Re: I found a FORTH stack optimzer, but could use a better version.

On Jan 6, 5:27 am, m...@[EMAIL PROTECTED]
 (Marcel Hendrix) wrote:
> none <""jan\"@[EMAIL PROTECTED]
(none)"> writes Re: I found a FORTH stack optimzer, but
could use a better version.
>
>
>
> > Marcel Hendrix wrote:
> [..]
> >> The algorithm might be useful for a hardware stack engine (although
> >> I would hope that such chips don't do e.g. "SWAP +" as two separate
> >> operations).
> > Did you perhaps mean "SWAP -" ?   because "SWAP +" should always be
> > reduced to one instruction!
>
> I hope so! Maybe I should have said "the compiler for such an engine
should
> be smart enough to not instruct the chip to do "swap +" when the source
code
> says "SWAP +". Is that clearer?

I can understand how SWAP is written in one place, + is written in
another,
and the compiler ends up inlining them and generating the nonsense
code
"SWAP +" so that the compiler now needs to also eliminate some of the
nonsense code that it generated.  But hopefully the idea is not that
the
programmer is being encouraged to write bad code in the first place.

> I guess an intuitive instruction set was not on the design requirements
> list :-)

Forth was designed to have an intuitive instruction set.  If the
computer
you have choosen does not have an intuitive instruction set then the
job
of the compiler is greater since you have to transform intuitive stack
opcodes in Forth to non-intuitive and perhaps non-obvious code
sequences.
This is sometimes refered to as eliminating the Forth from Forth code.

Best Wishes