@pthariensflame@social.treehouse.systems
Post #3213134
2023-11-03 18:31 UTC
@atax1a@infosec.exchange @thephd@pony.social Yeah, this and many other such things means there is no language at all that lets you “fully control the hardware”, since the lowest-level control you get (ISA and/or programmable microcode) is still a heavily abstracted interface.
And that’s why I’m developing an architecture spec that has linear statically-typed values behind a large number of available references (“forwards”) as the only directly accessible form of memory. Down with (architectural) registers, down with byte-addressed memory, down with transparent pointers, and down with the executable form of an ISA having to be the same as its stable form!
Replies (2)
-
@atax1a@infosec.exchange 2023-11-03 18:32
@pthariensflame@social.treehouse.systems @thephd@pony.social this sounds ideal for a highly optimized stack machine, actually
-
@wizzwizz4@fosstodon.org 2023-11-04 00:35
@pthariensflame@social.treehouse.systems Have you published anything on this ISA yet? That sounds like exactly the kind of thing I want to target to prove a point about programming languages and compilers (namely that we should be trying for general solutions, lest our abstractions become wedded to the implementation details of a particular family of ISAs, leaving us unable to consider alternative designs without massive performance losses).