Some developers (and clients) who are doing development on the basis of $100-$150/hour or more are going to probably prefer paying $1500 for a compiler that produces tight code from the get-go than going with a free alternative that might require 30 hours (i.e. $3000-$4500) to optimize by hand. This is even more true if you end up having to modify code that was already optimized... meaning you may have to optimize it again after the modification.

If you are willing to work for $1.00 or less an hour, you absolutely do not need commercial products. This, of course, will be true for many amateurs.

I can not recall ANY case where a tight tool budget has resulted in any savings.

Examples of "negative savings" I recall off the bat (I am sure there are more):
* Writing substantial code in assembler because the powers would not pay for a compiler.
* "reasoning" and experimenting to find the location of a narrow spike because the powers would not pay for a fast scope (only 20MHz available).
* Hunting for weeks for a miniglitch because the powers would not pay for an ICE.
* Fixing bad joints (quick, but the time to find them) all over the powers would not pay for thermostatically controlled a soldering station for production.

So it all comes down to the fact that, in most corporations, management can accept "I need more people" but not "we need more instrumentation" to produce more.

Erik

PS I am VERY fortunate to be working at a place where management uderstand that "more productive" is advantageous to "more people".