Re: A89: Shifting/Multiplying. . .


[Prev][Next][Index][Thread]

Re: A89: Shifting/Multiplying. . .




I think this would be a good place for a MACRO...

Mark Leverentz
----- Original Message -----
From: Cody R Maggard <ravenloft_15@juno.com>
To: <assembly-89@lists.ticalc.org>
Sent: Monday, September 13, 1999 6:36 PM
Subject: Re: A89: Shifting/Multiplying. . .


>
> I think that mulu would be a bit faster.  Give or take a byte, each
> instruction is three bytes long.  Seeing that the processor is 32-bit, it
> doesn't really matter in performance speed, but if you have many
> instances of it, go with the mulu.  You'll find that your program will
> run faster, since every time you want to call the subroutine, you need to
> tack on a few more bytes (to save program counter, push arguments, etc.).
>  But I'm just a beginner.  Zoltan, insert your words of wisdom. =)
>
> --------------------------------------------------------------------------
-------
> Cody Maggard, Technomancer
> RavenLoft_15@juno.com
>
> On Mon, 13 Sep 1999 19:06:03 -0400 "Scott Noveck" <noveck@pluto.njcc.com>
> writes:
> >
> >Another optimization (note: "optomization" is spelled with a _z_ =)
> >question:  If I'm multiplying one variable number by a constant that
> >is the
> >sum of several powers of two, is it better to flat-out multiply or to
> >perform several left shifts (multiplying by the powers of two) and
> >adding
> >those numbers?
> >
> >For example:  I want to multiply an arbitrary variable value contained
> >in d0
> >by 12.  Ten is the sum of 2 powers of two: 8 (2^3) and 4 (2^2).  I can
> >use a
> >simple "mulu.w #12,d0" or I can use the following little algorithm:
> >
> >mul_d0_12:
> > move.w d0,d1    ;arbitrary value is a word
> > lsl.w #2,d0     ;d0 = (4 * d0)
> > lsl.w #3,d1     ;d1 = (8 * d0)
> > add.w d1,d0     ;d0 = ((8 * d0) + (4 * d0))
> >                 ;   = (d0 * (8 + 4))
> >                 ;   = (d0 * 12)
> >                 ;[via distributive property of multiplication]
> >
> >I'm pretty sure that the algorith is a good deal faster and not too
> >much
> >larger, although Motorola never did send me my manual so I could tally
> >this
> >stuff up myself (Olle, how do I request it again?).  If somone could
> >just
> >give me a count of the size/speed of the algorith compared to the
> >multiplication (or even show me a better way to do it -- Zoltan =) I'd
> >appreciate it.
> >
> >Also, to what point is this more efficient?  If I were to expand upon
> >this
> >to multiply by 31 -- 2^4 + 2^3 + 2^2 + 2^1 + 2^0 (w/ 2^0 not requiring
> >any
> >shifting), would it still be more efficient than multiplying?  It's
> >beginning to seem to me like the 68k's multiplication/division
> >operations
> >are more or less wasteful. . .
> >
> >    -Scott
> >
> >
>
> ___________________________________________________________________
> Get the Internet just the way you want it.
> Free software, free e-mail, and free Internet access for a month!
> Try Juno Web: http://dl.www.juno.com/dynoget/tagj.
>
>


Follow-Ups: References: