Path: utzoo!utgpu!water!watmath!clyde!bellcore!faline!thumper!ulysses!andante!mit-eddie!apollo!nelson_p@apollo.uucp
From: nelson_p@apollo.uucp
Newsgroups: comp.sys.ibm.pc
Subject: floating point speedup
Message-ID: <3c623946.44e6@apollo.uucp>
Date: 31 May 88 17:21:00 GMT
Sender: user@apollo.uucp
Lines: 17
Posted: Tue May 31 13:21:00 1988
To: comp.sys.ibm.pc@news


 I have an application that's very floating point-intensive.
 It is written in (M.S. QuickC) C and run on my Leading
 Edge Model 'D' XT clone.  Since I don't have an 8087,
 the code uses the M.S. floating point emulator code.
 (BTW I have V1.01 which fixes the famous precision error 
 in V1.00.)                                          

 My question is this:  How does the speed of the software
 floating point code compare with doing it with an 8087 
 running at the same clock speed as the CPU?  I am especially
 interested in single-precision multiplies and divides.
 I need to decide if an 8087-2 will give me enough of a boost
 to justify its ~ $150 pricetag.   Than you in advance.

                                     --Peter Nelson