From owner-freebsd-hackers Tue Aug 20 06:12:13 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id GAA27583 for hackers-outgoing; Tue, 20 Aug 1996 06:12:13 -0700 (PDT) Received: from Campino.Informatik.RWTH-Aachen.DE (campino.Informatik.RWTH-Aachen.DE [137.226.225.2]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id GAA27575 for ; Tue, 20 Aug 1996 06:12:09 -0700 (PDT) Received: from gilberto.physik.rwth-aachen.de (gilberto.physik.rwth-aachen.de [137.226.31.2]) by Campino.Informatik.RWTH-Aachen.DE (RBI-Z-5/8.6.12) with ESMTP id PAA06594 for ; Tue, 20 Aug 1996 15:07:31 +0200 Received: (from kuku@localhost) by gilberto.physik.rwth-aachen.de (8.6.11/8.6.9) id PAA01722 for freebsd-hackers@freefall.cdrom.com; Tue, 20 Aug 1996 15:20:25 +0200 Date: Tue, 20 Aug 1996 15:20:25 +0200 From: Christoph Kukulies Message-Id: <199608201320.PAA01722@gilberto.physik.rwth-aachen.de> To: freebsd-hackers@freefall.FreeBSD.org Subject: max math performance - how? Sender: owner-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk I digged out a scientific test I compiled some time back in 386bsd times. At that time I hacked around with different libms and I had a binary lying around (statically linked) which outperforms every newly linked version of that benchmark. (It is a bunch of fortran programs, the so called 'Lund' benchmark from Lund University - a program that physicists are mainly interested in seeing perform fast). Now I have no idea what I did at that time - maybe I took some early verrsion of the libmsun or was there something different in 386bsd days (Bruce?) Anyway, with g77 or f77 (f2c) I get time lundtst 5.085u 0.015s 0:07.66 66.4% 545+402k 0+0io 0pf+0w time lundtst.withnormal.libm 7.032u .... time lundtst.static.386bsd 4.184u 0.023s 0:05.61 74.8% 454+422k 0+0io 0pf+0w That means with some unknown old method I'm getting around 20% better performance. Now give me back that mathlib :-) --Chris Christoph P. U. Kukulies kuku@gil.physik.rwth-aachen.de