Originally posted by DaPi
I don't think you want COUNT(NULL) - it ought to be fastest 'coz the result is always zero.
Doh. Of course, that realy was silly of me. But, BTW, it wouldn't be any faster (nor slower) than other options, because the explain plan will still be the same, the number of I/O operations will still be the same, the CPU load will still be the same - ONLY THE RESULT WILL BE WRONG! So technicaly speaking, my answer still stands: "performance wise COUNT(*) = COUNT(NULL) = .....".

If you really want to slow things down, COUNT(DISTINCT anycolumn_evenPK) will do so - so it's not quite anything_you_like.
Well, yes, of course, if you substitute "anything_you_like" with something stupid like some heawy power PL/SQL function that can bring your server to the knees. In this context you can treat DISTINCT in this case as such a function (BTW, DISTINCT is a funny beast, as it can be used both as a function and as an operator). What I ment with "anything_you_like" was any_constant, any_pseudocolumn, any not_null_column, any_you_know_what...