tilly, I wish I had ++ points left today. You've made excellent points; I agree with you 100%. If I had to pick one statement from McConnell's book that you linked to, it would be the quote from Donald Knuth¹:

Premature optimization is the root of all evil

I have just this to add: If your code is running slowly, don't guess where the slow spot is, measure. Far too many times, I've seen people confronted by a performance problem take off and try to optimize what they thought was the slow part, instead of trying to profile the program to validate their belief. Then, after reasonable meausrements are made, it turns out their optimizations just made the code complicated and hard to maintain, without making a dent in its run time.

I remember reading about a study, where programmers, asked to guess where they thought a slow program was spending its time, guessed wrong 80% of the time. (I can't seem to find a reference; perhaps someone here can compensate for my Alzheimer's?)

Here's an exercise: suppose I want to add all the integers between n and m. Try guessing which of these will run faster, then use Benchmark to validate your guess. The results will probably surprise you.

use constant N => 1;         # for example
use constant M => 10_000;    # for example
my $sum;

# Method 1: a simple for loop.
$sum = 0;
for (my $i = N; $i <= M; $i++) {
    $sum += $i;
}

# Method 2: an evil use of map by throwing away its result.
$sum = 0;
map { $sum += $_ } N..M;

# Method 3: the APL way
$sum = eval join('+', N..M);
[download]

¹Hmm, I just realized how strange that sounds; my pick from one author's book is a quote from someone else.