Sam Westrick @shwestrick@types.pl

Some of Millet devs at CMU made shirts, so I had to get one. It looks so good!
https://azdavis.net/posts/millet/

This is a good link to keep track of... you know, just in case you might need a graph with 128 billion edges down the line

https://webdatacommons.org/hyperlinkgraph/

A nice little data structure problem for efficient reversal of contiguous subsequences: https://cs.stackexchange.com/q/156028/66544

mmap/munmap are very expensive and very not parallel, oof

siri doesn't know know how to spell POPL

AFAICT the ocaml memory model guarantees this ❤️​

https://kcsrk.info/papers/pldi18-memory.pdf

and they were able to implement it with low overhead. But I wonder if it's actually portable to other languages?? OCaml boxes a lot of things, and heap boxes make it pretty easy to ensure "atomicity" in this sense.

relaxed behavior seems fine, as long as it's always "atomic" (each value read should always be a value "previously" stored).

I feel like the future of memory models is to achieve "data-race-free" semantics -- not in the sense of assigning good semantics to programs without data races, but rather to remove the notion of a "data race" entirely...

(noticed today that I was browsing CS slack, types.pl, typ.zulipchat, the birdsite, discord, and reddit, and enjoying each one separately for what it is)

it's fun to be involved in many different social media platforms, even if there is overlap between them. It's like having N different group chats with the same group of friends. Each group chat can serve a different purpose, and can carry a different conversation. I like that.