How does it really work?

Totally just for the heck of it...

It is my understanding that if I have a dual core system, I could create 20
LPARs of .1 processor/core each.
The system basically divides some unit of time by 10 and assigns 1 unit of
the resulting time per 10th of processor assigned.
On a single LPAR ( no LPAR ) box all time slots would be available. If I
assign say '1' core to LPAR A and '.5' each to LPARs B and C
LPAR A would get 10 time slots and B and C would each get 5 time slots.

First Question, with the HW above does LPAR A only use time slots from a
specific core or does the system handout time slots from both cores?
Second Question on a 6core system with only 2 cores licensed for IBM i and
no other workloads ( OSes ) on the system, is it spreading those time slots
across all six cores or just 2?

Just curious