From owner-freebsd-current Tue Aug 25 12:56:23 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id MAA23481 for freebsd-current-outgoing; Tue, 25 Aug 1998 12:56:23 -0700 (PDT) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from pcpsj.pfcs.com (harlan.fred.net [205.252.219.31]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id MAA23395 for ; Tue, 25 Aug 1998 12:55:31 -0700 (PDT) (envelope-from Harlan.Stenn@pfcs.com) Received: from mumps.pfcs.com [192.52.69.11] (HELO mumps.pfcs.com) by pcpsj.pfcs.com (8.8.8/8.8.8) via ESMTP id ; Tue, 25 Aug 1998 15:53:07 -0400 (EDT) Received: from brown.pfcs.com [192.52.69.44] (HELO brown.pfcs.com) by mumps.pfcs.com (8.8.8/8.8.8) via ESMTP id ; Tue, 25 Aug 1998 12:53:03 -0700 (PDT) Received: from localhost [127.0.0.1] (HELO brown.pfcs.com) by brown.pfcs.com (8.8.8/8.8.8) via ESMTP id ; Tue, 25 Aug 1998 15:53:02 -0400 (EDT) X-Mailer: exmh version 2.0.2 2/24/98 To: joelh@gnu.org cc: Harlan.Stenn@pfcs.com, garbanzo@hooked.net, mike@smith.net.au, entropy@compufit.at, wwoods@cybcon.com, freebsd-current@FreeBSD.ORG Subject: Re: gcc 2.8 In-Reply-To: Joel Ray Holveck's (joelh@gnu.org) message dated Tue, 25 Aug 1998 14:18:30. <199808251918.OAA00942@detlev.UUCP> X-Face: "csXK}xnnsH\h_ce`T#|pM]tG,6Xu.{3Rb\]&XJgVyTS'w{E+|-(}n:c(Cc* $cbtusxDP6T)Hr'k&zrwq0.3&~bAI~YJco[r.mE+K|(q]F=ZNXug:s6tyOk{VTqARy0#axm6BWti9C d Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Tue, 25 Aug 1998 15:53:02 -0400 Message-ID: <870.904074782@brown.pfcs.com> From: Harlan Stenn Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I tested two packages. One compares a number (6-8) byte move subroutines (memcpy, bcopy, a variety of Duff's devices (using char, short, and int), and some other "fast" byte copies I've snarfed over the years). I run a reasonable quantity of different size/alignments against each of these, and report the CPU time of each one. The second test is much less (?) useful: I compiled my Mumps implementation with each compiler, and ran 11 coarse-grained tests (for loops, subroutine calls, (string) arithmetic (integer and real number), symbol table stuff, a variety of string operations (catenation, justification, formatting), pattern matching, database ops, and probably a couple of others. I run this test on a "quiet" system, as the test uses wall-clock timing (but each test runs for 30-60 seconds, on average). If I could get TenDRA to produce an executable that can be run under gprof, I'd do that instead. H --- > > I just did another test of performance using TenDRA, comparing it to > > FreeBSD's "cc". > > What was your test methodology? To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message