Standardized benchmarks, implemented in multiple languages, using the language's idiomatic, default containers, iteration methods, type methods and so. An effort to allow an almost fair comparison of core characteristics of the various languages.

An initial comparison of a loop recognition algorithm, implemented in C++, Java, Go, and Scala, has been published at the Scala Days 2011 Workshop at Stanford: https://code.google.com/p/multi-language-bench/source/browse/trunk/doc/scaladays2011-paper/main.pdf

Publicity

This paper got picked up by The Register with a slightly more snappy title, and it was all over the Geek Blogs soon after that, e.g., Slashdot, or Reddit

Responses

Several people have pointed out flaws in the paper or suggested improvements. All of those are greatly appreciated and should help to better understand the various language issues.

GO

Russ Cox published an excellent Blog Post on how to improve performance. In the end, they even beat the tuned C++ version

Java

Jeremy Manson published an interesting comment on his Blog

News

• We've added a manually constructed JavaScript version
• We've added a manually constructed Dart version. It includes instructions on how to generate JavaScript from it.
• We've added an emscripten C++ to Javascript converter script.
• We've added a GWT version (basically, a wrapper around the Java version). This version is difficult to run and requires some manual massaging. This version also allows to generate JavaScript.
• We've added a NaCl version. This version uses a similar technique as the GWT version, but can be run more easily (on Chrome).