Following on from the bumper Ruby interpreter performance tests by Antonio Cangiano two weeks ago, Xue Yong Zhi decided to run the same performance suite on XRuby Ruby to JVM compiler and found that XRuby is faster than the Ruby interpreter in 26 out of 38 tests. It's interesting to note that the failing tests are the same as those for the official Ruby interpreter on Windows (stack exhaustion?) so in the scope of these performance tests XRuby appears to be more feature complete than interpreters such as Cardinal and Rubinius that fail many of the tests.

While XRuby's performance is certainly amazing for a project that's less than 2 years old, and the idea of compiling Ruby directly to Java bytecode is initially appealing, the results for YARV/Rite are still streets ahead in terms of raw performance, and where I'm placing my bets for the next de facto Ruby interpreter.