Query Approach and performance

Hey everyone,
On average, are multiple simple queries better performance-wise than joins?
i.e.
select A.col1 from table1 A
select B.col2 from table2 B where B.col1 = A.col1
etc

vs

select A.col1, B.col2 from table1 A, table2 B where B.col1 = A.col1

Are joins better for small/large numbers of tables?
Is there a diff?
My approach to date has been to keep queries as simple as possible, and
when I see a need for complicated joins, I create a view and then do simple
queries against that.

Does pg cache queries like Oracle does so that repeated queries don't need
to go through the compile phase and run faster? Is this configurable?

Thanks,
Morgan