🚀 go-pugleaf

I have been trying to understand the WAM by reading "The WAM: A (Real) 
Tutorial" by Hassan Ait-Kaci (it is a technical report from DEC's Paris
Research Labs). One of the things that I am puzzled about is which
WAM instruction is responsible for storing the permanent variables
in the environment when a new rule is called.

The situation is illustrated by the following:

	foo(args) :- ...., bar(args),....

This is translated to:

foo:  allocate
      get args of foo
      ..code for predicates before bar...
      put args of bar
      call bar/arity, N
      ...code for predicates after bar...
      proceed

bar:  allocate
      get args of bar
      ...code for bar's body..
      proceed

Now the number of permanent variables in bar is not necessarily equal to
its arity. In both Warren's original paper and the WAM tutorial referenced
above, the description of allocate does not seem to reserve space for the
permanent variables. So my question is which instruction allocates the space 
for bar's permanent variables ?

Thank you.

         Naim