We'll be in London next week! Come hang out with us 🤍

doing huge architecture changes and scope reduction is both so depressing and so exciting

want to change my versioning scheme but screwed myself as npm can't rewind versions. my options are: 1. bypassing this restriction somehow 2. going to 0.1xx.0 3. going to 1.0.0-alpha.x any opinions/ideas?

update: writing a borrow checker 🥴

any js (library/runtime) folks want to help/give feedback on an api design? :)

new gc is both so exciting and crash out worthy for debugging gc bugs 😤

Map and Set are now up to over 300x faster :)

do you want to see exciting demos even if impractical/will likely never come to fruition? yay or nay

(very!) early preview for a new experimental threading Porffor api. it enables intentionally low level concurrency to (try to) enable the most efficient communication feasible. here are 1 million threads alive simultaneously using only ~16gb. 16kb/thread, <1us start avg.

cleanup grind continues

optimized selfhosted porffor binary is now <20mb and only takes ~1 minute to compile (>90% clang). (% change is vs debug binary) still buggy though. also still too big ;)

self compilation is now ~15% faster after some optimizations for 2c itself, and emits a smaller (unoptimized debug!) binary

after more 2c optimizations

cleaning up code directly correlates to smaller (unoptimized!) selfhosted binary size 😇 (the ~release binary size is much much smaller)

elysia uses >2x less memory with ~15% lower latency and higher throughput vs ~a week ago :)

in the last week I have written: a (new) gc, a wasm engine, a wasm compressor (via huffman coding) 😅 (maybe gonna drop the wasm engine though)

why is homebrew 3 node minors behind :(

fun data from optimizing how builtin apis are embedded (they are precompiled). numbers are still unoptimized/not final! oh and btw porffor can selfhost now ;)

hono uses ~10% less memory with ~5% lower latency vs a few days ago, thanks to new gc and other allocation optimizations

Porffor's optimization phase is now 5-10x faster for large files

forgot I added some debug logs um

if you want a more traditional/detailed profiler I got you too ;)

new experiment

UK people: please write to your MP asking them to sign EDM 240. It disapproves the EHRC's latest Code of Practice which forces trans segregation. This is likely the only thing left that could stop it. edm.parliament.uk/early-day-mo... You can use www.writetothem.com, it only takes a few minutes ♥️🏳️‍⚧️

hmm

i guess everyone is making code review tooling now. idk if I'll release this yet

rewrote 2c and now it emits ~2x less lines of c :)

learning to get less attached to old code. you can just delete things

enjoy 2 new blog posts! - Porffor now has closures: surprisingly annoying for a compiler. goose.icu/porffor-clos... - Profile-guided dead code elimination: a novel approach for DCE in dynamic languages. goose.icu/profile-guid...

new tech (ignore the last one)

oops

nevermind, porffor.dev containers use ~600kb of memory now 🤯

deployed latest porffor to porffor.dev (previously running a ~week old version). the improvements are obvious 😍

now 5% less memory usage, faster and <180kb binary size :)

hello world http server is now under 200kb 🥳

unironically I want this to be lower but the majority of my memory usage is just libc(/++) being loaded 🥴

container running porffor.dev is using 1.1mb of memory btw

one week of uptime 🥹

some more fun graphs from load testing today

fun fact: you can see top secret porffor details of my sites with /metrics (built into my framework for grafana). goose.icu/metrics porffor.dev/metrics

would you prefer 1 big general blog post or multiple smaller ones focused on one thing?

still optimizing and fixing, eg below ;)

I don't think tech at large understands how hard it is to be an open source maintainer right now. So many great folks are on the brink of burnout. Funding is drying out. We live in a security minefield. Companies think humans are disposable. If you're a maintainer, we care. You deserve the world.

there are more demos which are an even larger deal btw #vagueposting

testing http stuff after fixing closure bugs, it crashed so I thought I had a regression but just forgot I changed my hello world to this yesterday

half my remaining regressions are actually tests which were false positive due to now fixed bugs 😳

fixing regression hell from a ~5k diff 🥴

real world load testing! with 3K req/s (2.5M requests total), I peaked at: ~22% usage of 1 core, ~2.6MB memory usage, 9ms p99 response time. went better than expected 🥳

today in unnecessary infrastructure: made my own status page from scratch with automated incidents, paging. it's also hosted independently so in critical incidents it would remain up status.porch.porffor.dev