| ▲ | Show HN: An ObjC-like compiler for Mac ARM, x86_64(Linux, Win), WASM, & others | |
| 2 points by spacedcowboy 6 hours ago | 1 comments | ||
Hi folks, So this is my announcement of a compiler[1] I've been working on for about 6 months now (so yes, to get the obvious out of the way early, this was written with Claude code). The compiler is called 'xc', stands for 'cross-C' or 'extended-C' or whatever you want, really. It's fairly similar to Objective-C in style (without the [] brackets), and in fact the first version of the language was written in ObjC. It works on Mac M-series, Windows, and Linux - and any of these hosts can create binaries for any of {Mac M-series, Windows, Linux, Android, iOS, WASM, Arm9 (Zynq), m68k and even 6502}. Building the compiler needs host-tools to bootstrap everything, either GNUStep on Linux, or a Mac with Xcode. Once built, however, the system is entirely self-contained, and you don't need any platform tools. I have a binary running on my iPhone which was compiled and signed on a Linux box... The suite ships with 5 host-specific assemblers, 10 object/executable writers, and a native Apple code-signer (we don't shell out to 'codesign') Downloads are available at [3] for all three hosts (xcc-osx-0.6.tar.bz2, xcc-linux-0.6.tar.bz2, xcc-win64-0.6.zip). Linux and Windows are larger because of static linking to make sure things are easy to run. The documentation for the language is all at [2], it goes into the language details, but you get ARC (Automatic reference counting, which automatically goes atomic when you have threads), blocks, callbacks, Classes, and Protocols. It's an optimising compiler with a shared IR, deliberately not LLVM's IR because one of the original targets was the 6502, and LLVM likes register-rich architectures. The compiler runs Vectorize, LICM, LoopRotate, LoopUnroll (fixed and variable trip), StrengthReduce, Inline, TailRecursion, NarrowIV, PointerIV, RedundantLoadCSE, IdiomMemset, LoopReductionCollapse,... One major inter-operability feature is the ability to say #import <GEM> (or #use, which also promotes its statics into the bare-call space) and have the compiler read libGEM.so, parse its DWARF, and expose the functions, types and enum constants inside. Xc uses the platform's native structure packing, so binding a C library is mostly just pointing -L at it. The Licence for the compiler is GPL3, for the standard library there's the "standard" library exception, since that code is bundled into most apps. Bottom line, you are free to create any applications you like, open or closed, without any licensing issues. If you make a fix to the compiler, I would appreciate the bug report/patch. A utility in the tools/ section is c2xc. This takes almost all C code (I used it on Kundert's Sparse 1.4 — sparse.sourceforge.net) and converts it to valid XC. It might not be the best XC possible, but it will compile, and if you're trying to produce WASM code and can't just #use a dynamic library, this is the important part :) There's also the beginnings of a cross-platform UI toolkit and Interface-Builder, though these are very early days. There are 568 fixtures, 608 unit-tests, two simulators (6502 and 68k) and ~50 differential scripts to check output vs clang / binutils. The compiler is written in its own language, forming a "reference" ObjC compiler and the "shipped" written-in-xc compiler (all 120k lines of it), meaning it has two chances to find obscure bugs. Other real-world tests have been: - transpiling Kundert's Sparse 1.4 to xc and diffing its test program's output against a clang build of the same library - a circuit simulator built on it, checked against ngspice-47 and against an independent numpy MNA implementation - writing a "test in anger" application, a server that listens for incoming web requests, parses them, uses them to query a DB and then sends the results out to a web-client (written in XC, and compiled to WASM) Enjoy :) | ||
| ▲ | Rochus an hour ago | parent [-] | |
Wow, this is huge. Is the goal really to have a new language better than Objective-C, and to actually use it in projects, or mainly to demonstrate what an AI can do? As far as I understand, there is a non-trivial optimiser, assumingly also generated by AI, including the associated IR and lowerings. Are there comparisons of the resulting machine code quality and performance compared to e.g. GCC? Did you design and specify the xc language, and did the AI implement the frontend based on your specification, or was it a "full package" AI approach? (sorry if some questions were already answered on the web sites and the code, but it was too much to quickly get the answers) | ||