Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!microsoft!brianw
From: brianw@microsoft.UUCP (Brian Willoughby)
Newsgroups: comp.sys.apple
Subject: Re: Beginner in Assembly
Summary: Try Fixed Point math - it's faster
Keywords: Assembly, source code, etc.
Message-ID: <36@microsoft.UUCP>
Date: 14 Jul 89 04:36:17 GMT
References: <8907060856.AA01161@crash.cts.com> <6783@sdcsvax.UCSD.Edu> <3425@portia.Stanford.EDU>
Reply-To: brianw@microsoft.UUCP (Brian Willoughby)
Organization: Microsoft Corp., Redmond WA
Lines: 39

In article <3425@portia.Stanford.EDU> kodali@Portia.Stanford.EDU (VAS KODALI) writes:
>
>..., can anyone tell me how to do
>floating point calculations from assembly?  I'm trying to calculate
>fractions in this one subroutine, and this is a BIG stumbling block.
>I got around the required deadline by jumping into BASIC and having it
>handle the FP calculations, but I view this as a kludge, while it did
>turn out to be an elegant one in the end.
>
>Chan

This may not be the answer you were looking for, but...

Try using fixed point math that is tailored to your need for accuracy.
I usually use 24 bit numbers, and "pretend" that there is a decimal
point after the 16th bit.  Then when you need the whole part of the
number, you just access the most significant 16 bits.  Of course, this
means that you'll have to write custom divide routines which take into
account your particular fixed point notation, but if you don't like
relying in the BASIC routines, then you probably shouldn't use the divide
routine in the Monitor area of the Firmware.

Now I did say that I usually use 24 bits.  At one point, this wasn't
enough, because the rounding errors (in my graphics output) were very
apparent.  So I just increased my notation to 32 bits, 16 whole and 16
fractional part.  Again, I had to rewrite my divide routine, but it was
a good exercise to be able to understand codeing fast 6502 divide
routines anyway.

p.s. The multiply routines also need to be tailored to the notation used,
but I find multiply routines to be an order of magnitude easier than
divide!

p.p.s. If speed is a factor, you'll really like fixed point!

Brian Willoughby
UseNet/UUCP     ...!{sun|uw-beaver|uunet}|microsoft!brianw
Bitnet          microsoft!brianw@Sun.COM
                brianw@microsoft.UUCP