 I've had a bit of time to kill between events and shows and things this week, and each time I had a few minutes to myself, I'd pour over the CADR lisp machine papers(528, 628, etc..) It seems clear now that implementing lisps in Rust and C doesn't serve any purpose whatsoever. Build-Your-Own-Lisp type books should really be focusing on teaching people to write for LISP machine architectures instead of software implementation. 11 comments
@zzz those are great, you can see a collection of them here: https://plover.com/~mjd/misc/hbaker-archive/ A reading list? You mean you'd like to have a complement readings to the MIT lab memos? I can put together a sort of crash course on lisp machine architecture this summer. There are so many papers published over the years. Something that's just here is the 20 papers you should read on lisp machine architecture / modern architecture would be awesome @neauoire it also makes me appreciate a time when you could design a computer architecture from the chips up, targeting a particular language and its memory and instruction needs. @neauoire IMHO the same goes for implementing your own Forth in a high level language. The beauty of Forth is in the bootstrapping: you can make one from nothing, and use it as the basis for making the rest of your stack. @neauoire that took a little thinking to work through, but I agree. I take it to mean focus on eg a stack machine, instead of creating C spaghetti for CONS cells etc  @cblgh a computer chips based around lisp's primitives |
Gregory's Server
@neauoire
I'd love if a reading list like that existed. I might ask around for something.
In the meantime, I'd check out some of Henry Baker's old papers
Like how to use stack allocation in a lisp: https://web.archive.org/web/20191231120254/http://home.pipeline.com/~hbaker1/LazyAlloc.html
And this one on what's a minimal core language that you can reliably make a simple, efficient, portable system: https://web.archive.org/web/20191008085100/http://home.pipeline.com/~hbaker1/CritLisp.html