NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/sys/sun/admin/6320

Internet Message Format | 1992-09-14 | 1.2 KB

Xref: sparky comp.sys.sun.admin:6320 comp.sys.sun.hardware:4383 comp.sys.sun.misc:4218
Path: sparky!uunet!cs.utexas.edu!news
From: [e-mail] (King-Yeung Yau)
Newsgroups: comp.sys.sun.admin,comp.sys.sun.hardware,comp.sys.sun.misc
Subject: Fine grained performance analysis on SunOS
Followup-To: comp.sys.sun.admin
Date: 14 Sep 1992 17:06:37 -0500
Organization: CS Dept, University of Texas at Austin
Lines: 14
Message-ID: <[e-mail]>
NNTP-Posting-Host: levelland.cs.utexas.edu

Has anyone tried to perform some really fine grained timing analysis
on SunOS, like context switch time? Specifically, I'm running SunOS 4.1.1
on a SPARCstation 1+. Since the performance figure is likely to be a few
hundred microseconds, I doubt very much if I can trust the microsecond
precision of any sequence of instructions that read the system clock.
On the other hand, it seems difficult to perform 100 context switches on
end while locking out other system activites that affect the time.
Would appreciate it if someone could offer any advice.

Please respond by email to [e-mail]. I'll summarize if there's
enough interest and good answers. Thanks a lot,

David