IRC channel logs

<ekaitz>siraben: we should talk later

<ekaitz>siraben: about mes performance

<siraben>ekaitz: ok, roughly what time works in general

<ekaitz>siraben: now?

<ekaitz>basically the thing is those changes are obsolete, as my work in mes removes those parts of the interpreter

<ekaitz>but we still need to make a proper release with the work I've done

<xentrac>possibly it would be useful to propose an additional possible meeting time in addition to "now", in case siraben is asleep

<xentrac>like "now, but tomorrow", for example?

<ekaitz>I'll ping them again

<ekaitz>siraben: https://codeberg.org/ekaitz-zarraga/mes this is the relevant repo

<ekaitz> https://ekaitz.elenq.tech/tag/gnu-mes-interpreter-speedup.html and this is the context for it

<notgull>Is mes still the slowest part of the bootstrap?

<siraben>ekaitz: ok i am in a different timezone so I probably will be asynchronous

<siraben>(about to sleep)

<ekaitz>notgull: i'm not sure if it is the slowest

<ekaitz>siraben: just wanted to let you know that you don't have very good chances to upstream those changes

<ekaitz>because the code is gone

<siraben>ok that's fine, thanks.

<ekaitz>but we could try to merge them in an intermediate release if you need them for the short term

<ekaitz>basically that's the thing

<ekaitz>we are working hard to make mes substantially faster

<ekaitz>we already did

<ekaitz>but there's still a lot to od

<ekaitz>do*

<siraben>basically the philosophy is very simple and easy to iterate on whatever you do next. identify the hot loops and ensure things are inlined properly or try to make things division free

<siraben>it just took me an afternoon of profiling to find those hotspots

<ekaitz>yeah! they are very good catches

<siraben>related in the spirit of replacing MES, i forget if you have seen this already so forgive me if it's a repeat http://github.com/siraben/blynn-bootstrap

<ekaitz>now we are other hotspots that we'll try to fix soon

<siraben>bootstrappable via haskell, typed language, but there's an additional complexity of adding another IR since it's space intensive to do the string represetation. i don't expect this to replace MES it's just an experiment

<ekaitz>i think I've seen it. Very cool stuff

<ekaitz>what i did in Mes is to compile to bytecode first

<ekaitz>so now we basically removed that crazy goto soup that we had in the interpreter

<siraben>ah, great

<ekaitz>oh! you have a face!

<ekaitz>:)

<siraben>heh

<ekaitz>anyway, i just wanted to tell you that and if you want to take a look to the new Mes internals, I could guide you

<ekaitz>for anything you need about Mes just ping me or janneke, maybe here or in #guile

<siraben>great, i might take a look again tomorrow, if i have time

<siraben>and another thing, https://github.com/siraben/live-bootstrap-bake/ , minimal build system to show that net LOC can be reduced

<xentrac>it is common for the humans to have faces, I have found

<ekaitz>xentrac: fake news!

<siraben>net -3,719 after converting all the kaem scripts to .bake files

<ekaitz>damn!

<siraben>ok, good night

<ekaitz>siraben: thanks for sharing this, good night

<matrix_bridge><Andrius Štikonas> notgull: it is slowest in the beginning, but e.g. modern GCC or guile or other big things towards nthe end of bootstrap are slower

<matrix_bridge><Andrius Štikonas> siraben: speak to fossy regarding bake files

<matrix_bridge><Andrius Štikonas> I don't particularly mind, should be fine i think

<matrix_bridge><Andrius Štikonas> Though i also didn't find kaem files too hard either

IRC channel logs

2026-08-25.log