The DFG currently does a lot of great high-level optimizations, but still keeps it pretty simple when it comes to low-level ones. One option for fixing this would be to write a great backend from scratch. That would be super fun. But before we do that, it would be cool to do a sort of limit study to explore how much LLVM would give us.
Just a joke: FTL in BSG means Faster Than Light, lets aim that speed ;)
Landed in http://trac.webkit.org/changeset/167958
*** Bug 120047 has been marked as a duplicate of this bug. ***