MPJ Express Blog

A talk on MPJ Express at CSAIL, MIT

2011-02-28T09:52:00.001-08:00

Aamir Shafi gave a talk on "Programming Parallel Hardware using MPJ Express" at CSAIL, MIT.

MPJ Express Software Citation in Research Papers

2011-01-07T15:05:00.000-08:00

Dear MPJ Express Community,

Please cite (reference) the following paper for acknowledging the use of MPJ Express in our research work.

Nested parallelism for multi-core HPC systems using Java [bibtex]

Thanks

MPJ Express Team

MPJ Express Release 0.37

2011-01-07T14:58:00.000-08:00

Dear MPJ Express Community,

We are pleased to announce the next beta release (0.37) of the MPJ Express software. This release is mainly a bug fix release.

CHANGELOG
---------------------

1. Added the executing and debugging tutorial of MPJ Express applications using Eclipse IDE. 14/5/2010

2. Modified the runtime to get rid of the shared directory requirement between the starter (head node) and daemons (running on compute nodes). This part is coded by Rizwan Hanif and Amjad Aziz. 6/1/2011

3. Fixed a bug that causes daemons to crash on Windows boxes. 4/1/2011

You may download the software from:

http://mpj-express.org

Feel free to contact us in case of any queries.

Best Regards,

Aamir Shafi
Bryan Carpenter
Mark Baker
Guillermo López Taboada

Debugging MPJ Express Applications using Eclipse and Netbeans in the multicore mode

2010-12-01T08:39:00.000-08:00

Acknowledgement: This work is done by Amjad Aziz and Rizwan Hanif.

Debugging parallel applications with Eclipse:

This video tutorial shows how to debug applications in the multicore mode using Eclipse IDE.

Debugging parallel applications with Netbeans:

This video tutorial shows how to debug applications in the multicore mode using Netbeans IDE.

We will be posting more videos for debugging parallel applications in the cluster mode shortly.

Executing and Debugging MPJ Express Parallel Applications using Eclipse

2010-05-13T21:51:00.000-07:00

Acknowledgement: This work is mostly done by Rizwan Hanif and Amjad Aziz.

We have written a tutorial (video and textual) that allows debugging parallel applications using Eclipse IDE. Currently it's only possible to debug parallel Java applications when MPJ Express is running in the multicore mode - this essentially means that the target platform is a laptop/desktop with shared memory or multicore processors in it. For more details, see the textual guide or the video guide.

A List of MPJ Express Methods

2010-04-20T02:01:00.000-07:00

Most (or let's say a good amount of) MPJ Express users have excellent knowledge of the Message Passing Interface (MPI) specification document (http://www.mpi-forum.org/docs/docs.html). For this reason, it will be helpful if there's a document that maps MPI methods to javadocs provided by the MPJ Express software.

We have made an effort to provide such a mapping from MPI routines to MPJ Express methods. See the document here (http://mpj-express.org/docs/guides/listofmpjmethods.pdf).

MPJ Express Version 0.36

2010-04-19T03:30:00.000-07:00

Dear MPJ Express Community,

We are pleased to announce the next beta release (0.36) of the MPJ
Express software. This release is mainly a bug fix release.

CHANGELOG
---------------------

1. Fixed the mpjboot error for Mac OS X reported by Ed Baskerville. The mpjboot script used x86 binaries on Mac OS X, which obviously was incorrect. 18/3/2010

2. The MPJ Express runtime could not start more than 100 processes. Guillermo Lopez Taboada identified and proposed and a fix for this. 18/3/2010

3. Fixed the bug which broke test cases when executed with odd number of processes. Many thanks to Guillermo Lopez Taboada for discovering this bug. 18/3/2010

4. Made changes to the MPJ Express runtime system and got rid of a couple of internal classes.

5. On some clusters, users complained that MPJ Express took abnormally long amount of time to bootstrap. We have fixed this bug in this release. 1/4/2010

6. Got rid of the Jetty dependency 9/4/2010

You may download the software from:

http://mpj-express.org

Feel free to contact us in case of any queries.

Best Regards,

Aamir Shafi
Bryan Carpenter
Mark Baker
Guillermo López Taboada

MPJ Express with CUDA

2010-02-10T23:15:00.000-08:00

One of the students at NUST did some work on exploiting GPUs from within the MPJ Express code. For more details, see http://bibrak.blogspot.com/2010/02/mpj-cuda.html

MPJ Express Version 0.35

2010-01-26T10:09:00.000-08:00

Dear MPJ Express Community,

We are pleased to announce the next beta release (0.35) of the MPJ
Express software. This release introduces some interesting and
innovative features in the MPJ Express software.

MPJ Express programs can now be executed in two configurations:

1. Multicore Configuration: This configuration is used by developers
who want to execute their parallel Java applications on multicore or
shared memory processors (laptops and desktops). This configuration is
preferable for development and teaching purposes.

2. Cluster Configuration: This configuration is used by developers who
want to execute their parallel Java applications on distributed memory
platforms including clusters and network of computers. This
configuration is preferable for executing large scale number crunching
applications.

You may begin by having a look at the talk "MPJ Express: An
Implementation of MPI in Java, Version: 0.35 (Beta Release)"
(http://mpj-express.org/docs/talks/mpjexpress_release0_35.ppt)

You may download the software from: http://mpj-express.org/download.php

Quick Start Guides:
----------------------------

README for Linux/UNIX/Mac: http://mpj-express.org/docs/readme/README
README for Windows: http://mpj-express.org/docs/readme/README-win.txt

Detailed User Guides:
-------------------------------

Linux User Guide: http://mpj-express.org/docs/guides/linuxguide.pdf
Windows User Guide: http://mpj-express.org/docs/guides/windowsguide.pdf

We are thankful to MPJ Express community of users for their continued
support and patience.

Feel free to contact us in case of any queries.

Best Regards,

Aamir Shafi
Bryan Carpenter
Mark Baker

MPJ Express on Infiniband

2009-08-28T22:07:00.000-07:00

Courtesy: Zafar Gillani (http://hpc.seecs.edu.pk/~zafar/about.html) who worked on porting MPJ Express on Infiniband.

MPJ Express can be executed on InfiniBand in couple of ways (so far):

Use IPoIB: For this just provide IP addresses in the "machines" file alloted to the InfiniBand HCAs. Run the MPJ Express as on GigE or FE.

Over SDP: First create a configuration file somewhere on the system (such as in ~/mpj-user/ directory). Lets say we created "sdp.conf". There are two approaches to this, either write bind rule or connect rule in "sdp.conf" file (as explained below). Comments are represented via a # sign.

bind
example: bind 192.168.2.1 *

connect
example: connect 192.168.2.1 *

* in port means it can use any free port on runtime. The port can be defined in range as well such as 15000-*. A separate bind rule has to be defined for each compute node that has configured IB HCA. This is sort of a machines file:

bind 192.168.2.1 *
bind 192.168.2.2 *
bind 192.168.2.3 *
and so on

When to use bind rule? SDP protocol transport should be used when a TCP socket binds to an address and port that matches the rule. In Java this is equivalent to socket.bind(SocketAddress). Bind rule is recommended since this will explicitly bind an unbound socket to an IP address using the SDP protocol.

When to use connect rule? SDP protocol transport should be used when an unbound TCP socket attempts to connect to an address and port that matches the rule.

To execute MPJ Express on InfiniBand simply use the following command:
mpjrun.sh -Dcom.sun.sdp.conf=sdp.conf -Djava.net.preferIPv4Stack=true -np 2 Application.

Switch -Djava.net.preferIPv4Stack=true is optional but is recommended since this will explicitly tell JVM to use IPv4. This prevents Java from using IPv6 if IPv6 is enabled on IB HCAs (InfiniBand Host Channel Adaptor analogous to NICs).

Nested Parallelism Using MPJ Express

2009-07-14T19:41:00.000-07:00

Steps to write a nested parallel Java application using MPJ Express and Java OpenMP (JOMP). JOMP can be downloaded from here.

Step 1: Write

Edit the HybridApp.jomp - the source code is shown in the Figure below. Here we rely on MPJ Express for node parallelism and JOMP for thread parallelism.

Step 2: Translate and Compile

Once we have written our application, we need to translate the .jomp file to .java. Later the Java file is compiled using the Java compiler.

The code is translated from .jomp to .java by using the command:

aamir@barq:~/tmp/jomp> java -cp $MPJ_HOME/lib/mpj.jar:jomp1.0b.jar jomp.compiler.Jomp HybridApp

Compile the Java application:

aamir@barq:~/tmp/jomp> javac -cp $MPJ_HOME/lib/mpj.jar:.:jomp1.0b.jar HybridApp.java

Step 3: Execute

Write machines file:

chenab1

chenab2
Start MPJ Express daemons:

aamir@barq:~/tmp/jomp> mpjboot machines
Run the hybrid code:

aamir@barq:~/tmp/jomp> mpjrun.sh -wdir ~/tmp/jomp -cp $MPJ_HOME/lib/mpj.jar:.:jomp1.0b.jar -np 2 -Djomp.threads=2 -dport 11050 HybridApp

Hello from process <1:0>
Hello from process <1:1>
Hello from process <0:0>
Hello from process <0:1>

Halt MPJ Express daemons:

aamir@barq:~/tmp/jomp> mpjhalt machines

Communicating Multi-dimensional arrays using MPJ Express

2009-05-12T21:23:00.000-07:00

Many scientific applications utilize multi-dimensional arrays for storing data. Naturally in the parallel versions, there is a frequent requirement of communicating these multi-dimensional arrays.

Currently the MPJ Express version "directly" supports communicating basic datatypes to and from single dimension arrays. But, it is obviously possible to communicate multiple dimension arrays. Here we'll see how 2D arrays are communicated using the MPJ Express software.

There are mainly two ways of communicating 2D arrays. The first is to communicate this data using the MPI.OBJECT datatype. The second is to map (or flatten) the 2D array onto a 1D array and communicate normally. If you are looking for performance, the second option is recommended. The main reason is performance, the first option is severely hampered by the performance of Java's serialization.

The MultidimMatrix class communicates data from the 2D array using the MPI.OBJECT datatype.

The MultiAsSingleDimMatrix class stores the 2D array as a 1D array. By doing this, the data can be communicated using the datatype of the array---in our example, the MPI.DOUBLE datatype.

Overhead of using Multi-dimensional arrays in Java

2009-05-08T02:41:00.000-07:00

As discussed many times in the related literature, the Java multidimensional arrays introduce a good amount of overhead because of the way they are stored in the memory. The purpose of this post is to quantify this performance overhead.

Let's first look at the matrix multiplication implementation that uses the multidimensional arrays (the 2D version).

Now let's look at the same code but implemented using single dimension array. Basically the two dimensional array is mapped onto the single dimension array (the 1D version).

Now put some performance results:

 
  aamir@barq:~/tmp> java MultidimMatrix
  time => 19.821604495
  aamir@barq:~/tmp> java Matrix
  time -> 11.928243662

The 1D code is 1.66 times faster than the 2D code.

Parallel Programming with Java

2009-05-05T20:19:00.000-07:00

I recently gave a talk on "Parallel Programming with Java" to Masters students at the University of La Coruna, Spain.

This talk gives a good introduction on how to get started with the MPJ Express software.

Ports Used by the MPJ Express Software

2009-05-04T20:25:00.000-07:00

There are three kinds of ports used by the MPJ Express software.

Daemons Ports (where MPJ Express daemons listen on compute-nodes). The value of this can be changed in two steps:

Edit $MPJ_HOME/conf/wrapper.conf and search for property "wrapper.app.parameter.2=10000". Change it to whatever you want- lets call it X.
Now you start your parallel application, use -dport switch to specify X. Some thing like this - mpjrun.sh ... -dport X ..

There are also ports used by MPJ Express runtime on the head-node. This is used to ship across the code. The default value of this is 15000. It can be changed by using the switch -sport Y - where `Y' is the port that you choose.
Each MPJ Express process uses a port for communication with peers. This can be changed by using -mpjport Z switch to mpjrun.sh - the default value is 20000.