> > - avoids comparing an unsigned long to -1

Just out of interest, why?  The cast is unnecessary:  there's no ambiguity 
or undefinedness (the int -1 gets promoted to unsigned long, with 
wraparound semantics), and neither gcc nor MSVC complains.

Other than that, the patch looks fine to me;  x ^= x >> 4 would be fine  
too.  I really don't see that it makes much difference either way, since 
both transformations are reversible and fast enough.