Poorly designed was a harsh choice of words. A better way of putting it would be "not yet optimized for single program execution processor counts > 6 or so." The allocation wall (http://portal.acm.org/citation.cfm?id=1639949.1640116) is something a lot of folks are seeing in practice across a variety of virtual machine-based languages (including Java) when scaling to about 8 cores.
Hrm, interesting. I would have thought the way allocation is done with threadlocal variables negate a lot of the scaling for allocs, but I could be wrong. Do you have a direct link to the PDF, or mind putting it somewhere? I don't have an ACM login.
Sorry, but unfortunately I can't. The authors are permitted to put it online (where scholar.google and CiteSeerX usually immediately find them) but in this case they don't appear to have done so.
