The fastest language is the one that you know well enough that you are actually able to implement and release your ideas to your users. Benchmarks speed is only one part of the equation.
Benchmarks are worthless unless they accurately model your application and workload. Is C 42 times faster than Python for doing numerical calculations? No. If you are using NumPy, the difference is more like 10%. (I pulled the 10% figure out of my nether orifice, but I believe it to be in the ballpark.) Yes, I know that NumPy is buckets of C with a tasty Python frosting.
I don't want the fastest possible language or the most powerful possible language. I want the simplest language that is fast enough and powerful enough. The Benchmarks game is a lot of fun, but if I were making a language (especially one not designed for speed), I probably wouldn't put it up there. It encourages contributors to think "oh crap, if we optimized our X better, we'd be like 4 higher in the listings!" instead of "What sort of useful code can I write?".
Seems quite different compared to what I remember previously. Have some communities been updating their interpreters/compilers to optimize for these tests?
(Reminds me of the 1990s when Intel built a special machine (which you could theoretically buy, to fulfill the demands) and wrote a special compiler for SPECs... :-)
