Actually I am very impressed from the FORTH simplicity and efficiency. I was developing one FORTH system for DOS a couple of years ago and I think it is very useful for small programs that have to perform low level/hardware tasks. Unfortunatelly it doesn't scale well for larger applications. I don't know how well it can be used as G-machine interpreter but its execution model has a lot of common with STG machine. In STG each closure has a header word which is a pointer to its entry code. In FORTH each procedure has a header word which also points to some entry code. The entry code receives the address of the entered procedure in special (usually EBX on x86) register just like in STG (where the used register is ESI). The execution of any FORTH program is a sequence of jumps from one entry code to another just like in STG. Of course STG is much more complicated because the closures are dynamically allocated and there is a garbage collector too. Thanks to its simple execution model I still think that FORTH is the fastest interpreted language, just like STG is the fastest virtual machine for Haskell. Cheers, Krasimir On 6/1/05, Andrew Harris wrote:

Hi -

Brace yourself... I work in an environment where FORTH is still used.

I've been thinking about writing a G-machine interpreter in FORTH so that one could write Haskell like programs that would compile down and run "graph-reduction" style on the FORTH machine.

Many developers think FORTH is nice, but the language is so, shall we say, "terse".

I'm curious about what people think about this; having the expressiveness of a Haskell-like language that compiles to this environment might provide the best of both worlds, simple hardware architecture and an advanced programming language.

let me know what you think, -andrew _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Andrew Harris wrote:

I've been thinking about writing a G-machine interpreter in FORTH so that one could write Haskell like programs that would compile down and run "graph-reduction" style on the FORTH machine.

Many developers think FORTH is nice, but the language is so, shall we say, "terse".

I'm curious about what people think about this; having the expressiveness of a Haskell-like language that compiles to this environment might provide the best of both worlds, simple hardware architecture and an advanced programming language.

IMHO Koopman's "TIGRE" (http://www.ece.cmu.edu/~koopman/tigre/) might be a close approximation of what you want. I had some (rather non-serious) experience with Forth, and I had great pleasure reading Koopman's paper on TIGRE. Unfortunately I couldn't find any further development on this topic (although I saw some later Koopman's publications on Forth itself). I guess, only someone "thinking Forth" (as in the Brodie's book) might develop such a thing. Dimitry Golubovsky Middletown, CT