Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!apple!usc!elroy.jpl.nasa.gov!lll-winken!sun-barr!newstop!exodus!binkley.Eng.Sun.COM!tomw
From: tomw@binkley.Eng.Sun.COM (Tom Westberg)
Newsgroups: comp.arch
Subject: Re: ~8-job "knee" in response curves on Suns (was Re: IBM RS6000)
Message-ID: <6123@exodus.Eng.Sun.COM>
Date: 16 Jan 91 23:57:00 GMT
References: <amos.663857722@shum> <5257@auspex.auspex.com> <3956@skye.ed.ac.uk> <1991Jan16.231017.2530@csn.org!datran2>
Sender: news@exodus.Eng.Sun.COM
Organization: Sun Microsystems, Mt. View, Ca.
Lines: 36

In article <1991Jan16.231017.2530@csn.org!datran2> smb@csn.org!datran2 writes:
>
>I just tried Richard Tobin's little context switch benchmark on my
>NeXT 68040
>3   : 1460 switches / second
	.
	.
>36  : 1280 switches / second

And I just tried it on my SPARCstation2 (4/75) running a few window systems,
but not much else:

2: 4243 switches / second
3: 4140 switches / second
4: 4022 switches / second
5: 4088 switches / second
6: 4048 switches / second
7: 3881 switches / second
8: 3755 switches / second
9: 3615 switches / second
10: 3612 switches / second
11: 3653 switches / second
12: 3568 switches / second
13: 3548 switches / second
14: 3523 switches / second
15: 3381 switches / second
16: 1966 switches / second
17: 1908 switches / second
18: 1929 switches / second

As you can see, it falls off bigtime at 16. SunOS reserves
one of the contexts (number zero) exclusively for DVMA,
which is why it doesn't fall off at 17.

Tom Westberg
tomw@eng.sun.com