Re: Re: Re: Re: Re: Resulting class performance

I'm no Hotspot expert but here is one observation I would make:  Both  
the Java and your main method require on-stack replacement to replace  
the original interpreted code with the JIT'ed code.  I notice in your  
bytecode that you reuse the String[] argument slot to store another  
local variable.  Perhaps this slight stack frame weirdness confuses  
the on-stack replacement logic enough such that it cannot install the  
JIT'ed code?

Chris

On May 7, 2010, at 4:19 PM, Jeremy Cowgar wrote:

> On 5/7/2010 4:08 PM, Scott Vachalek wrote:
>> I know.  Just saying, it's still tricky -- for example if the  
>> problem is really the iconst/ldc difference it's not really going  
>> to make a major difference to fix it, although it will seem like it  
>> for your test.
>
> That took it to 2.6 seconds, but I think the real problem is the  
> compile failed that I posted a minute ago.
>
> Jeremy
>
> -- 
> You receive this message as a subscriber of the [email protected] mailing  
> list.
> To unsubscribe: mailto:[email protected]
> For general help: mailto:[email protected]?subject=help
> OW2 mailing lists service home page: http://www.ow2.org/wws

-- 
You receive this message as a subscriber of the [email protected] mailing list.
To unsubscribe: mailto:[email protected]
For general help: mailto:[email protected]?subject=help
OW2 mailing lists service home page: http://www.ow2.org/wws