The Fastest BigInt in the West

&gt;&gt;&quot;Unlike the Benchmarks Game, we aren’t requiring all the test programs for the same language to be identical. This is a controversial rule...&quot;&lt;&lt;<p>Benchmarks Game programs for the same language are not required to be identical. Wilfred Hughes has been asked to correct that misstatement.<p>&gt;&gt;&quot;It’s also not clear how representative the test programs are of typical performance of that language.&quot;&lt;&lt;<p>Without sampling programs &quot;in the wild&quot; how could anyone possibly claim that other programs were &quot;representative&quot;?<p>See <a href="http://research.microsoft.com/en-us/projects/jsmeter/" rel="nofollow">http:&#x2F;&#x2F;research.microsoft.com&#x2F;en-us&#x2F;projects&#x2F;jsmeter&#x2F;</a>

Wilfred, please correct your statement that the benchmarks game requires &quot;all the test programs for the same language to be identical&quot;.<p>It isn&#x27;t true. It wasn&#x27;t true 4 years ago.<p>For sure, my preference was to show PyPy programs that also <i>worked</i> with CPython -- that made clear that optimizing for PyPy could make performance worse with CPython and <i>vice versa</i>.