On Tuesday 30 May 2006 23:38, Alex Blewitt wrote:
> On 30/05/06, Chris Gray <[EMAIL PROTECTED]> wrote:
> > If the implementation is an empty method and is final, a straightforward
> > static flow analysis will show that the evaluation of the arguments can
> > also be optimised away.
>
> Not necessarily. Evaluation of arguments may have side-effects, and
> therefore even if the call to the logging gets optimised away, the
> evaluation may not be.
You're right, and the more I look at it the less likely it seems to me that
the evaluation of parameters will be optimised away. To start with string
concatenation: the expression ("hello " + "world") has no side-effects, but
the optimiser has to perform an escape analysis in order to recognise that
new StringBuffer("hello ").append("world")has none. (StringBuilder I guess
these days). And in ("Frobbing the " + foo), if (as is likely) the type of
foo is a non-final class or interface, all bets are off; a subclass could
override the toString() method to have all kind of side-effects. Logging
("Connecting to " + remotehost.getHostName()) certainly has side-effects.
It's probably also not a good idea to rely too much on JIT optimisations,
given that Harmony should run on a number of VMs and not all of these will
have a fully optimising JIT in all circumstances. It should be possible to
compile the class libraries with or without logging.
Chris
--
Chris Gray /k/ Embedded Java Solutions BE0503765045
Embedded & Mobile Java, OSGi http://www.k-embedded-java.com/
[EMAIL PROTECTED] +32 3 216 0369
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
