Hacker News Reader

Comment by mruts | Hacker News Reader

A Lisp machine isn't just some hardware with Lisp on top. You can do that today with x86. A Lisp machine is supposed to have support for lisp primitives at a hardware level.

View on HN

The MIT Lisp Machine didn't have lisp primitives in hardware. It had two extra hardware features, it could fetch two 16bit instructions at a time, and a fast multi-way branch instruction. Other than that it looked very much like a modern RISC architecture.

I would say that the main feature of a Lisp Machine is that you do everything from within Lisp.

The architecture the developer saw on a Lisp Machine like a Symbolics was tagged, 36 or 40 bit, mostly stack oriented and had lots of primitive Lisp functionality, type checking /dispatching, some GC support, etc.

That was all done in software on the MIT/LMI/TI machines.

I don't know of any papers on the Symbolics hardware that are not behind the ACM paywall, how much did it really differ from the other Lisp Machines ?

Try https://people.eecs.berkeley.edu/~kubitron/courses/cs252-F00...

https://3e8.org/pub/scheme/doc/lisp-pointers/v1i6/p13-dussud...

If that was the case, why are we talking about lisp machines vs a lisp OS?

What about microcode? I thought the point of this whole computing thing was that we could abstract away the fine details and solve problems on the level they're posed at. Who gives a damn if there's a single atom representing a cons or not?

Lisp is good as a language for humans, and it used to be used in hardware because that was necessary to run lisp code fast. That hasn't been the case for ages.

The stated processor (Zilog/Z80) is not suitable for that and isn’t suitable for emulating such an architecture in real time or instantiating suitable virtual machines.

I find this to be an interesting question. Given a modern, high-performance compiling lisp implementation on a modern CPU, say SBCL on x86_64.

What modification or addition to the CPU (or wider computer architecture) would be most effective in making typical lisp programs run faster?

Who gives a damn if there's a single atom representing a cons or not?

Hipsters?

So it needs an Address register and a Data register?

Or does it need hardware parenthesis support too? ;)

Yup. As soon as I reached “Zilog/Z80” I thought “yeah, uhm, nope”.

It's fine as a toy, but yea I'd prefer a powerhouse chip to make it worthwhile.

What instruction extensions would you have to add to something like RISC-V to accelerate Lisp Machines?

The previous design, RISC-IV (named SPUR), was a Lisp machine.

http://pages.cs.wisc.edu/~markhill/papers/computer86_spur.pd...

The basic instructions (Table 3) are close enough to RISC-V but it also had a few Lisp specific instructions (Table 4). Though the name J Extension for RISC-V might lead you to think it is only about Java or Javascript, the group working on it is also interested in making it as good as possible for Lisp.

Reply on news.ycombinator.com