📢 Exciting FPTalks coming up! 🗓️ Join us April 3rd at 9:00am PT for Artem Yadrov talk on Mixed Precision Error Bounds! Updates / discussion: fpbench.org/subscribe Nominate contributions: fpbench.org/nominate

Really proud of my latest project on incremental layout in web browsers. It's got PL (attribute grammars), my favorite algorithm (order maintenance), cache optimization, and even inline assembly. Incredibly satisfying to see every level of the stack work together like this. (...)

Every Marisa "Long Middle Name Looks Cool" Kirisame project is like this: whole stack working together, algorithm through assembly. He's graduating in a year, folks, if you want to hire someone like this, reach out!

Very cool work reducing latency in web rendering engine through clever use of priority queues and order maintenance data structures .. improves upon the current Double Dirty Bit algorithm .. co-authored by Marisa "Long Middle Name Looks Cool" Kirisame Link: 👇

the government just disappeared a computer science professor, I don't know much more than anyone else, but at face value this is extremely alarming

📢 Exciting FPTalks coming up! 🗓️ Join us May 1st at 9:00am PT for David K. Zhang 's talk on Fast Branch-Free Algorithms for Extended-Precision Floating-Point Arithmetic! Subscribe for more: fpbench.org/subscribe Nominate a speaker: fpbench.org/nominate

I think the appetite for software is nearly infinite. I've been using AI extensively to write code and yet the number of things I still need to code has increased, not decreased. It's like we added three lanes to the highway and still have traffic.

Incredibly exciting to write a microarchitecture simulator and have it match cycle counts on the actual machine, even under load.

🚨 FPTalks 2025 is coming July 10 🚨 Join online for the latest in floating-point and numerical computing from speakers at CNRS, ARM, Intel, Graphcore, Cornell, Inria, and more. ⚙️ Research, tools, hardware, theory: all in 10-min lightning talks + Q&A! fpbench.org/talks/fptalks2…

Over the 10 days, I wrote a new Racket library for "expect testing", a style that Yaron (Ron) Minsky has advocated for in OCaml. github.com/samth/recspecs/ As an experiment, I built it entirely with Codex (the OpenAI async AI programming tool). I have some thoughts.

# better diff views for AI agents +1 to the quoted post. To work well with AI agents, we need better diff views! your ability to check the agent's work is the bottleneck. here are some ideas for how to do that, based on our research at Ink & Switch : Zoomed out diffs: show

Hi all! FPTalks FPTalks, our free annual workshop on floating-point and numerics, is in a week, on Thursday morning PDT. There's still time to register: fpbench.org/talks/

Pavel Panchekha Yay finally arrived (!) Much excite !!!

I use LLMs a lot, but I think the killer app is learning stuff fast. I am hoping LLMs make it possible to do much more creative research as a result.

Today's the last day to register for FPTalks 2025: fpbench.org/talks/ Come for amazing numerics talks across hardware, software, and applications. Workshop is tomorrow over Zoom.

Next FPTalks: Thu Aug 7 @ 9AM PT Detecting and diagnosing FP exceptions in GPUs and CPUs Baranowski & Gopalakrishnan (Utah) Come learn how we track down NaNs and INFs in closed-source GPU kernels launched from Python, Julia, etc! fpbench.org/subscribe fpbench.org/nominate