis. Maybe I will eventually suggest Cursor rules for these problems.
Reverse Engineering Reality
Stephen Wolfram investigates the discovery of structures since the Game of Life was invented--searching possibilities versus invention by explicit effort. A thorough look at oscillators, modularity, glider guns.
Types in coding projects are good. Don’t be afraid to create them when you need to.
Build a Container Image from Scratch — Danish Prakash
LLMs are making software engineers dumber.
This post is about writing technical documentation. More specifically: it's about writing documentation for programming languages and libraries.
How to self-host a personal archive of web content even if stuff get's taken down
After the Graz, Austria February 2025 WG14 Meeting, I am now confident in the final status of the defer TS, and it is now time.
Today sees the announcement of the latest additions to Raspberry Pi's microcontroller line-up - the RP2350 family. I've had prototype units for a while, and you can run Rust code on it today. To my knowledge this is the first ever microcontroller launch with Rust support out-of-the-box.
As I’m getting older I look back on all the things I’ve done as a creative developer, and I see so many cool projects! But I never wrote down any development
This post illustrates a key lemma that is used in proving the fundamental theorem of Galois theory (FTGT). Note that FTGT is not covered in this post. The focus of this post is on understanding and proving this lemma only. Here is the lemma from the book Galois Theory , 5th ed. by Stewart (2023):
An indie project for early 2010s internet addicts
libexpat is a fast streaming XML parser. Alongside libxml2, Expat is one of the most widely used software libre XML parsers written in C, specifically C99. It is cross-platform and licensed under the MIT license .
History of the evolution of my C compiler and operating system for transputer when I was age 16.
So far, we’ve only really touched on some of the fundamentals of reverse engineering with one game example, but this time we’re looking at something a bit different. Shenmue (1999) is widely known for its pursuit of realism and has remained an influential series to this day, with its third game releasing in 2019. This…
Building on my previous post on sorting algorithms, I implemented the same algorithms using CUDA to explore performance improvements through parallel computing. The goal is to see how we can leverage the power of parallel computing to speed up our sorting algorithms. I went for a NVIDIA recruiting event some days ago…
Not long ago, I was auditing a codebase for work (looking for bugs) when I realized that despite the quality of the code, I was becoming mentally fatigued extremely quickly and had a hard time working on it for long stretches of time…
A deep dive into the performance of Python 3.14's tail-call interpreter: How the performance results were confounded by an LLVM regression, the surprising complexity of compiling interpreter loops, and some reflections on performance work, software engineering, and optimizing compilers.
And by LLMS I mean: (L)ots of (L)ittle ht(M)l page(S). I recently shipped some updates to my blog. Through the design/development process, I had some insights…
This post describes how to compile a single C++ source file to an object file with the Clang API. Here is the code. It behaves like a simplified clang executable that handles -c and -S…
Premise The 0.4.0 release of uv does everything I currently do with pip, pyenv, pipx, pip-tools, and pipdeptree. Because of that, I'm in the process of switching to uv. This blog post covers switching
To start off, we can leverage the STL for this calculation. std::count_if() more precisely: