Value of microeffiency (was: Re: Optimal ...)

[T. William Wells]

In article <3401 at rpp386.UUCP>, jfh at rpp386.UUCP (John F. Haugh II) writes:
> i don't know what value microefficiency has this week, but in general,
> writing good solid algorithms is what is important.

Just so you know: around here, what you are calling
microefficiency makes the difference between staying in business
or not. There is no argument against the idea that a good
algorithm is the proper base from which to start (though there is
certainly room for discussion as to what constitutes a good
algorithm). But, a good algorithm is ONLY THE BEGINNING.

Once you have an algorithm, you must IMPLEMENT it. BOTH are
necessary. And as important. It is not either/or.

Some people argue against the notion that altimately, resource
consumption is irrelevant to what constitutes a good algorithm.
Their "reason" is that processors get better, memory gets
cheaper, disks get faster, etc.

Actually, what happens is this:

	Consumption expands to the limits of resources.

So, I might accept that the days of PC's with less than a meg
have gone (and yes I know that that is BS; but, just for the
sake of the argument), but that does not mean that our spelling
checker can get larger. Oh no.  What that means is that our
customers expand their product to use the extra memory. Should
our spelling checker get significantly larger, they will go to
another vendor.  And should it be slower than another's...

Anyone who thinks that "microefficiency" is irrelevant either is
not in the real world or has customers who have indefinitely
deep pockets (or who just do not understand the problem, but
time will cure that ignorance).

>                                                      does anyone still
> use profil anymore?  what about size?

Profilers are essential tools around here; so are any other
performance measurement tools that are appropriate to the task.
Including size.