The blog post compares the performance of quicksort implemented in Java, Scala, Groovy and Groovy++.
(Groovy++ is an effort to create a statically typed variant of Groovy, so that dynamic and static programming can be mixed).
Bottom line: Scala and Java compare in performance, Groovy is 512 times slower and Groovy++ is only 42 times slower. This is obviously just a micro-benchmark, but I don't see how manipulating any other sort of data will make the difference in performance smaller.
Update: It turns out that using the shift operator instead of division dramatically improves groovy++ performance: http://stronglytypedblog.blogspot.com/2010/02/groovy-performance-now-were-talkin.html. bottom line, groovy++ is only 50% slower than Java/Scala after this improvement. Looks promising!, although I hate the idea that choosing the wrong way of implementing a line of code can cause performance to degrade so significantly.