Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!rutgers!cbmvax!fred
From: fred@cbmvax.UUCP (Fred Bowen)
Newsgroups: comp.sys.cbm
Subject: Re: C Power 128 Float to Int conversion clarified
Message-ID: <2717@cbmvax.UUCP>
Date: Tue, 10-Nov-87 13:55:06 EST
Article-I.D.: cbmvax.2717
Posted: Tue Nov 10 13:55:06 1987
Date-Received: Fri, 13-Nov-87 00:06:31 EST
References: <5519@rutgers.rutgers.edu>
Organization: Commodore Technology, West Chester, PA
Lines: 40

In article <5519@rutgers.rutgers.edu>, prindle@NADC.ARPA writes:
> From: prindle@nadc.arpa (Frank Prindle)
> 
> [...]  It is the routine to convert floating point to
> integers (jump table address $AF00) that does not return it's result in A,Y.
> Additionally, *as I said before*, that routine does not work correctly unless
> the FAC has had any fractional part removed by INT first.  I don't know if
> either of these has been changed by the upgrade ROM, but my fix makes it work
> on any 128 (in theory!).  It is simply not sufficient to fix the A,Y register
> problem, you also have to do the INT fix.

The routine is called AYINT (C128 jump table $AF00).  It returns its result
in FACMO (msb: $66 in C128, $64 in C64) & FACLO (lsb: $67 in C128, $65 in C64).

The confusion here involves the use of three very similar routines- AYINT,
QINT, and INT.  Each does basically the same thing- INT(FAC), but with a few
differences.  Actually, both AYINT and INT call QINT.  They all have left
the result in the FAC somewhere since the beginning (same applies to the 8-bit
Apples, too:-)  

AYINT provides some range checking (+32767 to -32768), then jumps to QINT,
which leaves the result in FACMO & FACLO.  INT calls QINT to remove any
fractional part, then refloats the result in the FAC.  QINT is the
"Quick greatest INTeger" routine itself, leaving the result in FACMO & FACLO.
I suspect the problem with C-power is some confusion between INT and QINT,
from what Frank Prindle says.

There is a difference between the C128 math pack and all earlier versions
(PET, Apple, C64, etc.), but this fix is found in even the original C128
ROMs.  INT now *rounds* the FAC before calling QINT.  This fixes the Microsoft
math bug where INT(.9 + .1) yeilds ZERO.  Try it on your Apple, C64, and C128.

Can I go home now?
--
-- 
Fred Bowen			uucp:	{ihnp4|rutgers|caip}!cbmvax!fred
				arpa:	cbmvax!fred@RUTGERS.EDU
				tele:	215 431-9100

Commodore Electronics, Ltd.,  1200 Wilson Drive,  West Chester,  PA,  19380