Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!mit-eddie!genrad!decvax!ucbvax!CORY.BERKELEY.EDU!dillon
From: dillon@CORY.BERKELEY.EDU (Matt Dillon)
Newsgroups: comp.sys.amiga
Subject: Re: 32bit = 16bit x 16bit
Message-ID: <8710121921.AA23591@cory.Berkeley.EDU>
Date: Mon, 12-Oct-87 15:21:03 EDT
Article-I.D.: cory.8710121921.AA23591
Posted: Mon Oct 12 15:21:03 1987
Date-Received: Wed, 14-Oct-87 00:49:27 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Lines: 27


	Aztec does this correctly by using MULS/MULU (thank the lords!)... I
depend on it heavily for my graphics work as well... for speed.  Lattice C
isn't quite as smart... you need to put in some casts, but it still does it.
Unfortunetly, most other compilers do not optimize this.   If:

	long z;
	short x,y;

	z = (long)((short)x * (short)y);

	Doesn't work, there probably is no way to do it without using some
assembly.  BTW, the above applies to 32 bit compilers.  16 bit compilers
will usually barf on it since theoretically the result of the multiplication 
is a short if the arguments are short and the compiler would then extend it to 
a long destroying the upper 16 bits.

	You could try the following for 16 bit compilers, but they have to
be *smart*:

	z = (long)((long)x * (long)y);

						-Matt

P.S. that first (long) cast is not strickly required, but some compilers are 
very stupid when it comes to checking the result variable type when applying 
optimization.