Path: chuka.playstation.co.uk!news
From: gil@snsys.com (Gil Jaysmith)
Newsgroups: scea.yaroze.freetalk
Subject: Re: C versus C++
Date: Fri, 09 Jan 1998 11:47:12 GMT
Organization: PlayStation Net Yaroze (SCEE)
Lines: 94
Message-ID: <34b60720.835391@www-s.playstation.co.uk>
References: <34b54e3d.9156136@news.scea.sony.com>
NNTP-Posting-Host: trish.snsys.com
X-Newsreader: Forte Free Agent 1.11/32.235

On Thu, 08 Jan 1998 23:34:09 GMT, matth@bellatlantic.net (Matthew
Hulett) wrote:
>   I suppose what really gets me curious for your opinions is that Gil
>was defending the language up above, and then he said later on in a
>little reply to me, ".And it's still the GNU compiler, but you can use
>C++. If you're mad." Now, c'mon, you can't have it both ways!!!

LOL! Okay, I'll clarify what I mean.

First you have to bear in mind that I amn't a Yaroze or PSX programmer
myself. My perspective is from the SN support and tool-development
desk, and so I tend to get feedback from the pro developers about the
various languages and tools, and try to do something helpful about it.

We write our PC-hosted dev tools in C++, because it aids team
development and compartmentalised design. We have an IDE coming out
shortly which is being worked on by eight people. I strongly doubt
this would be as straightforward as it's proving were it not for C++ -
or rather, for the concept of classes and interfaces. We could do it
in Java for that matter. The important thing is that C doesn't have
any formal mechanisms for making sure your project remains reasonably
streamlined and cohesive while several programmers are working on
different modules. You have to work at it. It's possible, and most
games programmers are familiar enough with C to make it work
adequately. But C++ simply makes it easier.

On the PSX I know of perhaps a dozen companies using C++. The reason
for much of this use is that PC games are being ported to PSX and the
language of choice for PC games developed under Windows is C++.
Microsoft has worked hard making its implementation of C++ fairly
rigorous, extending it and ignoring the delays of the ANSI committee,
perhaps causing trouble for itself later on, but nonetheless supplying
all its customers with a solid C++ compiler now, as opposed to some
years in the future.

However, the GNU compiler everyone uses for C++ on PSX is neither as
stable nor as powerful as VC++. I know it works because there are
people using it. I also know that there are people using it who don't
like being made to use it, and who feel there are speed problems which
are unacceptable in an embedded system. This point of view isn't
restricted to console programming; there's now something called
Embedded C++ which attempts to strip out the unimportant bulk of C++
to make it more efficient for embedded systems of any kind.

The argument on here a while ago about C++ boiled down to: is C++
*always* inefficient and thus simply the wrong choice, or is the
slight loss of efficiency worth the gain to developers in making
projects run smoothly, and in some cases run at all? There are plenty
of cases of games, applications, whatever, on many chips, being
designed and half-implemented in C, and then abandoned because the
program's complexity was running out of control. My view is that C++
will help you control this problem, but you need to display due care
and attention when designing and writing a game in C++, so as not to
fall prey to the inefficiencies which the compiler will insert for you
if you let it. If you don't let the compiler make assumptions which
involve extra temporary variables, lots of class member calls, and
horrible convolved operations to call virtual functions, then you can
get C++ code which runs *almost* as fast as C - and will be more
easily tweaked at points where the C code would become "something we
don't mess with in case it breaks". (I don't believe in having code
which "works, but no-one really understands how". It's an offence
against design and more seriously an offence against common sense; if
you don't know how it works, how can you be *sure* it works? A million
examples aren't enough.)

Some years ago the argument was about the tradeoff between C and
assembly language. To put it bluntly, assembly language lost: C
compilers which had been poor were suddenly capable of out-optimising
human beings. The issue is now whether good C code is ruined when
rewritten as good C++ code. And I think it'd be difficult at first,
and perhaps impossible for some people, but that C++ is certainly a
viable option, and probably the one I'd choose.

All of which might seem like a longwinded way to explain what is
probably now obvious, that my comment "if you're mad" was somewhat
flippant... but I thought you ought to know the background...

>   So please, if you could, why should the aspiring developer  learn
>or ignore  C++? 

You should always learn everything, is my opinion. Or at least you
should learn meta-programming: learning a language in terms of its
concepts, without stressing yourself out too much about the fact that
it's specifically 'C++' or 'Java', will mean that other languages can
be understood that much more readily. The context of a new language
may help you to see how to implement more efficiently in an old one,
or make you suddenly understand why a particular language feature is
regarded as evil by other programmers. (For example, Java is the
logical conclusion of the argument to remove the C preprocessor.)

So now you know...

- Gil