I was researching general optimization in C++ and I found 2 very interesting articles/books/websites that offer a lot of useful information on optimizing C++ code.
http://www.tantalon.com/pete/cppopt/main.htmhttp://en.wikibooks.org/wiki/Optimizing_C%2B%2B
Both articles cover similar techniques, but I think that the latter one covers some additional stuff. I knew most of these techniques, except for a few very exotic ones. But I have to admit that I was surprised by the heavy speedup when using templates with constants (e.g. factorial calculation), omitting vtable initializations (useful with abstract/pure virtual classes) and that prefix increment/decrement operators are faster than postfix ones (++i is faster than i++, especially when implementing the operator in a class).