Re: [Cxtable-devel] RE:The last code edit I had done....

[Borne Goodman-M. <bm...@eg...>]

The most expensive object to create an instance of in Java is an array,
so if this process is done often, then it could certainly have an impact
on performance.

--bjgm

On Wed, 2001-12-05 at 10:46, Williams, David wrote:
> Borne...
> 
> (This code edit is reflected in the current source on CVS...I have not worked on it since)
> I did something that I felt should have increased speed a bit on the inbound-processing end...but in testing it there was no difference in the benchmark.
> 
> At the time the Listener-dispatch was started, it took System.currentTimeMillis().  Once it had finished cycling through all of its read-delegating, it did the same, took the difference.....  It was either 0 or 50,60,.... in both cases...  
> 
> Prior to this change.... ALL xListener objects stored in the xListenerRegistry for that particular xClientConn would receive the message and determine if it was related to their function or not...  I thought that the new way (checking first if they accept ALL messages with a if (xlistener.readAll()) {/*send it on*/} ....   and then checking which Strings (String[] keys = xlistener.readKeys();)
> 
> I think the lack of performance boost comes from the fact that the String[]'s returned for that method are created each time the method is called, rather than final static String[]'s...  I think if I make them final static it might increase speed on processing...
> 
> Does that make sense, do you think?
>