don't agree with you. it is always challenging to hit or to be faster than any standard library from the original vendor etc. this holds for C/C++ as well (for Julia I presume this holds also in certain areas).
this is not an argument at all towards using Mathematica. An argument is, that it is closed source and years of engineering went into this product and sometimes you have absolutely no clue how these guys are doing it. If you write something that isn't built in you need to have something you measure against with... How did you come to the conclusion that something that does not exist as a built in is naturally slow?
