The Same is the Same? or How to Compare for Equality in Computer Performance Evaluation