There is no simple answer to that. It all depends on the layout. As you are probably aware, as long as there are currents flowing through the ground traces, there is going to be a small amount of voltage difference between various points of the ground path because the impedance of the copper in the traces isn't zero. The solution is to minimum that impedance in order to minimize that phenomenon.
A separate ground plane layer is most convenient because there is a lot of copper for the lowest impedance, and parts that need a ground connection can be tied to the ground plane right where it is. Moreover, the ground plane serves as an electrical shield against external interference (although for a power supply this is of minimal importance).
Single point grounding is basically "star grounding" applied to printed circuit layout (where each part that need to connect to ground has its own path, leading to a common ground point that is used as reference). In reality it is usually unfeasible to design a PC board with so many separate ground paths, for space and routing reasons. Also, if any branch of these paths is long, then the trace impedance will cause voltage to be developed, which is undesirable.
For single layer boards, you can arrange the parts that need to go to ground such that they are located near each other, and then have a large ground copper area (or a plane) to tie them all together. This would then essentially be the same thing as single point grounding done right.