Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.7.0.10 $; site uicsl
Path: utzoo!watmath!clyde!cbosgd!ihnp4!inuxc!pur-ee!uiucdcs!uicsl!mather
From: mather@uicsl.UUCP
Newsgroups: net.lang.f77
Subject: Re: timing code segments
Message-ID: <12900002@uicsl>
Date: Mon, 11-Nov-85 09:35:00 EST
Article-I.D.: uicsl.12900002
Posted: Mon Nov 11 09:35:00 1985
Date-Received: Wed, 13-Nov-85 07:24:13 EST
References: <2356@watale.UUCP>
Lines: 26
Nf-ID: #R:watale.UUCP:2356:uicsl:12900002:000:863
Nf-From: uicsl.UUCP!mather    Nov 11 08:35:00 1985


Look at dtime(3F) and etime(3F). Briefly:

real function etime(t)
real t(2)
returns elapsed (sys,user) time of the process since beginning.

real function dtime(t)
real t(2)
returns delta time since last call (or beginning if first call)


Both functions return the sum of sys+user times with granuality of 1/60 sec.
the real array contains the system time t(1) and the user time t(2) separated.

Because process timing is very rough, take the results with a grain of salt.
The results are load dependant and so can be confusing. This is because the
times are calculated by sampling the program counter (I was told) peiodically
and seeing where it is and whose program it's in.

Also look into profiling with the f77 -p switch. You can get interesting
statistics about functions/subroutines, etc.
----
				b.c.mather
				Software Surgeon
				uiucdcs!uicsl!mather