Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 beta 3/9/83; site basser.SUN
Path: utzoo!watmath!clyde!floyd!harpo!ulysses!unc!mcnc!decvax!mulga!munnari!basser!andrewt
From: andrewt@basser.SUN (Andrew Taylor)
Newsgroups: net.lang.prolog
Subject: Re: Why I think partitions are faster than tagging.
Message-ID: <256@basser.SUN>
Date: Mon, 2-Apr-84 21:54:10 EST
Article-I.D.: basser.256
Posted: Mon Apr  2 21:54:10 1984
Date-Received: Sat, 7-Apr-84 02:28:44 EST
References: <4096@edai.UUCP>
Organization: Dept of C.S., University of Sydney
Lines: 33

I can not agree that a partitioned scheme is faster than tagging.
edai!ok claims that the loop for de-referencing a variable
is faster in a partioned scheme:
{
	I can get this loop down to 4 instructions using C-Prolog's data
	representation, but with *byte* tags I can only get it down to
	5 whichever end I put the tag.  The VAX code is 11 bytes -vs- 19.
	The number of instructions would be the same on an M68000.
}

I use a tagged scheme. The compiled VAX code for the aforesaid loop
involves only 3 instructions, 8 bytes. It lokes like this:
	
		jbr	L1
	L2:	movl	4(r11),r11
	L1:	tstl	(r11)
		jeql	L2

A tagged scheme has other advantages. The unification routine can be
table-driven. This is, I think, faster and it certainly is much cleaner.

A partition scheme (or at least the Cprolog version thereof) involves
horrible, heavily machine dependent, manipulations so that the bit patterns
for integers, floats and pointers can be distinguished.

Using a tagged scheme the various storage areas (stacks) can be small
when created and "grown" to what ever size is necessary. This is harder
to do in a partitioned scheme. The C-prolog approach of initially allocating
~1 megabyte is totally impractical in our environment. We have a swapping
(non-virtual memory) system with 4Mb of memory and sometimes 20-30 students
using prolog.

				Andrew Taylor