Posts

Writing Something Real in OctoGo: A PDP-11/40 That Boots Unix V6

-
Image
The OctoGo post in July ended with a question: whether the zero-allocation constraints feel workable once you try to write something real. This post is the something real. p2-11 is a PDP-11/40 emulator for the Propeller 2, written in OctoGo. It boots Unix V6 and RT-11 from RK05 disk packs kept as files on the SD card of a P2 Edge. V6 logs in, and compiles and runs C with its own compiler. Here it is doing that, recorded on the board at real speed: That is 95 seconds from ogo run to the last sync . The first twelve are the build, the load, and the emulator timing itself; then V6 boots, root logs in, and cc takes about half a minute over hello, world. How it was built Plainly: with AI help, and a lot of it. Claude, Anthropic’s model, working in Claude Code sessions, wrote the emulator, its tests, its scripts and the project’s notes. 57 of the repository’s 58 commits carry its Co-Authored-By line; the one that does not is the initial commit, an empty README. My part was the dire...
Post a Comment
Read more

Building OctoGo: A Go-like Language Where Goroutines Are Cores

-
The last few posts here have been about pieces of a toolchain without saying what they were for. egg , the LL(1) parser generator. The flat AST and the ergonomics of walking one when you have decided that allocation is the enemy. And underneath both, ccgo, which turns C into Go. This post is what they were (also) for. OctoGo is a Go-like language for the Parallax Propeller 2. egg generates its parser. Its AST is a flat []int32 . Its backend is a C compiler that ccgo transpiled into the binary you install. And the thing that makes it worth writing about is not any of that plumbing - it is that on this target, a goroutine is not a scheduler entry. It is a core. The target, briefly Most readers will not have met a Propeller 2, so: it is an eight-core microcontroller with no interrupts in the usual sense, no MMU, and no operating system. The cores are called cogs. There are eight of them, they are symmetric, and they run genuinely in parallel. There is 512 KB of shared hub RAM, ea...
Post a Comment
Read more

Taming the Regex Monster: Optimizing Massive Literal Alternations

-
Image
A recent discussion on the Gophers Slack caught my eye regarding the performance of Go's standard library regexp package when handling massive alternations of literal strings. The user posted the following observation: "I've been looking at why a certain regexp is slow, and I find that anything with over 500 instructions (maxBacktrackProg) gets run by the NFA engine. However every literal character in the regex consumes an instruction, so this penalizes simple regexps which contain a bunch of long strings." The "Monster" Regex The user provided an obfusctaed version of the regular expression causing the bottleneck. It is essentially a massive list of pipe-separated literals, wrapped in a dot-star sandwich to search for substrings: ^(?s:.*zQPbMkNO.*|.*NNSPdvMi.*|.*iWuuSoAl.*|...many more...|.*wYtnhdYb.*)$ Because the standard library counts every literal character as an instruction, this regex exceeds the threshold for the efficient One-Pass...
Post a Comment
Read more

Taming the Flat AST: Ergonomics in the Age of Zero Allocations

-
Image
Last week, I introduced egg , a modern LL(1) parser generator for Go. The response from the community was insightful, particularly regarding egg 's most distinct feature: the flat AST. To achieve high performance and low garbage collection pressure, egg encodes the entire Abstract Syntax Tree (AST) into a single []int32 slice. While this is excellent for the CPU cache, it can be intimidating for the developer. One commenter on the Gophers Slack put it perfectly: "The 'Traversing the AST' section really dissuades me from using this tool... I think coming up with a clean, well-typed API might be harder than I think." They were right. In my previous post, I showed a raw recursive walker that involved manual index arithmetic. It was fast, but it wasn't pretty. However, shortly after that post, I implemented ImportEBNF —a feature allowing egg to bootstrap itself by reading EBNF from HTML files. In doing so, I had to "eat my own dog food"...
Post a Comment
Read more

Building a High-Performance JSON Parser in Go with Egg

-
Image
Egg: A Modern LL(1) Parser Generator for Go Parsing is a staple of computer science, yet finding the right tool for the job in Go often feels like a choice between heavy, complex frameworks or writing a brittle parser by hand. Enter egg (Expression Grammar Generator). egg is a new parser generator that prioritizes simplicity, performance, and Go-native conventions. It takes an EBNF grammar—very similar to the one used in the Go language specification—and generates a fast, dependency-light recursive descent parser. In this post, we’ll build a fully functional JSON parser from scratch to demonstrate how egg works, and we'll dive deep into its most controversial and powerful feature: its flat AST encoding. The Grammar: Familiar and Concise One of egg 's strengths is that it doesn't invent a complex new syntax. If you've read the Go language spec, you already know how to write an egg grammar. It combines lexical definitions (regex and literals) and syntax de...
Post a Comment
Read more

ccgo/v4 experiment: Trying the new runtime.Pinner

-
tl;dr: Looking forward future Pinner.Pin  performance improvements. The upcoming Go version 1.21, scheduled for release next month, is currently available for download as Go 1.21rc2 in the "Unstable version" section here . Go 1.21 introduces a new runtime type, Pinner . ccgo/v4, the next, also not yet released version of the C to Go transpiler, uses pinning to "freeze" addresses of local  Go variables, addresses of which are passed around in the original C code. ccgo produces Go code where any C pointer points to memory not managed by the Go runtime. So ccgo simply puts such "escaping" variables in the memory not visible to the garbage collector, with stable, immovable addresses. Those are provided by the modernc.org/memory package. Otherwise a goroutine stack resizing can change the address of a local variable. Another problem ccgo has to solve are the runtime pointer validity checks. I'm not aware of the details being documented somewhere outside of...
Post a Comment
Read more

Producing a Go scanner in 1,219 bytes of code

-
modernc.org/rec is a regexp to Go code compiler tool. It is still a bit rough around the edges. For example, it converts the regexps to a DFA, but it does not yet support intersecting character classes ending up in the same DFA state. Anyway, rec can already handle some nontrivial tasks, like generating a usable, working Go scanner. Here are those 1,219 bytes - in scanner.sh . The shell script is used in the generate target of the Makefile . Note the Perl Unicode character classes in the regexp for Go indentifiers. The respective EBNF  lexical grammar   part of Go specification is identifier = letter { letter | unicode_digit } . The above production expands and translates to the regexp (\pL|_)(\pL|_|\p{Nd})* .  As mentioned above, constructing DFAs from regexps using character classes is a bit challenging per se when considering Unicode. A similar program, lx(1) , part of libfsm , seems to not support Unicode so far, possibly because facing similar difficulties. The...
Post a Comment
Read more