The Fil-C Optimized Calling Convention | Hacker News Reader

The Fil-C Optimized Calling Convention

fil-c.org

160 pointspizlonator32 comments

Great video here by the man himself: https://x.com/filpizlo/status/1976831020566798656

From my cursory glance, the real magic (InvisiCaps) appears to be a unique take on fat pointers to track types, access rights, etc. Pretty clever, and the website is a great technical read.

Dig the posters in the background; I just saw Burning Ambition in theaters last week. Up the irons, Earth dog! Ghost opened for Iron Maiden a few years ago; I saw them all together in Oakland.

> Fil-C is a personal passion project by Filip Pizlo.

Do I understand correctly that this project is based on the work of just one person, Filip Pizlo? If so, that's amazing.

Mostly. A handful of people have made some very nice contributions though

So you just need safe unicode identifiers I guess, fixing the longstanding unicode C11 spec bug, which made identifiers unidentifiable. Restricting to ASCII would be safest. In my rcc compiler I use my libu8ident

I came around to it a few weeks late, but Zef and this article by Filip are also great work!

https://zef-lang.dev/implementation

https://news.ycombinator.com/item?id=47843194

It has list of more than 20 optimizations for interpreters, with measured speedups. I'm pretty sure I was looking for something like this 3-5 years ago, but it didn't exist

One person who spent 15 years learning and building in the domain. He very much knows what he's doing, what questions to ask, and what machines do.

So for interpreted languages with types that are written in C, how is the engine supposed to tell C it already checked all the arg types manually in the interpreter? In other words: it's safe to go ahead and dereference this function and invoke it with these args.

Seems like C technically requires function declarations for every possible signature. That quickly explodes into hundreds or thousands of function declarations in the header and switch statement.

Edit: clarification

I’ve thought about how to let folks prove to Fil-C that Fil-C’s checks are obviated by some higher level checks.

It’s a super hard problem! I don’t have a good answer, but I also can’t prove that it’s impossible

Something something compile Fil-C to WASM64?

I don’t see how that would help

I guess it is the imaginary security that WebAssembly advocates tend to sell, without telling the part that linear memory segments don't have bounds checking within their internals.

If you have an interpreted language, you don't have a C function corresponding to each language function. You have a C interpreter loop with a "current instruction" pointer. When the current interpreted instruction is a call, you check all the things you need to check, push the current IP to a stack, and set the IP to the first instruction of the function.

C's type checker never sees the interpreted language's functions.

> Where my_thread is a pointer to the current Fil-C thread, which Fil-C passes around as the first argument in all calls.

Does this just mean you reserve a register for the current thread? In which case you could explain it as a reserved register (like FS used for TLS). Describing it as "passes around as the first argument in all calls" makes it sound inefficient–but whether it actually is depends on how you implement it.

It is exactly as inefficient as “passing it around as the first argument” implies

There’s a speedup to be had by either reserving a GPR or using one of the segment registers

Lots of obvious stuff like this hasn’t been done yet! If you want to have the satisfaction of landing speedups then Fil-C is a fun thing you could contribute to :-)

From what I understand, on x86 Linux stores a thread-local pointer to its TLS block in %fs. Could that simply be re-used?

https://groups.google.com/g/comp.arch/c/IT2dhS4q2M8?pli=1

Yes it could.

It would require more than zero work. Basically you’d need to unify yolo libc’s internal definition of pthread with libpizlo’s filc_thread

Are there any examples how to force C/C++ libraries within a Rust build to use Fil-C instead to improve security? Is it just a matter of overriding CC/CXX?

Won’t work

Can’t link Fil-C code to regular C code

And rust uses regular C ABI

You could make it work, if you teach Rust and Fil-C about each other. Nobody has done that (to my knowledge)

This would be a fun and popular project for the right sort of person

I keep day-dreaming about how to leverage the ideas in Fil-C. (a) use it for both Python interpreter and all C libs i want to call from there. (b) use the ideas for extra security in the OS kernel.

https://drive.google.com/file/d/1yVlKs_GPspxTq95MXLPgj5QuggT...

https://drive.google.com/file/d/14HG52S0TrrBbqwzisTupebCD1Im...

Do you think it’s possible to have fil-c ideas applied to protect unsafe blocks in Rust?

You’d have to teach rust about Fil-C pointers and their special rules. You’d also have to teach the rust compiler to play nice with the Fil-C GC.

It would be a big change. Probably not backwards compatible with today’s rust.

Interesting project in general. I wonder whether it could be adapted to behave reasonably without relying on threading. E.g. run the GC only when *alloc is called.

EDIT: misread the post! Never mind

You even read the comment you’re responding to? They’re saying no threads.

You're right. I can't delete anymore unfortunately

Pretty interesting, but what’s the reason of being for Fil-C?

Memory safety for existing C and C++ codebase.

Especially in systems where CHERI, MTE, ADI and similar harware isn't available.

There's a whole lot of C and C++ software out there, and Fil-C makes it memory safe, frequently with minimal work.

Can't speak to how everyone else is using it but at my job we run all of our unit tests under Fil-C as part of CI, in addition to the UBASAN, TSAN, and Valgrind pipelines we already had for them.

Reply on news.ycombinator.com