Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!udel!princeton!njin!uupsi!sunic!news.funet.fi!funic!santra!news
From: gson@niksula.hut.fi (Andreas Gustafsson)
Newsgroups: comp.sys.handhelds
Subject: Re: Sass vs ASAP
Keywords: ASAP Sass CHIP-48
Message-ID: <1990Nov16.105051.20653@santra.uucp>
Date: 16 Nov 90 10:50:51 GMT
References: <1442@fs1.ee.ubc.ca>
Sender: news@santra.uucp (Cnews - USENET news system)
Reply-To: gson@niksula.hut.fi (Andreas Gustafsson)
Organization: Helsinki University of Technology, FINLAND
Lines: 47

In article <1442@fs1.ee.ubc.ca> jmorriso@snickers.ee.ubc.ca writes:
> As far as I know, there are two assemblers for the Saturn processor:
> Sass which I believe was the first, and ASAP which was written later.

Actually they were written roughly at the same time, but Sass was
published a bit before ASAP.

> Most people seem  to be posting code in ASAP form. Although the mnemonics
> are the same, there might be differences in assembler directives.
> Could someone mention why they are using ASAP instead of Sass? Is it better?
> As far as I know it is written in PERL, and runs on  Unix workstations only.
> Sass however is self contained. It is written in yacc and lex, and the c
> source code generated compiles on anything . I know  that it runs on unix
> and the  ibm pc.

I wrote ASAP as a quick hack just to get CHIP-48 assembled somehow,
because there was no other Saturn assembler available at the time.

Sass is certainly smaller, faster and more elegant, but the main
reason I haven't switched over to it is that when I last looked, it
lacked a crucial feature: it doesn't let you calculate an
assembly-time constant by subtracting two labels.  The CHIP-48 source
relies heavily on this construct to calculate the addresses of fixed
tables within a relocateable Code object.  Consider the following
example:

	move.a	pc,a			; get address of next instruction
ref17:	move.p5	hexfont-ref17,c		; Sass can't do this
	add.a	a,c			; calculate address of "hexfont"
	move.a	c,d1			; now d1 points to the hex patterns
	...

hexfont:
	data.5 #F999F
	data.5 #72262
	...

I think with Sass you need to either do the subtraction at runtime
(making the program bigger and slower) or keep the Code object load
address in a register (which could probably be put to better use).

If some later version of Sass supports this construct, I will probably
switch over to Sass.
-- 
Andreas Gustafsson
Internet: gson@niksula.hut.fi
Voice: +358 0 563 5592