Xref: utzoo comp.sys.amiga:49744 comp.sys.amiga.tech:9965 comp.lang.forth:2172
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!zaphod.mps.ohio-state.edu!swrinde!emory!mephisto!mcnc!rti!sas!bts
From: bts@sas.UUCP (Brian T. Schellenberger)
Newsgroups: comp.sys.amiga,comp.sys.amiga.tech,comp.lang.forth
Subject: Re: Thanks...Amiga Forth quest. answered
Message-ID: <1548@sas.UUCP>
Date: 13 Feb 90 03:59:37 GMT
References: <3178@trantor.harris-atd.com> <1990Feb8.154759.21230@wilbur.uucp>
Reply-To: bts@sas.UUCP (Brian T. Schellenberger)
Organization: SAS Institute Inc, Cary NC
Lines: 46

In article <1990Feb8.154759.21230@wilbur.uucp> scott@wilbur.UUCP (Scott Beckstead) writes:
|Forth is extremely fast for an interpreted language but not
|quite  as fast as most C programs.  The difference is admittedly small

This is true of MOST FORTH IMPLEMENTATIONS, but not an inherent property of
the language.  In fact, JForth (which the original artical primarily discussed
is not a threaded interpreted (traditional Forth) at all.  It compiles each
word directly to 68000 assembly (JSR-threaded), *and* it supports in-line
expansion of simple words based on the length of the definitions.

Thus, it produces code just like a C compiler for it is, in fact, a compiler.
It lacks only an optimizer.  In Forth, this not all that serious, though; 
optimizers mostly optimize out common subexpressions (rarely a problem in
Forth; this is a real benefit of the stack approach) and other common code
(also rarely a problem because of the ease of factoring problems sufficiently.

All in all, the difference is likely to be miniscule when comparing C to
JForth.  Certainly, being able to compile in about 1/100 of the time more
than makes up for whatever slight difference there might be.

Now, the *other* point, about the paradigm differences is certainly valid.
Forth requires a different kind of approach to problems.  I think that it
is just plain more fun, and certainly better suited to tasks like graphics
and user interfaces because of the ability to test as you develop and the
ease with which you can re-compile small sections interactively.  On the
other hand, C code is a lot easier to share.  But that's only because it's
more popular, and if you were into popularity, you'd have bought a PC, not
an Amiga anyway.  Right?

(Actually, C *is* better [modern C, anyway, like Lattice] for applications
which can't easily be verified by testing--eg, low-level stuff that must
deal with extreme situations.  With C, you can use prototypes and various
static, compile-time verification to be sure that you are doing what you
thought you were doing.)

(Also, if you need to run on more than one machine, C would be my
recommendation for almost any application; Forth is, alas and alak, a
tower of Babel.  Great fun for your own, one-machine stuff but a nightmare
to port.)


-- 
-- Brian, the Man from Babble-on.		...!mcnc!rti!sas!bts
-- (Brian Schellenberger)
"No one will ever write a song called 'Nitro Burning Funny Cars'"
                 -- THE DEAD MILKMEN, "Nitro Burning Funny Cars"