It doesn't matter much what the backends are, it's important to be able to demonstrate the effective implementation of the problems posed by the tasks being benchmarked. Especially for the new language. Of course, if your language is designed as to have the limited chances of being universaly fast, it's another story. But Rust is designed to be able to reach at least the speeds of C and also allow for a more modern approach. That's the real worth of the contributions to the "benchmark game:" demonstrating the capabilites of the language, even if it's just on some of all of possible tasks.

See also Rust vs Haskell: http://benchmarksgame.alioth.debian.org/u64q/benchmark.php?t...

And Rust vs Go: http://benchmarksgame.alioth.debian.org/u64q/benchmark.php?t...

It's not that easy to implement a language that's expressive with this kind of performance, even if you use llvm. Kudos to the Rust guys!

The Mono C# programs also use llvm ;-)

A more accurate comparison? A different comparison would be against clang.

The benchmarks game might well be described as "a good starting point" http://algs4.cs.princeton.edu/faq/ but let's not confuse that with some kind of definitive final analysis.

(Edit: Say what you disagree with, don't just down vote.)

I was under the impression that clang was still, generally, slightly slower than GCC for C++. Perhaps this has changed in recent years (as clang and llvm have been seeing tons of development effort and funding).