Personally I think performance is underrated. My programming experience started off with 86 assembler and C, writing games targeted for the 286, through the first step of the Pentium.

I can tell you, trying to make a 3d texture map renderer for a 386sx-16, you need every tweak, hack, and trick you can think of.

Fast forward awhile and I find myself starting with Perl. At first, the whole idea of a quasi-interpreted language made me bristle. I always had a constant nagging guilt over those poor wasted clock cycles and bytes. With time, that subsided somewhat with the great amount of ease and speed I could make complex things.

I think there is a balance. I still try to make things quick, occasionally I still have that pang. Though if it comes down between making something fast but convoluted, or easy to maintain and expand, I'll forego speed for ease. Though I think my early experience (and that pang) have come in handy. Especially when something ends up getting a lot more use/users then orginally intended, and it holds up.

Though there's a difference between extreme optimization (loop unrolling, self modified code) and putting a little thought into algorithms and data organization.

Think about Windows bloat. Outside of games, why the hell have hardware requirements gone up so much? I think a lot of that is just laziness. Good for dell, not so good for end users.

Linux is a good example of the opposite. Not saying it's perfect, but it's certainly much better IMHO. Lately I've been working on a full screen graphic intensive Perl/SDL app for my car. The target, a 500mhz laptop I got circa 2000. Runs blazing fast with CPU to spare. No way I could do that running WinXP. Probably not even win98

On the other extreme, I've also been working on a bunch of microcontroller stuff related to it. It's all clock cycles and bytes. Kind of a drag, but I've been able to take what I've learned from my Perl experience and make it a little more managable and configurable even with only 2k of ram to work with.

Personally, I'd say as a rule of thumb, if it's something that is going to see more than a one time use, get it working, then make it as fast as you can without making it hard to continue developing.