Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uunet!zephyr.ens.tek.com!uw-beaver!ubc-cs!alberta!ncc!idacom!andrew
From: andrew@idacom.uucp (Andrew Scott)
Newsgroups: comp.lang.forth
Subject: Re: Vectored Execution
Message-ID: <1990Aug28.160157.3938@idacom.uucp>
Date: 28 Aug 90 16:01:57 GMT
References: <12398@sdcc6.ucsd.edu>
Organization: IDACOM, a division of Hewlett-Packard
Lines: 35

John Wavrik writes:
> There is a good reason for wanting words (like 
> DEFER and IS) for vectored execution to be part of the language 
> standard rather than user defined. Here are some timings for 100,000 
> executions of a NOOP -- both directly and using various approaches to 
> vectored execution: 

[ timing figures deleted ]
                        
> In F83 the mechanism for vectored execution is coded in assembly 
> language -- and the overhead for executing a word indirectly is 
> surprisingly small.


I don't think execution speed of a word in a certain implementation is a
compelling reason to include it in the standard.  I think that optimization
is a job best left for the compiler and not the language designer.

For the same reason, I think words like 1+ and 1- should at least be moved
from CORE to EXT CORE.  They unnecessarily clutter the language.  I realize
they have existed for many years and are noticably faster in *some* Forth
implementations.  I think it's a wart to make addition or subtraction by
1 a special case.  It's another wrinkle a newcomer to Forth has to learn.

Anyway, the EXT CORE word PERFORM is useful for vectored execution:

VARIABLE ACTION            ( our execution vector )
' SOMETHING ACTION !       ( assign something to the vector )
ACTION PERFORM             ( execute the vector )

I prefer this to the DEFER and IS solution because of the prefix-postfix
reasons we've been discussing for a while.
-- 
Andrew Scott	| mail:		andrew@idacom.uucp
		| - or -	.!uunet!ubc-cs!alberta!idacom!andrew