If the user is working with enumerated sets of pc-groups that are all subgroups of a common over-group G, then the following optimization is strongly recommended. Define the set to have the universe PowerGroup(G). For any subgroup H of G, we can find a canonical form for the generators of H. This allows us to have a very good hashing function for the subgroups.
The following example illustrates the optimization.
> G := ExtraSpecialGroup( GrpPC, 3, 3 );
> P := PowerGroup(G);
> time s1 := { P | sub< G | Random(G), Random(G) > : x in { 1..500 } };
Time: 1.140
> time s2 := { Parent(G) | sub< G | Random(G), Random(G) > : x in { 1..500 } };
Time: 9.769