The Fastest BigInt in the West

>>"Unlike the Benchmarks Game, we aren’t requiring all the test programs for the same language to be identical. This is a controversial rule..."<<

Benchmarks Game programs for the same language are not required to be identical. Wilfred Hughes has been asked to correct that misstatement.

>>"It’s also not clear how representative the test programs are of typical performance of that language."<<

Without sampling programs "in the wild" how could anyone possibly claim that other programs were "representative"?

See http://research.microsoft.com/en-us/projects/jsmeter/

Wilfred, please correct your statement that the benchmarks game requires "all the test programs for the same language to be identical".

It isn't true. It wasn't true 4 years ago.

For sure, my preference was to show PyPy programs that also worked with CPython -- that made clear that optimizing for PyPy could make performance worse with CPython and vice versa.