DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG 
RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
<http://nagoya.apache.org/bugzilla/show_bug.cgi?id=7058>.
ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND 
INSERTED IN THE BUG DATABASE.

http://nagoya.apache.org/bugzilla/show_bug.cgi?id=7058

QName.toString() should use StringBuffer concat to improve perf





------- Additional Comments From [EMAIL PROTECTED]  2002-03-21 17:34 -------
Dave suggested, offline, that what the Java compilers are doing is creating a 
default 16-byte StringBuffer and issuing multiple appends to it, rather than 
allocating it larger.

I would actually expect that a JIT compiler would rewrite that to check the 
lengths of the inputs and do something more intelligent. If not, someone should 
feed that back to JVM developers as a suggestion; it's an obvious opportunity to 
improve how a common code pattern is executed.

In any case, there's the question of whether changing this particular instance 
of stringbuffer concatenation would make a measurable difference in Xalan 
performance. I'd like to see that validated before we start doing this sort of 
microoptimization. A 99% speedup of something that accounts for 1% of runtime 
is, at best, a 0.99% total gain... and sometimes trying to write around the 
compiler hampers, rather than improves, its ability to optimize, or helps one 
JVM while harming another.

Worth investigating in detail across multiple platforms, when someone has the 
cycles to do so... but I'm inclined not to worry about it unless a major gain 
can be demonstrated.

Reply via email to