Comments on: BEA plans chip-based Java speed boost

The company is expected to announce plans to speed up its version of Java for servers by letting the software run directly on processors.

Native Code Compiler

Why not just write a native code compiler? Those who want
increased performance at the cost of portability would compile
their Java programs into native machine code instead of Java
Bytecode.

[bboyes]

Native execution chips

Native execution Java chips have been available for a few years. Currently these target smaller editions and configurations of Java, such as J2ME/CLDC. My company developed JStamp (www.jstamp.com) which is 1 x 2 inches and will run on a 9V 'transistor' battery for up to 40 hours. It's good for robotics and industrial applications. It's not a PC or server replacement, but you might want to know that these solutions exist.

To speed up servers, you could certainly build a chip capable of running J2EE natively but it's probably easier to just cross compile to the native instruction set of the hardware. These days PC-type hardware is so cheap that using a multi-processor CPU built with off the shelf hardware is probably much more cost-effective than doing any custom hardware design.

"Raw Iron"

Reminds me of an initiative launched by Oracle years ago, called "raw iron", but as far as I can tell that initiative has not gone anywhere. A desperate move by BEA?