Welcome to the early version of www.sunspot.systronix.com
This web site is the work of Systronix Inc, a SPOT Beta site.
The official SPOT resource is sunspotworld

Newer entries are at the bottom of this page

Benchmarks (2006 Sep 25):

I pasted my benchmark code into the SPOT ApplicationTemplate example and this is the result:
http://www.practicalembeddedjava.com/benchmark/11a/benchmark_11a.html

I was very pleasantly surprised how good the performance was, especially for a beta release. TheSPOT is right between the Imsys SNAP (80 MHz, 32-bit CPU, 8-bit external data path, micro-coded core) and JStamp (70 MHz, 32-bit CPU, 8-bit external data path, native execution), but in some cases as good as JStik (100 MHz, 32-bit CPU, 32-bit external data path, native execution). It's surprising that the SPOT Beta is initially so close to native-execution, 32-bit CPU performance.

There are some results worth a comment or two:

• SPOT doesn't benefit as expected from System.arraycopy of a byte array. It does benefit more in the System.arraycopy of an int array case.
• Byte multiply, and int multiply and divide are all stellar, better than JStik. This is due to the native ALU of the ARM?
• String handling, especially concatenation, lags behind JStamp.Since Strings are used so much in communication this would be worth improving.
• Object creation is astonishingly fast, much better than JStik.
• floats and doubles (the primitives) are about 1/8-1/10 the speed of JStik.

Overall, considering how highly integrated SPOT is, and the fact that it runs on a controller with a much lower price point than JStamp or JStik, these are very good performance results. It's reasonable to assume they will only improve over time.

These benchmarks are just a start. The other items critical to quantify in the near future include:

• SPI burst (6 MBits on JStik, ARM should do at least as well) and sustained.
• I²C burst (under 50 kbits JStik - ARM should be much better) and sustained
• Interrupt response latency and determinism. JStamp and JStik do very well at this, with sub-millisecond response and a few microseconds max of jitter.
• Thread context switch (JStik switches in about 1 usec)
• Periodic thread capability - JStamp and JStik do well at Periodic Threads though the tools to support them could be much more friendly and helpful.
• Asynch serial data burst and sustained, including hardware and software FIFO support with internal and external UARTs.
• Debugging options

Credits, thanks and whom to blame:

All the photos here are courtesy of Bruce Boyes - peruse their metadata for a bit more information. SunSPOT, Java, etc are trademarks of Sun Microsystems. Thanks to the great SunSPOT team at Sun Labs (no, not Sun Laboratories) for letting us participate in the alpha and beta programs. And to Radio Paradise for great music to program and write by.

Go to sunspot.systronix.com Index