SQL & PL/SQL

Does anybody know where I may find documentation discussing the performance differences between a PIPELINED function versus using a BULK COLLECT?

Or, does anybody have any advice on which of these two approaches is faster? I am almost certain that the level of performance for these two approaches is tied to the specifics of the application and what it should do, which is why I asked for documentation that I could read to asses the differences as they would apply to my need. But, if you have some experience choosing between these two approaches, I would love to hear about it.

Thanks.