On Thu, Feb 25, 2010 at 1:07 AM, Gregg Wonderly <[email protected]> wrote: > For more than a decade, there has been http://www.ajile.com out doing direct > silicon execution of Java, complete with some microcode assist. Their web > site now speaks more directly at a market in the JME side of things. The > first version of their chip did JDK1.1.8 with some hiccups in GC that took > some time to work out. > > Sun would not advertise nor support them too much for some reason. The > original team was composed of people who left HP, Sun and other companies > and bought the rights to the chip manufacturing from one of the companies > (I've forgotten which one).
I have worked with the aJile chips. Also the Typhoon chips from Velocity Semiconductors. The latter was much more performant than the former, and could either compile down the Java to assembler (MIPS IIRC) or do dynamic classloading on the target. aJile on the other hand had a stronger security model, where there was actually 2 separate memory spaces, intended for "system" and "user" to ensure full safety for a kernel. I have not looked at it recently, but I think they are both struggling heavy from ARM architectures, which now supports Java acceleration modules in silicon as well, which is the obvious choice for large-volume manufacturers. As for "real-time Jini", you need to understand the constraints, and need to set the expectations accordingly. I am not sure that even RTSJ is capable of delivering Jini within a predictable time frame, so the point of bringing realtime networks to the table doesn't necessary improve the situation (some argue that realtime networks are not possible by definition (for instance, data corruption)). OTOH, there are many "realtime-like" systems, I normally call them "soft realtime", where the time constraints are not "hard", i.e. system just temporarily degrades with failure to meet the time. For such systems, even ordinary Java is often "good enough", if you take great care in object allocations. Having to worry about that through-out Jini though is probably not feasible at this stage. IIRC, RTSJ supports the separation of a time-critical loop versus the overall "enterprisey" functionality, and if Jini is in the later, it might work. But in that case, again, the realtime networks brings very little value. Cheers -- Niclas Hedhman, Software Developer http://www.qi4j.org - New Energy for Java I live here; http://tinyurl.com/2qq9er I work here; http://tinyurl.com/2ymelc I relax here; http://tinyurl.com/2cgsug
