Rational benchmark

[Cherubin62]

Add a new benchmarh dedicated to rational number type

[bluescarni]

What type of operations would you like to be benchmarked for the rational type?

[Cherubin62]

Actually I would like to benchmark exact geometry based on rationals, as done with gmp in CGAL package. I was wondering if mp++ could compete favorably. This implies +,-,*,/ operations at least.

[bluescarni]

Ok, so benchmarking arithmetic operations against GMP's rational. What kind of limb/bit sizes are you normally using for numerator/denominator? Do your numbers usually fit, say, in 128bit of storage or do they need more?

[Cherubin62]

128bit might be enough in a wide range of use cases. But, actually, exact geometry is only interesting in very degenerate cases, where more precision might be needed. To answer shortly, impossible to assert 128 bit will be enough in all cases.

[Cherubin62]

By the way, I do really appreciate your interest in my problem and look forward to compare your mp++ vs gmp benchmark and, afterwards, gmp-based geometry vs. mp++-based geometry becnhmarks.