Message-ID: <3742ED58.4D88E9D@t-online.de>
Date: Wed, 19 May 1999 18:56:56 +0200
Organization: LISA GmbH
X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.5 i686)
X-Accept-Language: en
MIME-Version: 1.0
To: pgcc AT delorie DOT com
Subject: Re: pgcc... Do I really got it ?
References: <002501be9285$af72e200$d94902c1 AT 63970047>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Sender: 0303277050-0001 AT t-online DOT de
From: hpj DOT lisa AT t-online DOT de (Hans-Peter Jansen)
Reply-To: pgcc AT delorie DOT com

Damjan Glad wrote:
> 
> I compiled bzip2 using:
> gcc 2.7.2.3
> gcc 2.8.1
> egcs 1.1.2
> pgcc 1.1.2
> 
> performed compresion tests on 20 meg tar file (10 megs of binaries and 10
> megs of texts)
> 
> running on celeron 333a I got:
> 
> gcc 2.7.2.3 53 seconds
> gcc 2.8.1 53 seconds
> egcs 1.1.2 54 seconds
> pgcc 1.1.3 56 seconds
> 
> ??
> 
> gcc switches were -O3 -funroll-loops -fomit-frame-pointer -m486
> egcs 1.1.2 -O6 -funroll-loops -fomit-frame-pointer -mpentiumpro
> pgcc 1.1.3 -O2 -funroll-loops -fomit-frame-pointer -mpentiumpro
> 
> pgcc 1.1.3 generates wrong code if I use -O[3456] so I didn't test it (I
> didn't try to trace whitch specific -f optimization flag causes this).
> 
> bzip2 compiled using -O2 on egcs was compressing for more than 57 seconds.
> 
> I didn't try to time decompressing...

Well, last time, I did that, I was disappointed about the performance
decrease, too. Only the pervertest options gave a slight increase, but
all egcs/pgcc versions were significantly bigger, than the gcc versions.

Interessingly, most of the time during compressing with bzip2 is
taken by one one routine... I thought a certain time about it,
but got no idea, how to improve it efficiently... 
 
> Are there any other tests? Is it really worth it?
>
> Damjan Glad

Hans-Peter