c64vector / c128vector are (presumably deliberately) under-specified.
I'm guessing the intent is that a c64vector could/should be implemented
as using the same primitive numeric type as f32vector (i.e. a c64vector could be
implemented using a double-length f32vector).  Similarly c128vector/f64vector.

If that is the intent, I suggest some wording to that effect should be added.
(If it is there already I didn't notice it.)
--
	--Per Bothner
xxxxxx@bothner.com   http://per.bothner.com/