> News flash: ETA: today! :-) It's in git HEAD now. That was fast! Great! Thanks! > Ok, so this does speed up the malloc case by 2.5x. Funny enough, > now the malloc case is compiled, but the VLA allocation is *not* > compiled (same as before). But due to the finalizer overhead the > malloc case is still 3x slower. I think that is enough for my current needs -- the 'benchmark' being a kind of extreme case ;) I am looking forward for luajit 2.1, he he. Cheers Ronan