1. Home
  2. Embedded Linux
  3. looking for performance statistics (native JAVA processors)

looking for performance statistics (native JAVA processors)

hey.

Does anyone have references to performance stats or whitepapers about performance of native java processors vs. software jvm for embedded apps? I read the hardware specs on a couple of 100mhz processors and they didn't seem very impressive, even for running bytecode in hardware. One spec showed a benchmark of copying a 10,000 byte array in 64ms. On the surface that seems fast but consider that 10000/0.064secs=1.25mbits/sec. and they include an ethernet tranceiver on the eval board. I haven't used a 1mb/s ethernet in nearly 15 years!

I guess what I'm really wondering is whether there are java native processors out there that can approach risc mips levels. I understand that there will be emulation overhead even in hardware but my requirements will be to keep up with 802.11b wireless speeds: 11bm/s. Anyone know of java native processors that can deliver that kind of thru-put and support the FULL J2ME specs?

tanks

Reply to
noone
Loading thread data ...

640 cycles per iteration. Running interpreted on my desktop machine gives around 200 cycles per iteration. However, CLDC (and even RTSJ) have block copying methods which will presumably be written natively.

ARM processors with Jazelle can execute Java byte code directly at a reasonable pace. They also support Ahead-Of-Time (AOT) compilation for speed critical parts.

formatting link
formatting link

Tom Hawtin

--
Unemployed English Java programmer
http://jroller.com/page/tackline/
Reply to
Thomas Hawtin

the intent of a Java processor would to conserve RAM. It could run as an interpreter without chewing up megabytes of RAM with the machine language equivalents.

Sun abandoned its PicoJava chips when it turned out they ran too hot. You need a low power chip for a hand held.

Have a look at

formatting link

I have always wanted to see a machine with a cached stack with a mindless little processor that worked in parallel using spare memory cycles to keep the stack from overflowing and to keep it filled.

--
Canadian Mind Products, Roedy Green.
http://mindprod.com Java custom programming, consulting and coaching.
Reply to
Roedy Green

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.