Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: $Revision: 1.6.2.16 $; site pbear.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!harvard!think!pbear!peterb
From: peterb@pbear.UUCP
Newsgroups: net.micro
Subject: Re: Re: Z-100 blows away IBM-AT
Message-ID: <7000004@pbear.UUCP>
Date: Thu, 30-May-85 12:29:00 EDT
Article-I.D.: pbear.7000004
Posted: Thu May 30 12:29:00 1985
Date-Received: Sat, 1-Jun-85 11:35:41 EDT
References: <463@bgsuvax.UUCP>
Lines: 27
Nf-ID: #R:bgsuvax:-46300:pbear:7000004:000:738
Nf-From: pbear!peterb    May 30 12:29:00 1985




I just had to throw in my two cents worth:

Using an IBM PC/XT running PC/IX as its OS, the program to find those
numbers who square roots are integers ran with the following times:

Using the 8087 coprocessor:

Real	11.7 (seconds!)
User	4.7  (run time in program)
Sys	6.7  (since context switch unloads 8087 registers)

Using subroutines for sqrt and floor (instead of using 8087 instructions):

Real	54.4
User	45.1
Sys	7.0

This just shows that here you are testing basics against one another, and
the math implementations. Start doing benchmarks in assembler to get the hard
core speed operation out of them, and also lets you have an exact grasp
of what you are comparing.

Peter Barada
{ihnp4!inmet|{harvard|cca}!ima}!pbear!peterb