Xref: utzoo comp.sys.amiga:13722 comp.sys.mac:11705
Path: utzoo!utgpu!water!watmath!clyde!rutgers!sri-spam!ames!claris!apple!dwb
From: dwb@apple.UUCP (David W. Berry)
Newsgroups: comp.sys.amiga,comp.sys.mac
Subject: Re: Code Optimization
Message-ID: <7309@apple.UUCP>
Date: 1 Feb 88 22:23:06 GMT
References: <8801281857.AA09606@cory.Berkeley.EDU>
Reply-To: dwb@apple.UUCP (David W. Berry)
Organization: Apple Computer Inc., Cupertino, USA
Lines: 23

In article <8801281857.AA09606@cory.Berkeley.EDU> dillon@CORY.BERKELEY.EDU (Matt Dillon) writes:
>:	In terms of floating point, things are very different here. Floating
>:point operations on the Amiga are not compiled "inline", they are called.
>:If you have a 68000, you can use the standard software emulation (IEEE). If
>:you have the 68020, and have the 68881 or 68882 present (also detected on boot) you can simply replace the math library, and use 68881/68882 instructions
>:directly. No recompiles, no changes, existing software is upward compatabible.
>
>	What do you mean floating point operations are not compiled inline?
>The *real* answer is that you can have it both ways... inline if you don't 
>care about downward compatibility to a 68000, and via shared library calls
>if you want downward compatibility (i.e. 68020 machines would have a different
>library implementing the same functions).
	The same thing is true of the Macintosh if you regard the trap
mechanism as a "shared library" which it really is.  Most of the existing
68020 upgrades for the SE replace the standard traps (which run on a 68000
and even faster on an 020) with direct calls to the 881.  You lose a little
to the extra trap overhead, but you do with shared libraries too.
-- 
	David W. Berry
	dwb@well.uucp                   dwb@Delphi
	dwb@apple.com                   973-5168@408.MaBell
Disclaimer: Apple doesn't even know I have an opinion and certainly
	wouldn't want if they did.