Path: utzoo!attcan!utgpu!watmath!att!dptg!rutgers!tut.cis.ohio-state.edu!unmvax!bbx!bbxsda!scott
From: scott@bbxsda.UUCP (Scott Amspoker)
Newsgroups: comp.lang.misc
Subject: Re: Fast conversions, another urban myth?
Message-ID: <147@bbxsda.UUCP>
Date: 21 Sep 89 22:51:55 GMT
References: <832@dms.UUCP> <688@UALTAVM.BITNET> <9dAz02zs58y201@amdahl.uts.amdahl.com> <27935@winchester.mips.COM> <136@bbxsda.UUCP> <1444@brazos.Rice.edu>
Reply-To: scott@bbxsda.UUCP (Scott Amspoker)
Organization: Basis International,  Albuquerque, NM
Lines: 22

In article <1444@brazos.Rice.edu> preston@titan.rice.edu (Preston Briggs) writes:
 >In article <136@bbxsda.UUCP> scott@bbxsda.UUCP (Scott Amspoker) writes:
 >
 >>Fortunately, business applications are usually I/O bound so the
 >>decimal math can be simulated in software without a perceived
 >>performance penality (unless you bring in an unethical benchmark).
 >
 >But isn't this what the chip-makers say?
 >If a function is so unimportant that it can be simulated in software
 >without performance penalty, than it seems better not
 >to spend hardware on it.

The keyword in my first posting was "usually".  Sometimes you do get in
a CPU intensive calculation in business applications.  Decimal math is not
the most important feature in the instruction set but it does help.  At the
very least it lets the developers know that the hardware manufacturers
have some inkling of an idea about the needs of the business user.

-- 
Scott Amspoker
Basis International, Albuquerque, NM
(505) 345-5232