You might remember my last post about IronJS performance, and that there were some trouble with the speed of implicit globals in the loop-empty-resolve.js benchmark, after a lot of improvements in both the generated code, the compiler and the way globals are handled I can now report that the performance of implicit globals (and globals in general) have increased ten fold.
Do note that these benchmarks were run on a different machine than the original benchmarks were, which is also why I included the left most staple showing the old code run on the same machine as the new code, TraceMonkey and V8.