26 May
2025
26 May
'25
12:08 a.m.
Besides which x86 or other machine models you want, and how much of
which sort of optimization is enough, the other challenge is which C
language variant, exactly, you want. Simple C compilers that do one
thing well, often, well, do just one of these things.
On 05/25/2025 04:46 PM, Bakul Shah via TUHS wrote:
In May 23, 2025, at 5:45 AM, Steve Nickolas
<usotsuki(a)buric.co> wrote:
On Fri, 23 May 2025, Steffen Nurpmeso wrote:
There are
a number of such efforts. Apart from the ones
already mentioned, there are at least
https://git.sr.ht/~mcf/cproc & https://www.simple-cc.org/
May be you can help them out....
The difficulty is in maintaining such compilers, standard
compliance, complete toolchains for supported platforms, other
support programs, dealing with platform changes etc.
Machines are fast enough (for me) and I would love it if I
can compile the (FreeBSD) kernel + most of the userland using
a fast & simple compiler but that is just not possible as
some compiler specific dependencies have slowly crept in
standard header files.
It is *such* a pity! I said similar sad words
just two days ago
when shortly touching linux-man@. That we lost (i only track your
git mirror of) it, and are left with only gigabyte monsters that
go universes beyond Ken Thompson's "reasonable optimizations"
(iirc), and tcc (luckily this we have). Here the built gcc ball
is 243 times larger than tcc's, and clang is 284 times larger
even!
I wish I had any idea what I was doing when it came to language interpreters
and compilers... These swiss-army-nukes epitomize "no kill like overkill", but I
prefer small, single-purpose tools.
A new lightweight C compiler with a focus on various varieties of x86 is something I
think would be useful and would do if I had any idea how to go about it.