libera/#lisp - IRC Chatlog
Search
20:32:45
aeth
But some features might not have even been implemented at the ASM level... it might have been below that, in microcode.
20:36:02
aeth
I wouldn't be surprised if the output of a Lisp Machine Lisp compiler looked a lot closer to CLISP's bytecode disassemble than SBCL's x86 one.
20:39:44
aeth
You definitely had some form of memory addressing because they had C compilers, though.
21:04:33
White_Flame
although there's no reason that modern machines can't boot into Lisp, either, like mezzano
21:05:20
aap
though i don't know anything about symbolix. i've only looked into cons and cadr and haven't even looked into what the microcode does very much
21:06:39
White_Flame
but things like + are still general primitives that can do many numeric types, some with traps to software implementations
21:06:43
aeth
oh, good, archive.org crawls bitsavers (although maybe it hasn't gotten to every PDF). https://web.archive.org/web/20220601000000*/http://bitsavers.trailing-edge.com/pdf/
21:08:21
aap
i recently implemented a good chunk of the CONS cpu in verilog but unfortunately the schematics are not quite public (yet) and also it's a bit hard to piece things together
21:08:51
aeth
White_Flame: well, yeah, but it's like those CPUs that are specifically designed to run Java bytecode from the JVM (possibly even involving some of the same people).
21:09:12
aeth
rather than x86-64 asm where everything's very verbose and low-level and any kind of high-level asm shortcut is probably there for legacy reasons and discouraged
21:16:18
White_Flame
or does it all still end up in serial execution timing for the various low level bits it does?
21:17:36
aap
i'm not familiar with what the microcode actually does, but i don't see how it could do that. the cpu has dispatch memory which can be indexed by certain bits of...some quantities
21:18:07
White_Flame
and of course the thing that most resign to is that since modern hardware can do such high parallel IPC, that the various checks can execute "parallel enough" that it still isn't worth trying to do such things in built-in parallel hardware to try to catch up
4:57:01
jcowan
sonny: yes, lisp machines (both East and West Coast versions) didn't do virtual memory.