Good to hear from an Iowan!
I keep thinking I should switch to Java from C/C++ one of these days, but it will still probably be another few years or so until we see Java consistently on all embedded platforms running Linux. The consistency and availability of C/C++ on embedded platforms is really hard to beat right now, and I have a lot more experience with C/C++. When is it going to be easy enough that you guys will support Java out of the box on every board shipped?
On embedded applications needing a good real-time garbage collector, I imagine there are applications that are continuously allocating and de-allocating memory, but the common ones I write, you typically allocate space for input and output objects, data logging, scripts, etc, at start up, and never delete or create anything from there on out. In C/C++ with the dangers of allocating and de-allocating and getting it right, it is often best to allocate for max number of inputs, etc, and never use new or malloc() except at start-up where possible.
Can anybody give us a good example of an embedded real-time application that really does need to do a lot of allocating and de-allocating after start-up?
If I were selling real-time garbage collectors for embedded systems, I would probably not talk about embedded real-time applications that don't need it.
I still think they are in trouble with Perk if they do not open source it. The standard JVM and compiler will be what everybody uses eventually if they don't. They need to figure out how to switch to a support model as soon as possible.
Don.
On 11/8/06, Jesse Off <> wrote:
--- In ts-7000%40yahoogroups.com, "Don W. Carr" <> wrote:
>
> I know Kelvin Nilsen from my days at Iowa State University where
he was a
> professor for a while, and where they started out in the research
park
> before moving to Tempe.
Small world! I worked at that research park in Ames, IA. NewMonics
(now Aonix) moved engineering to Tucson actually (not Tempe) and
thats how I arrived here in Arizona.
He is a very bright guy, but I wish they would
> figure out a way to open source at least the run-time or it will
never take
> off in embedded devices. They will just slowly improve the
standard JVM for
> embedded systems and Perc will become irrelevant if it is not open
sourced.
> If they did open source it, they would become ubiquitous on
embedded systems
> and would make a lot more on service and support. Heck, it would
hasten the
> move to Java on embedded devices. If there was a JVM with real-
time features
> on virtually all platforms, things would be a lot different.
Its probably harder to make something open-source when many people
and investors have already spent lots of $$$ building it. :-)
Indeed, it would be nice to have PERC open-source, but from what I
understand, the Aonix PERC products are doing well for the company's
bottom line right now.
>
> Another small issue also is that with most real-time applications,
you
> allocate objects once and never stop using them. Though it is good
to have a
> good garbage collector, the garbage collection issue is not very
big for
> most embedded real-time applications.
I know some people at Aonix that would probably disagree with you.
//Jesse Off
-- Dr. Don W. Carr J. G. Montenegro 2258 Guadalajara, Mexico +52-333-630-0704 +52-333-836-4500 ext 2930
__._,_.___
![]()
SPONSORED LINKS
__,_._,___
|