Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!wuarchive!udel!haven!mimsy!mojo!burgoyne
From: burgoyne@eng.umd.edu (John R. Burgoyne)
Newsgroups: comp.windows.ms
Subject: Re: Floating point for windows??
Message-ID: <1991Jan8.131553.13379@eng.umd.edu>
Date: 8 Jan 91 13:15:53 GMT
References: <1991Jan3.080328.1@acad3a> <1991Jan7.153421.25502@sunee.waterloo.edu> <1991Jan7.164119.10490@news.iastate.edu>
Sender: news@eng.umd.edu (C-News)
Organization: College of Engineering, Maryversity of Uniland, College Park
Lines: 24

In article <1991Jan7.164119.10490@news.iastate.edu> i1neal@exnet.iastate.edu (Neal Rauhauser -- ELT Computer Applications Group) writes:
>
>
>    Does windows use a floating point chip if its present? I have no
>programs that specifically require one, but $300 for a 387sx would
>be much easier to take than upgrading my sx to a full size machine.
>I'd like to here from someone who has done a side by side comparison
>of two identical machines except that one has a 387. I'll post
>results if they're interesting enough
>
>   Neal i1neal@exnet.iastate.edu

This thread went through the group last fall. The bottom line is that
interprocess comm. between the 386 and 387 is not that efficient, but the
386 by itself doesn't do badly for arithmetic. Thus, there is almost
no improvement until you do trig or transcendental math. Some functions
might execute in say 10 clock cycles with a 387 but 80 with a 386 only.
So unless you have a loop of a large number of such equations, I'm
not confident you'll notice.

BYTE or PC Magazine did this test sometime last year. I'll look for
the article date or perhaps someone else already knows.

Robert