Date: Wed, 25 Feb 1998 17:26:54 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Rylan <rylan AT intekom DOT co DOT za>
cc: djgpp AT delorie DOT com
Subject: Re: Is PGCC really worth it?
In-Reply-To: <01bd4150$afde1c00$LocalHost@default>
Message-ID: <Pine.SUN.3.91.980225172621.2648O-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On 24 Feb 1998, Rylan wrote:

> I am writing an application that might benefit a lot form intensive
> optimisation. Using -O3 (or -O2 with various switches) does seem to
> help,

How so?  Do you mean that your program's speed doesn't change, no
matter if you do or don't use -O2?  I find this hard to believe.

Did you try the switches and advice mentioned in the DJGPP FAQ list,
section 14.2?

> but I'm thinking that Pentium opti's might be even better.

In my experience, it depends on the structure of your program.  If most of
the time is spent in a small number of hot spots which are tight loops,
then Pentium-specific optimizations might help, perhaps as much as 25% or
even more.  But if the CPU usage histogram is flat, you are unlikely to
gain much by using PGCC. 

In any case, I would suggest profiling your program first, to see
where does it spend most of the time.  See chapter 13 of the FAQ for
more details.