Tools/frameworks for Java multi-core programming?

[anthonycwmak]

I am trying to find tools/frameworks that help with writing Java program that can utilize multi-core CPU. So far I only found jconcurr, JOMP, JAMP, Ateji, but they all either requires special shell command to compile and run the programs, or, in Ateji's case use a special file extension which is converted to java file. This doesn't seem to be ideal if I require to build multi-core programs that can be built on Maven or similar, and requires them to run on an enterprise environment or application server.

Besides writing concurrent programs, by using Java Concurrency API directly, to utilize the multi-core, are there other easier/simpler way to create Java enterprise applications that utlize multi-core CPUs? Preferably, just annotate my POJOs to make them multi-core capable?

Kind Regards,
Anthony Mak

 

[mr.amitghuge1]

http://tutorialswithexamples.com/category/java-tutorials/java-logging-apis/

 

[Robert Klemme]

I am trying to find tools/frameworks that help with writing Java
program that can utilize multi-core CPU. So far I only found
jconcurr, JOMP, JAMP, Ateji, but they all either requires special
shell command to compile and run the programs, or, in Ateji's case
use a special file extension which is converted to java file. This
doesn't seem to be ideal if I require to build multi-core programs
that can be built on Maven or similar,

Why is that an issue? Typically you can include custom build procedures
in a build tool / framework.

and requires them to run on an
enterprise environment or application server.

If you are on an JEE application server that will usually take care of
the concurrency. You typically need to use those APIs the JEE spec or
AS at hand provides to not interfere with the framework in bad ways
(e.g. by breaking transactions). That does not seem like a bad option
although AS come with their own learning curve.

Besides writing concurrent programs, by using Java Concurrency API
directly, to utilize the multi-core, are there other easier/simpler
way to create Java enterprise applications that utlize multi-core
CPUs? Preferably, just annotate my POJOs to make them multi-core
capable?

"Making a POJO multi-core capable" is meaningless. Using concurrency is
more like an architectural property of your application (even though you
need to take care to use POJOs properly - but it's not enough to make
POJOs thread safe). At this point I am suspecting that someone told you
modern Java programs must use multiple cores and you want to get it done
easily. Unfortunately that won't work: you need to understand the
particular issues associated with concurrency to properly craft such
applications - even if you use a framework that takes away a large part
of the work from you.

Kind regards

robert

 

[Lew]

Java programs can "utilize" multi-core CPUs out of the box.
You need do nothing special.

Why is that an issue? Typically you can include custom build procedures
in a build tool / framework.

If you are on an JEE application server that will usually take care of
the concurrency. You typically need to use those APIs the JEE spec or
AS at hand provides to not interfere with the framework in bad ways
(e.g. by breaking transactions). That does not seem like a bad option
although AS come with their own learning curve.

"Making a POJO multi-core capable" is meaningless. Using concurrency is
more like an architectural property of your application (even though you
need to take care to use POJOs properly - but it's not enough to make
POJOs thread safe). At this point I am suspecting that someone told you
modern Java programs must use multiple cores and you want to get it done
easily. Unfortunately that won't work: you need to understand the
particular issues associated with concurrency to properly craft such
applications - even if you use a framework that takes away a large part
of the work from you.

As Robert said, if you're writing enterprise apps using Java EE, then the
threading is already done. You should pretty much never explicitly code
a Java EE component to spawn a thread.

Java SE is another matter. You do have to use thread calls to spawn threads.
Once you have at least two threads in a program (and all Java programs do),
then you are multi-core capable.

Java programs that use threads are already multi-core capable.

Boom, done.

What are you really after?

 

[anthonycwmak]

hi Robert, Lew,

Thanks you guys for reminding me about not to spawn threads on JEE applications directly. I should look into perhaps using EJB Timer and other ways toimplement concurrency properly in enterprise app.

I am interested Loop Parallelization (and Tash Parallelization, Pipeline Paralleization etc) as supported by tools such as jconqurr etc. Which could be useful for scientific applications and matrix operations. (I looked into divide and conquer approaches like MapReduce as well.) But I like the idea like jconqurr and other tools, that I can implement parallelization with just code annotations. At least in those simple case that locking and consistency is not a problem. Thanks for pointing out I can make custom building procedure with a build tool, just that if the parallelization tool/frameworks don't require command line building it is preferable to me.

I am interested in multi-core programming because a while ago there have been some buzz about how to utilize multi-core or even GPU for example for scientific applications. And, I was wondering how to do that in Java. Justcurious.

Anthony

 

[markspace]

I am interested in multi-core programming because a while ago there
have been some buzz about how to utilize multi-core or even GPU for
example for scientific applications. And, I was wondering how to do
that in Java. Just curious.

Introduction here:

<http://docs.oracle.com/javase/tutorial/essential/concurrency/forkjoin.html>

 

[anthonycwmak]

Thanks about the ForkJoin, Mark. Most interesting read! (Even though in some enviroments I still have to use Java 1.6!)

Anthony

 

[Arne Vajhøj]

Thanks about the ForkJoin, Mark. Most interesting read! (Even though in some enviroments I still have to use Java 1.6!)

I believe that you can get the 'y' code from here:
http://g.oswego.edu/dl/concurrency-interest/
and make it work on 1.6!

Arne

 

[Arne Vajhøj]

http://.../category/java-tutorials/java-logging-apis/

What has logging API's to do with this question??

Arne

 

[Arne Vajhøj]

I am interested Loop Parallelization (and Tash Parallelization,
Pipeline Paralleization etc) as supported by tools such as jconqurr
etc. Which could be useful for scientific applications and matrix
operations. (I looked into divide and conquer approaches like
MapReduce as well.)

markspace has already pointed you to fork join.

But I like the idea like jconqurr and other
tools, that I can implement parallelization with just code
annotations.

Now we are starting to talk about something rather specialized.

Maybe you will need to write something yourself.

At least in those simple case that locking and
consistency is not a problem. Thanks for pointing out I can make
custom building procedure with a build tool, just that if the
parallelization tool/frameworks don't require command line building
it is preferable to me.

You can do practically verything from ant and you can call
ant from your IDE.

I am interested in multi-core programming because a while ago there
have been some buzz about how to utilize multi-core or even GPU for
example for scientific applications. And, I was wondering how to do
that in Java.

Basic thread support has been in Java since version 1.0.

Arne

 

[anthonycwmak]

What has logging API's to do with this question??

I was wondering the same.
Also, for some reason this threads seem to have a lot of repeated postings.

Anthony

 

[Lew]

I was wondering the same.
Also, for some reason this threads seem to have a lot of repeated postings.

You must be using Google Groups.

Someone is spamming the cljp newsgroup via some sort of bot that duplicates
old messages.

 

[Martin Gregorie]

You must be using Google Groups.

Not necessarily.

I'm reading my ISP's NNTP server with the Pan newsreader and I'm seeing
the duplicates, though they're easy to recognise since Pan shows repeated
To and From: header pairs in front of the message text.

Someone is spamming the cljp newsgroup via some sort of bot that
duplicates old messages.

Two thoughts:

- what evidence is there that this is really a 'bot and not some bug in
an inadequately tested server, possibly in Google Groups.

I didn't see it before I went on holiday on the 25th of July but it was
doing its thing when I got back online on the 18th of August.

- I follow a few newsgroups but this is the only one that is showing this
odd duplication bug.

 

[Martin Gregorie]

Not necessarily.

I'm reading my ISP's NNTP server with the Pan newsreader and I'm seeing
the duplicates, though they're easy to recognise since Pan shows
repeated To and From: header pairs in front of the message text.


Two thoughts:

- what evidence is there that this is really a 'bot and not some bug in
an inadequately tested server, possibly in Google Groups.

I didn't see it before I went on holiday on the 25th of July but it
was doing its thing when I got back online on the 18th of August.

- I follow a few newsgroups but this is the only one that is showing
this
odd duplication bug.

I think I've found the source: every duplicate message I've looked at in
any detail has duplicated trailers in the message text: the number of
these trailers matches the number of leading To:/From: pairs and all have
this form:

--- BBBS/Li6 v4.10 Dada-1
* Origin: Prism bbs (1:261/38)
--- Synchronet 3.16a-Win32 NewsLink 1.98 Time Warp of the Future BBS -
telnet://time.synchro.net:24


Does anybody know anything about synchro.net? The domain is registered to:
Rob Swindell
330 Latigo Ct.
Norco Hills, California 92860
United States

I've emailed him about this.

 

[Lew]

Not necessarily.

I'm reading my ISP's NNTP server with the Pan newsreader and I'm seeing
the duplicates, though they're easy to recognise since Pan shows repeated
To and From: header pairs in front of the message text.

Two thoughts:

- what evidence is there that this is really a 'bot and not some bug in
an inadequately tested server, possibly in Google Groups.

There were some posts at around the time this all started announcing the
"new service" and someone else mentioning the infamous NewsMaestro
spamware.

I didn't see it before I went on holiday on the 25th of July but it was
doing its thing when I got back online on the 18th of August.

- I follow a few newsgroups but this is the only one that is showing this
odd duplication bug.

That's evidence of malice-aforethought 'ware.

If the other hints are relevant, that is.

 

[Martin Gregorie]

There were some posts at around the time this all started announcing the
"new service" and someone else mentioning the infamous NewsMaestro
spamware.


That's evidence of malice-aforethought 'ware.

If the other hints are relevant, that is.

In support of that, the e-mail address used in the domain's whois is
rejected. However, as the registrar is GoDaddy, I suspect that would not
be considered unusual.