Christian Thalinger wrote:
On Tue, 2005-12-06 at 10:00 -0600, Archie Cobbs wrote:
That's getting into the micro-optimzation realm, which is
fraught with danger and mistaken assumptions :-) E.g., on
some machines the time overhead of setting up a try/catch in
a method that wouldn't otherwise have one is higher than
the single comparison required to test for null. In particular,
any interpreter is going to have to test for null anyway,
so the second time it's already in cache, blah blah, etc.
"...setting up a try/catch..."? What do you mean? Agreed on the
interpreter thing, but who does benchmarks on interpreters?
On gcj using SJLJ exceptions there is overhead for entering a try block.
And for the dwarf unwinder there is overhead in the unwinding tables
(but not in the code) for each try/catch.
That you don't care about the implications for platforms other than your
own, does not imply that they do not exist.
David Daney
_______________________________________________
Classpath-patches mailing list
Classpath-patches@gnu.org
http://lists.gnu.org/mailman/listinfo/classpath-patches
