X-pop3-spooler: POP3MAIL 2.1.0 b 4 980420 -bs-
Date: Mon, 20 Jul 1998 18:16:46 +0200 (MESZ)
From: Philippe Elbaz-Vincent <mat910 AT sp2 DOT power DOT uni-essen DOT de>
X-Sender: mat910 AT spo109
To: beastium-list AT Desk DOT nl
Subject:  speed PGCC vs GCC for DIEP
Message-ID: <Pine.A32.3.93.980720175944.98124I-100000@spo109>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: Marc Lehmann <pcg AT goof DOT com>
Status: RO
Content-Length: 2890
Lines: 86

Try the following flags:

-O2  -malign-double  -funroll-all-loops  
-malign-functions=2 -malign-jumps=2 -mamdk6


or test directly

-O6  -malign-double  -funroll-all-loops 
-malign-functions=2 -malign-jumps=2 -mamdk6

NB: occasionnally, -malign-loops=2 give some improvement, but sometimes
slow down performance

if you begin with -O2 and this gives you some improvement,then try to test
the different specifics flags for the  gcc Pentium (that you can find in
the FAQ)

you can also try to play with the following flags, which could be
relevants depending of your code:

-fexpensive-optimizations

-fomit-frame-pointer

and, could be

-ffast-math (as you use mainly integers, it should be safe, but don't
expect 'huge' improvement if it works) 

---- benchmark ----

using the above flags, I get the following benchmark, the best that I have
tested

harware config: PII/266, Asus PL297 motherboard, 64MB SDRAM 10ns

OS: Linux 2.0.33

Distrib: SuSe 5.2

Load average (from procinfo): 0.03 0.06 0.28 2/61 20760

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          121.44  :       3.11  :       1.02
STRING SORT         :           10.87  :       4.86  :       0.75
BITFIELD            :      3.9693e+07  :       6.81  :       1.42
FP EMULATION        :          11.107  :       5.33  :       1.23
FOURIER             :          2209.2  :       2.51  :       1.41
ASSIGNMENT          :          1.7355  :       6.60  :       1.71
IDEA                :          287.77  :       4.40  :       1.31
HUFFMAN             :          125.85  :       3.49  :       1.11
NEURAL NET          :          3.0864  :       4.96  :       2.09
LU DECOMPOSITION    :          89.968  :       4.66  :       3.37
==========================ORIGINAL BYTEMARK
RESULTS==========================
INTEGER INDEX       : 4.764
FLOATING-POINT INDEX: 3.872
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA
BELOW===============================
C compiler          : gcc version pgcc-2.91.47 19980707 (gcc2 ss-980609
experime
ntal)
libc                : libc.so.5.4.44
MEMORY INDEX        : 1.223
INTEGER INDEX       : 1.163
FLOATING-POINT INDEX: 2.148
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3,
libc-5.4.38
* Trademarks are property of their respective holder.


Hope this could be helpful, Ph.

 
Ph. Elbaz-Vincent, Marie-Curie Fellow,
DFG-Forschergruppe "Arithmetic and Geometry", FB 6, Mathematik,
Universitaet GH Essen, D-45117 Essen | Email:elbaz-vincent AT uni-essen DOT de 
Phone: +49 201 183 2198 | Fax: +49 201 183 2426