Path: utzoo!utgpu!water!watmath!clyde!ima!johnl
From: andrew@frip.gwd.tek.com (Andrew Klossner)
Newsgroups: comp.compilers
Subject: Re: Assemblers can be fast
Message-ID: <9621@tekecs.TEK.COM>
Date: 15 Jan 88 17:10:59 GMT
References: <813@ima.ISC.COM> <833@ima.ISC.COM>
Sender: johnl@ima.ISC.COM
Organization: Tektronix, Wilsonville, Oregon
Lines: 27
Approved: compilers@ima.UUCP

> "This [one-pass assembly with fixup chains] is easy if you
> don't have to worry about instructions that change their size
> depending on the expresison value (eg long, short and medium
> branch instructions).  Many assemblers spend time optimizing
> such branches, and this generally requires an extra pass to do
> it properly.  The gains of optimizing such branches are really
> worthwhile."

Branch length optimization is a win, but you don't need multiple passes
to do it.  I recently wrote a one-pass assembler with various sorts of
span length optimization.  There was nothing particularly clever
involved; I just told myself firmly that I wasn't going to do multiple
passes, and after some thought the necessary algorithms became clear.
It helped, but was not vital, that I had a large virtual memory space
and so didn't need to bother with temporary files.

In general, a good guiding principle is that it's much quicker to
review your generated object code in conjunction with in-memory tables
than it is to make another pass over the source code.

  -=- Andrew Klossner   (decvax!tektronix!tekecs!andrew)       [UUCP]
                        (andrew%tekecs.tek.com@relay.cs.net)   [ARPA]
--
Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU
Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima
Please send responses to the originator of the message -- I cannot forward
mail accidentally sent back to compilers.  Meta-mail to ima!compilers-request