Hi Marcos,

I said to take my contribution with a grain of salt because I don't have a lot of experience in this field. In other words, if it makes sense to you, then verify it for yourself. Don't just take my word for it.

That said, I suspect that your observations, as well as the input you're asking for, are extremely compiler dependent. I suspect that you can also affect your results/observations by changing configuration parameters in a given compiler.

You might poke around on the site here for a disassembler that includes clock cycles in the output. I seem to recall that someone has one they were going to upgrade to do just that. Maybe someone will read this and point you to what I'm talking about. If not it should be a fairly easy search. The point is that it will allow you to make quantified comparisons of the output from various compilers, or of multiple compilations with your own compiler using various configurations. If you're really concerned with such things, it would be a handy tool to have.

Joe