Sender: salvador AT delorie DOT com
Message-ID: <3BCD8E62.24652BA7@inti.gov.ar>
Date: Wed, 17 Oct 2001 10:57:54 -0300
From: salvador <salvador AT inti DOT gov DOT ar>
Organization: INTI
X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.19 i686)
X-Accept-Language: es-AR, en, es
MIME-Version: 1.0
To: djgpp AT delorie DOT com
Subject: Re: MAKEing Turbo Vision 1.1.3
References: <im4mstk71bb1ondv81goeqsededhsiqj0b AT 4ax DOT com> <3BCB274A DOT 34F2631 AT inti DOT gov DOT ar> <j0fnst84g4o75b64184kvs3dhlvg3j5gki AT 4ax DOT com> <3BCC49EC DOT 5FB4D8ED AT inti DOT gov DOT ar> <3995-Tue16Oct2001175743+0200-eliz AT is DOT elta DOT co DOT il> <20011016130041 DOT A19158 AT kendall DOT sfbr DOT org> <ofgpst05mjgdf36je1pdp48r3r6tumjq1m AT 4ax DOT com> <200110162355 DOT f9GNtqt13233 AT envy DOT delorie DOT com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp AT delorie DOT com

DJ Delorie wrote:

> > and THIS is supposed to encourage people to start moving "up" to
> > 3.0 ???
> > >                        gcc 2.7.2.1     gcc 3.0
>
> Well, a couple of things.
>
> The comparison omits gcc 2.8.1 and 2.95, which were important
> releases.
>
> Gcc 3.0 has a lot more functionality than 2.7.2.1.
>
> Bigger code does not mean slower code - it can sometimes be faster
> code by avoiding cache-busting jumps.
>
> The sizes do not indicate code vs data vs overhead (debug etc).
>
> "Hello world" is a bad program to do comparisons with.  You should
> choose a more real-world program, like gzip or make, and time how long
> it takes to build those and how big they are.

I agree with you DJ, but it looks like regular switchs (-O2 for example)
have a bad impact on old CPU models when using gcc 3.x.
I didn't run my battery of benchmarks yet so I can give a detailed
conclusion, but the new C++ library and code generation rules made my editor
20% bigger (we are talking about more than 200 Kb of increase) and 11%
slower.
I know the speed difference could be just because I tested with a K6 and
perhaps a Pentium II isn't affected, but isn't a nice thing. Specially when
I can clearly see the compiling time difference. I didn't meassure the time,
but I know that these differences are really notorious when a factor of at
least 2.5 or 3 is applied.
I don't blame the gcc team and I'm thinking that is time for a hardware
upgrade. I know the ISO C++ 1998 standard introduced a heavy use of
templates in the C++ standard library  and is one of the reasons for the
increase in code size and compilation time. But I also realize that
compiling C++ code won't be possible using gcc 3.x and small machines (386
and 486) and quite annoying with anithing less than Pentium II of 500 MHz
with 128 Mb of memory.
When I finish with TV and setedit releases I'll try to run the
BYTE benchmarks to compare gcc 3.x with all the data I already have.

SET

--
Salvador Eduardo Tropea (SET). (Electronics Engineer)
Visit my home page: http://welcome.to/SetSoft or
http://www.geocities.com/SiliconValley/Vista/6552/
Alternative e-mail: set AT computer DOT org set AT ieee DOT org
Address: Curapaligue 2124, Caseros, 3 de Febrero
Buenos Aires, (1678), ARGENTINA Phone: +(5411) 4759 0013