Is it cheating to have a single benchmark that uses random distributions within it to vary the proportion of CPU/GPU/IO?

Example of an Inkle loop. A structure with pegs to weave a long, narrow fabric.

Learnt about the inkle loom over the break and it sparked a cool family discussion about fractional dimensions! In an ideal world, on an ideal loom, how long would the longest shoelace you could make on an inkle loop be?

@tomdeakin.bsky.social gave a great tutorial at #SC5 today. My GPU programming fears are dissuaded! Thanks for entertaining all my questions.

#sc25 going to hit up the Museum of Westward Expansion over lunch. DM to join me if you'd like to :)

Bring back embroidered patches on the shirts!

Thank you! I'll see you tomorrow

It's my first SC #sc25. Would love to meet a lot of new folks over the next week! Please feel free to hit me up and also to adopt me into your well formed groups :)

Q: Do folks have some paper pointers to studies of how human-usable HPC frameworks are?

I'm thinking about BLIS, Halide, TVM etc., where the claim is that these serve as productivity multipliers.

(Disclaimer: I also have one such framework, wondering if it is usable)

Had a worlds colliding moment when using concepts from my Evidence-Based Teaching in STEM class to get LLMs to generate microkernel for matrix multiplication actually worked! Exciting stuff.
#FacDev #HPC #GenAI

Hello BlueSky!
Missed the Twitter train so I'm excited to meet and learn from folks on here :)