>compilers today
>overly complex code
>full of assumptions and heuristics
>all backed by llvm
>definition of bloatware
why didn't they keep compilers simple?
>compilers today
>overly complex code
>full of assumptions and heuristics
>all backed by llvm
>definition of bloatware
why didn't they keep compilers simple?
Other urls found in this thread:
ethoberon.ethz.ch
dbai.tuwien.ac.at
arxiv.org
en.wikipedia.org
github.com
twitter.com
You can always use the old versions.
did you post this on a teletype machine? Oh you did over a global packet-switching network that can heal itself when nodes fail? Using a machine that dynamically lays out hypertext in sub millisecond times? Seems complex to me.
because reality isn't simple suckless homo
What you call bloat are wanted features for others.
Look at a compiler like TCC. Sure, it's fast and simple. The dream of every minimalist. But if you're looking for optimized output rather than fast compiling it's not for you.
speed of compiled code justifies everything, that's why.
because having compilers that generate better code is better
the point of compilers is not for you to tinker with the source code, the point is for them to generate good binaries
>that can heal itself when nodes fail
That was the goal but it seems the internet fell short of it.
That book sucks so much. What a horrible way to write my first compiler backend.
Only compilers backed by thousands of devs (and not even most of them) use an llvm backend because llvm changes in incompatible ways in every single minor point release, both for APIs (which go out of sync with the opcodes) and opcodes.