Comments on: Deriving a Virtual Machine

By: Dave

Dave — Sun, 30 Mar 2008 20:53:04 +0000

After a quick skim of the Igarashi paper, I think the difference is they produce linear Compt terms — and it’s this linearization of the initial tree-structured term that prevents eval from trivializing to the identity.

By: MigMit

MigMit — Tue, 25 Dec 2007 11:12:22 +0000

Correction: “that’s all it does”

By: MigMit

MigMit — Tue, 25 Dec 2007 09:52:04 +0000

Sorry, but… I don’t see the point. Let’s look at our data types:

Isn’t that exactly the same thing (up to renaming of constructors)? Moreover, if we identify this two types, then eval seems to be an identity - OK, it forces almost everything, but that’s all id does. I don’t see, why you call id a “compiler”.

By: Shin

Shin — Tue, 27 Nov 2007 02:00:34 +0000

Christophe, thanks for the comments. It was indeed a fun exercise.

And I’ve corrected the typo. :)

By: Christophe Poucet

Christophe Poucet — Mon, 26 Nov 2007 12:35:32 +0000

Interesting article :) I have found it very informational to have a better understanding how these different techniques work. It’s usually easy to roll an interpreter but going from there to a compiled system is not always evident and your article illustrates the different steps in a crystal-clear fashion.

Thanks!

P.S: I do think you have a small typo the first time you write the following:
appK (Cont1 e2 env k) (Num i) =
eval e2 env (Cont i k)

should be

appK (Cont1 e2 env k) (Num i) =
eval e2 env (Cont2 i k)