X-POP3-Rcpt: mlehmann AT universe DOT sgh-net DOT de
21 Jan 1998 22:24:51 +0100 (CET)
From: Holger Burbach <holly AT cthulhu DOT rhein-neckar DOT de>
X-Sender: holly AT uggae DOT rhein-neckar DOT de
To: Ronald Wahl <Ronald DOT Wahl AT Informatik DOT TU-Chemnitz DOT DE>
cc: Jack Duan <jduan AT sprintmail DOT com>, Brian Makin <makinb AT ctc DOT com>,
        beastium-list AT Desk DOT nl
Subject: Re: PGCC optimizing AMD K6?
In-Reply-To: <Pine.LNX.3.96.980121190344.748B-100000@goliath.csn.tu-chemnitz.de>
Message-ID: <Pine.LNX.3.96.980121222144.998A-100000@uggae.rhein-neckar.de>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: Marc Lehmann <pcg AT goof DOT com>
Status: RO
Content-Length: 1822
Lines: 45

On Wed, 21 Jan 1998, Ronald Wahl wrote:

> No test w/o -funroll-loops? I've had discovered that it will result in
> slower code. Maybe this has changed with pgcc/egcs-980115 but it would be
> nice to see the results.

Okay, here they are!

K6 200MHz, 512kb 2nd-level Cache, linux-2.1.79, glibc-2.0.5
===========================================================
                      (0)  (1)  (2)  (3)  (4)  (5)
===========================================================
NUMERIC SORT         0.92 0.84 0.84 0.92 0.79 0.90
STRING SORT          0.84 0.83 0.84 0.80 0.82 0.84
BITFIELD             0.72 0.87 0.88 0.72 0.86 0.86
FP EMULATION         0.60 0.87 0.87 0.66 0.85 0.85
FOURIER              1.22 1.21 1.21 1.23 1.22 1.24
ASSIGNMENT           0.98 1.15 1.20 0.99 1.18 1.22
IDEA                 0.94 0.93 0.95 0.90 0.90 0.85
HUFFMAN              0.87 0.90 0.94 0.73 0.78 0.85
NEURAL NET                          0.80 0.92 0.93
LU DECOMPOSITION     0.97 1.35 1.31 0.89 1.34 1.37
===========================================================
MEMORY INDEX         0.84 0.94 0.96 0.83 0.94 0.96
INTEGER INDEX        0.82 0.88 0.90 0.80 0.83 0.86
FLOATING-POINT INDEX 1.06 1.18 1.17 0.96 1.15 1.16
===========================================================

(0) pgcc-980115 (haifa sched) -O6 -mamdk6
(1) pgcc-980115 (haifa sched) -O6 -mamdk6 -funroll-loops
(2) pgcc-980115 (haifa sched) -O6 -mamdk6 -funroll-all-loops
(3) egcs-980115 (haifa sched) -O6 -mpentiumpro
(4) egcs-980115 (haifa sched) -O6 -mpentiumpro -funroll-loops
(5) egcs-980115 (haifa sched) -O6 -mpentiumpro -funroll-all-loops

So, that's enough benchmarking for today ... 

Ciao
  Holly

-- 
Holger Burbach                              <holly AT cthulhu DOT rhein-neckar DOT de>
G.-Hauptmann-Strasse 16/67                  <burbach AT cck DOT uni-kl DOT de>
67663 Kaiserslautern