Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!columbia!rutgers!husc6!linus!philabs!ttidca!woodside
From: woodside@ttidca.TTI.COM (George Woodside)
Newsgroups: comp.sys.atari.st
Subject: Re: Mark Williams ver 2.0 info
Message-ID: <692@ttidca.TTI.COM>
Date: Thu, 7-May-87 10:46:18 EDT
Article-I.D.: ttidca.692
Posted: Thu May  7 10:46:18 1987
Date-Received: Sat, 9-May-87 18:05:18 EDT
References: <8704302259.AA00661@ucbvax.Berkeley.EDU> <4722@utcsri.UUCP>
Reply-To: woodside@ttidcb.UUCP (George Woodside)
Distribution: world
Organization: Citicorp/TTI, Santa Monica
Lines: 34


I have the MWC 2.0 update (to my original 1.1 compiler). While I can't
answer all your questions, here's what I can offer:

Inline Assembly - Still not supported.

Optimizer - Documentation says it is improved. Having looked at the code,
            I'd say it could use a bit more improving. It will still write
            the same instruction two (or even three) times consecutively.

Documentatioin - Buy the upgrade just for it, even if nothing else is of 
                 interest to you. It is now the most thorough documentation
                 on GEM and VDI anywhere.

Libraries - Again, the documentation says it is improved. Object modules
            are a little bit smaller, execution is about 15% - 20% faster.

The compiler itself is a bit faster. It comes with a vastly improved shell
(which I still don't use) and, at last, compatibility with Beckemeyer's
C-Shell (which I use almost exclusively). It also includes a RAMdisk, in
source. You are supposed to send in the page from your old book with the
copyright notice on it, and a fee of $50 plus $5 S&H in the US (I don't
remember the S&H for international customers). I think anyone who likes
their old version should upgrade, there is plenty of reason to.

The change for Beckemeyer shell compatibility requires that you have
C-Shell 2.0, which changed the way C-Shell saves and parses the PATH
environment variable.

Known bugs - there is a glitch in the assembler which causes the value of
generated long constants to have their high order byte zeroed (ie. 0x87654321L
yields 0x00654321L in the object. Printf will still not write any output until
it gets a line feed (this is not a bug as far as MW is concerned, but a design
"feature").