Curt Sampson wrote:
> On Mon, 30 Aug 2004, Shawn wrote:
>
>
>>>BREW is five times faster than Java
>>
>>...if *true* would be a comparison of handsets without a Java
>>accelerator chip, would it not?
>
> Unless the engineers have gone really nuts with the Java accellerator
> chips, which doesn't seem likely to me, BREW is likely always to be
> faster than Java on the same platform.
The family of VMs that is starting to appear on devices this
year uses techniques similar to the "hotspot compiler" used
in J2SE. As you probably already know, this means that byte
codes are either selectively, or uniformly, dynamically
compiled down to native instructions.
Depending on the quality of the compiler, the result can be
as good as, or in some cases, better than statically
compiled code. Some discussion of why this might (should?)
be the case is presented here:
http://www.idiom.com/~zilla/Computer/javaCbenchmark.html
j.
ps. as should be obvious, i work for Sun but i'm commenting
here as an individual. :)
Received on Tue Aug 31 03:23:35 2004