>>19
I've read somewhere that those benchmarks were made strict because the lazy ones cut corners and didn't do the same amount work that their equivalents in, say, Java did, thus being unfairly faster. Allegedly the strictness was not introduced to gain speed, but to lose it.