Re: Strength reduction of constant multipliers

sastdr@unx.sas.com (Thomas David Rivers)
Thu, 15 Oct 1992 13:47:08 GMT

          From comp.compilers
Related articles
Strength reduction of constant multipliers cliffc@cs.rice.edu (1992-10-13)
Re: Strength reduction of constant multipliers sastdr@unx.sas.com (1992-10-14)
Re: Strength reduction of constant multipliers sastdr@unx.sas.com (1992-10-15)
Strength reduction of constant multipliers cliffc@cs.rice.edu (1992-10-15)
Re: Strength reduction of constant multipliers davidm@voltaire.rational.com (1992-10-14)
Re: Strength reduction of constant multipliers preston@dawn.cs.rice.edu (1992-10-20)
| List of all articles for this month |
Newsgroups: comp.compilers
From: sastdr@unx.sas.com (Thomas David Rivers)
Organization: SAS Institute Inc.
Date: Thu, 15 Oct 1992 13:47:08 GMT
Keywords: arithmetic, optimize
References: 92-10-057 92-10-063

Norman Diamond (diamond@jit081.enet.dec.com) caught me on this:
> >two adds are faster than faster than a shift by 4.
>
> But produce a different result. Maybe you should settle for just one
> "faster than"? :-)


Oops; well I hope everyone, of course, simply _knows_ what I mean, it
should be "an exercise to the reader" to show that what I meant should
have been what I said! ;-)


[Two adds are _not_ equivalent to a shift by four! However, on the IBM
370; a logical left shift of one, intending to multiply a value by two can
be executed faster by logically adding the value to itself, which is what
I intended to say, but alas, didn't.]


    - Dave Rivers -
--

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.