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