Open atlaste opened 3 months ago
I would bet the compiler optimizations were explicitly turned off, to fully control all of the variables between the tests.
Yes, in a proper setting they would ideally be turned on, but it'd be hard to tell if the change made is to be ascribed to the change the programmer implemented or some strange adjustment made by the compiler optimizations. Disabling the auto optimizations would likely help control the variables.
I was just surprised at these numbers until I saw a loop that the compiler should optimizations away and inspected the Makefile. You guys disabled the optimizer by not adding -O3.
Please fix it, because otherwise these benchmarks are useless, which is a shame considering the work that went into this.