There is a theorem: http://en.wikipedia.org/wiki/Finitely_generated_abelian_group"
which states that every finite Abelian group can be decomposed into the direct product of groups of prime order (any group of prime order is cyclic btw).
Let's look at the subgroup A_p. The order of each x\in A_p must be p, it's clear that it can't be larger than p, and because p is prime it can't be smaller either. If the order of x is r which is smaller than p, then r must be a divisor of p, but p is prime.
Clearly, every element of A with order p is included in A_p.
And we have:
A_p=P_1 P_2... P_n where every P_i is a cyclic group of order p and n is the number of times that these groups occur in the decomposition of A.
Let's look at the subgroup A^p, if A^p\bigcap A_p\neq\{e\} then \exists x\in A_p,\ \ x=y^p and then \exists y, with order p^2.
But then let's look at the subgroup generated by various power of y. this is a cyclic group of order p^2 and as every other abelian group can be decomposed into a product of two order p cyclic groups. so y=hg where both h and g have order p. but hg must also have an order p in that case. and so we have that the order of y is p, and thus
A^p\bigcap A_p=\{e\} . If we look at the homomorphism:
\phi :A \rightarrow A^p \ \ \ \ \phi (a)=a^p
the kernel of this homomorphism is clearly A_p and we have:
im(\phi) \cong A/ker(\phi)\ \ \Rightarrow \ \ A^p \cong A/A_p
now we have enough information to conclude that A is decomposed into the product of A_P and A^p which answers both parts of your question!
During the proof there was ambiguity between semi-direct product and direct product of subgroups but when it comes to abelian groups the two are the same up to isomorphisms.
I am absolutely certain that the proof contains a lot of unnecessary stuff and can be simplified so good luck, and I hope I got the main idea through.
P.S. You can also say that let A=P_1 P_2...P_n Q_1 Q2...Q_m...Z_1...Z_l be the decomposition of A into cyclic groups of prime order where each two groups with the same letter have the same order. Then the product of all these groups with the P_1...P_n left out should be isomorphic to A^p, the reason for this being that taking the p-th power of each element is an isomorphism because p is not a divisor of any of the orders of the elements in this subgroup.
