Re: 10 rules for benchmarking (was Re: Teaching new tricks to an old dog (C++ -->Ada))

[Georg Bauhaus <sb463ba@user1.uni-duisburg.de>]

Marin David Condic wrote:
> Here is an amazing thing to me: We keep getting sucked up into the "C is 
> faster than Ada" troll. We all know Ada is capable of being just as fast 
> as C (and occasionally faster, because the additional semantic content 
> allows for more optimizations), so we get offended by the claim.

We might know that Ada is capable of being just as fast as C,
but others might never know this, unless Ada is given the rank it
deserves now for technical reasons based on evaluation. And evaluation
invariably includes speed, like it or not. The reputation of a
programming language has an influence on its use. The reputation
of a programming language is a social, human fact, not a technical
fact. (A big part of marketing is about establishing a reputation
for a product, and little else. The trick is to make this work.)
So it is important to see how speed comparisions add to the
reputation of languages.

Another social, human fact is that 20 years old arguments are still
influential when it comes to comparing current languages.
Ignoring technical progress helps arguing a lot. One of the
old "arguments" is that Ada compilers are big, expensive, slow,
complicated, and buggy.

So a social, human fact is that many people care for
speed ranking. Rankings are popular. This is important because
it makes the winners popular. It adds to the winners' reputation.
It also adds to the reputation of the loosers.

As long as test results show Ada in an inadequate position,
in particular wrt execution speed, the ranking will add to the
reputation of Ada, properties that have once again been stated in
this and related threads (about C++ as well).
The mentioned matrix test has a scientific appearance. Whether the
results are real or bogus, intentionally or not, doesn't matter much.
But visibility of a high rank does matter. Just submit yourself to
the fantasy that some reputable list of test results shows an Ada
program at number one. _This_ will be a killer application.

You might frown upon this kind of evaluation, but consider:

Situation 1: An Apple employee (program development) at a developer
conference tells his audience, in the style of an serious hint,
"You can always sell speed, so try some fpt optimizations, and be
careful".

Situation 2: Remember the card games where each card is essentially
a small technical data sheet of a car, with a small image?
Player A beats player B if A's maximum rounds per minute
exceeds B's.

 From a normal (formal) life-course point of view,
many of us are speed fans, and stay speed fans. So speed is one
obvious criterion when comparing things, and for sure it is
considered important. In computing, there is no "too fast",
in general...


> I think it is important to note that in an incredibly large percentage 
> of the apps that ever get written, whatever differences might exist 
> (hypothetically, Ada being 2x slower than C - or vice versa) it will 
> NEVER make any difference.

Speed will make a difference in the perception of a quite numerous
population buying CPUs, DVD recorders, high speed graphics cards,
and of course, programs. Games, fast and slow, digitally animated
movies, wave file manipulation programs, search engines. Also, number
plate recognition programs ...


Georg