Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!ut-sally!brian
From: brian@ut-sally.UUCP (Brian H. Powell)
Newsgroups: comp.sys.mac
Subject: Re: Turbo vs. Lightspeed
Message-ID: <8879@ut-sally.UUCP>
Date: Fri, 28-Aug-87 12:03:55 EDT
Article-I.D.: ut-sally.8879
Posted: Fri Aug 28 12:03:55 1987
Date-Received: Sun, 30-Aug-87 01:05:36 EDT
References: <AVB6Vny00V4CEI40Es@andrew.cmu.edu>
Organization: U. Texas CS Dept., Austin, Texas
Lines: 22


     I'm (also) fairly certain Turbo C doesn't exist on the Mac.  If it did,
and if it were as good as the IBM-PC version, it would be the best
code-generator around.  The IBM version made some awfully big ripples (nay,
tsunamis) in the C pond in which Microsoft was sitting.  I don't like LS-C's
code generation (mainly just the complete lack of optimization) at all.  I
could probably do a lot of the optimization myself in C if I really wanted to,
but I don't.

     (The main optimization I want is common subexpression elimination:

	x = ((*current_textedit_hdl)->viewRect.bottom -
	     (*current_textedit_hdl)->viewRect.top) /
	    (*current_textedit_hdl)->lineHeight;

     The handle gets dereferenced three times in the machine code also. (Yuck)
I usually declare the handle as register, and that helps quite a bit, but it
would be better to declare a TEPtr and dereference it myself.  I'd rather the
compiler do that.  I guess there could be some debate, since the user knows
better than the compiler about when handles might change.)

Brian