IRC channel logs

<siraben>stikonas: running some tests to ensure the m2-planet pr isn't going to change stuff downstream

<siraben>we really should have CI for all this tbh

<matrix_bridge><Andrius Štikonas> siraben, yeah thanks

<matrix_bridge><Andrius Štikonas> it is very helpful if you can test it with some downstream stuff

<matrix_bridge><Andrius Štikonas> yeah, something that clones stage0-posix, replaces submodules with PR and checks it completes

<siraben>Looks like removing IR optimizations help the overall time-to-first-tcc metric

<siraben>while barely making a difference on the performance of the first tcc compiled by hcc, and 470 lines removed too

<roconnor>Is time-to-first-tcc the best thing to optimize?

<roconnor>I guess removing lines of code is good.

<siraben>roconnor: it's one of the things to optimize. I'm trying to balance: simplicity of implementation, LOC, time to first tcc

<stikonas>ability to port to other architectures is even more important

<stikonas>we have some very fast builds to tcc

<stikonas>but x86 only

<stikonas>time is also hard to measure, depends a lot on the system

<stikonas>e.g. on riscv board that I have it is maybe 2 weeks to tcc

<siraben>Yeah I'm only doing relative timings of course, some things are clearly linear wrt. source code

<siraben>I've only ever run riscv in qemu heh

2026-05-16.log