As I understand it, wchar_t might be UCS-4 but I'm not sure that is the case in all implementations. I guess the best bet is to first figure out the actual encoding, and then convert it from/to utf8 perl strings (and remember to read/set the utf8 flag). In other words, it looks like the default typemap does not convert the encoding.

update: a google search for wchar_t encoding perl turns up some interesting articles as the first few results.