Watching now

Here are some projects that are occupying some brain space and that I am keeping an eye on, in no particular order (to demonstrate that, I shuffled before publishing). Some are because they do a lot with a small amount of code/people, some are because they feel like they are poking at something very new and cool, and some are excellent learning resources.

• Roc, a fast, friendly programming language
• YJIT, a JIT for Ruby based on Maxime’s research on basic block versioning
• bril, an educational compiler IR
• AtomVM, a small BEAM (Erlang) VM for small devices
• onramp, a bootstrapping C compiler
• wild, an incremental linker
• Taskflow, a graph-based parallel task system
• Alan, an autoscalable programming language
• elfconv, an ELF to Wasm AOT compiler
• bifrost, a p2p library that supports multiple transports
• dulwich, a pure Python Git implementation
• HVM, a massively parallel interaction combinator VM for the GPU
• Pyret, a functional programming language with some excellent ideas and a great pedagogical focus
• ir, the JIT internals for PHP
• Iroh, a toolkit for building distributed applications
• simple-abstract-interpreter, just what it says on the tin
• ssa-optimizer, an educational SSA-based optimizer
• hindley-milner-python, a small Damas Hindley Milner implementation in Python
• Hindley–Milner in Python, another one of the same (but this one is a bit “trickier”)
• micrograd, ccml, and cccc, small, educational autodiff libraries
• Natalie, an AOT Ruby compiler
• plzoo, which has multiple different small PL implementations with different semantics
• tinygrad, a small autodiff library with big plans
• MaPLe compiper, a compiler for automatic fork/join parallelism in SML
• try, to inspect a command’s effects before modifying your live system
• mold, a fast and parallel linker
• monoruby, a full Ruby VM and JIT by one person
• MicroHs, Haskell implemented with combinators in a small C++ core
• Cake, a C23 frontend and compiler to older C versions
• Toit, a VM from the Strongtalk lineage and management software for fleets of ESP32s
• Cosmopolitan, a C library that can build very portable self-contained binaries
• gf, a very usable GUI GDB frontend
• MatMul-free LLM, an implementation of transformer-based large language models without matrix multiplication
• Bun, a fast JS runtime based on JavaScriptCore
• Porffor, a from-scratch AOT JS engine
• weval, a WebAssembly partial evaluator
• Riker, correct and fast incremental builds using system call tracing
• arcan, a new display server and window system
• container2wasm, to run containers in WebAssembly
• LPython, a very early stages optimizing Python compiler
• Pydrofoil, a fast RISC-V emulator based on RPython, the PyPy internals
• Fil-C, a project where Fil Pizlo is making a memory-safe version of C
• Oil shell, a new Bash-compatible shell with fresh ideas and its own Python-esque compiler
• bigint, a small arbitrary precision integer library for C
• chibicc
• bcgen, which is kind of like Ertl’s VMGen, and copyjit, which is like Copy and Patch
• joos, a KVM virtual machine manager in JavaScript

Not quite code but presenting very cool ideas:

• Verifying your whole register allocator too hard? No problem, just write a verifier for a given allocation and abort if it fails. This also lends itself nicely to fuzzing for automatically exploring large program state spaces.
• Copy and Patch (PDF) compilation, which generates pretty fast code very quickly
• Egg, and more broadly egraphs, for program IRs
• Implementing a Toy Optimizer and union-find in general
• (To-be-written) Using Z3 to prove your static analyzer correct
  • More Stupid Z3Py Tricks: Simple Proofs is a good jumping-off point, as is Compiling with Constraints
• (To-be-written) From union-find to egraphs; exploring the tradeoffs
• a simple semi-space collector explains how semi-space GCs work in ~100 lines of C
  • Note: the zero-length array member payload is non-standard and probably not needed
  • Note: is_forwarded should actually check if if the masked tag is == 0
• just-in-time code generation within webassembly and its accompanying wasm-jit
• Regular Expression Matching: the Virtual Machine Approach
• Interaction nets and HVM
• Make Your Self
• Cheney list copying / GC
  • Cheney on the MTA
• Escape analysis and dead code elimination as compile-time GC
• PGO for DCE: tweet one and tweet two
• Nostr, but without all the blockchain stuff

What I am working on:

• Scrapscript, an interpreter and compiler for a small functional language
• Dr Wenowdis (PDF), where CF Bolz-Tereick and I are working on making C extensions faster in PyPy
• A small rasterizer for my graphics class that uses fenster
• weval-ing CPython
• Introduction to Software Development Tooling