Newsletters March 2008

March 2008

Welcome to our first Newsletter of 2008. Much has happened in 2008 already, with January seeing South Africa waking up to a power crisis that will affect us for years to come. Like you, OPTI-NUM solutions is taking steps to mitigate the power crisis, with our offices running on significantly reduced lighting, and our staff doing a great job of saving energy where they can (but we’re all a little warmer with the air conditioners off or turned way down!)

This newsletter was delayed until March to coincide with the latest release of The MathWorks product family. R2008a is now available for download for customers current on Software Maintenance Service. Read more about this release in our special R2008a section.

As always, you will find information on our upcoming public seminars, public training courses, and webinars available to you at your desktop. With broadband access becoming accessible to more South Africans, webinars are an effective way to learn about technology from The MathWorks without having to leave your office. Of course, if you would like us to visit you, you can always contact us on 011 325 6238, or e-mail info@optinum.co.za.

Wishing you a prosperous 2008,

The OPTI-NUM solutions team. 

Special feature: R2008a

R2008a is the first DVD release of The MathWorks products for 2008. R2008a includes major new functionality in MATLAB, plus new features in Simulink and 72 other products!

We have selected a few key new features to discuss in this newsletter. Among them, the new Object-Oriented Programming (OOP) capabilities in the M-language, enhancements to Parallel Computing Toolbox and MATLAB Distributed Computing Server, and an overview of Simulink new features. For a full list of new features, and for more information on R2008a, visit the R2008a page on The MathWorks website.

Customers who are current on Software Maintenance Service as of 1 April 2008 will receive their R2008a DVD automatically. If your Software Maintenance Service is expired, contact us to renew today, by e-mailing sales@optinum.co.za.

Beginning with R2008a, all products in the MATLAB® and Simulink® product families will require activation. For more information, see the Activation Overview on The MathWorks web page.

Object-Oriented Programming in MATLAB

New object-oriented programming capabilities have been introduced in MATLAB R2008a, whereby you can define classes and apply standard object-oriented design patterns in MATLAB that let you benefit from code reuse, inheritance, encapsulation, and reference behaviour without engaging in the low-level housekeeping tasks required by other languages.

Classes in MATLAB can be defined as value or handle classes. A value class behaves like standard numeric and structure classes in MATLAB, while handle classes have a variable that identifies a particular instance (much like MATLAB’s Handle Graphics objects). Classes can inherit properties from a single or multiple classes such that you can organise your classes in to hierarchies.  With MATLAB’s new object-oriented programming capabilities, you can control aspects of property behaviours by setting property attributes and by defining property-specific access methods that determine what happens when a property value is specified or queried. Support for packages allows you more flexibility in scoping your new classes, while the built-in support for the new class behaviour in tools such as the editor, including M-Lint support JIT Accelerator support, make programming and using classes in MATLAB as easy as programming a procedural function.

Operator overloading of value classes allows you to build complex mathematical constructs like polynomial classes, and operate on them in a natural mathematical notation. Click here for an example of a polynomial class built using the R2008a object-oriented features.

For more information on object-oriented programming in MATLAB 7.6, see The MathWorks object-oriented programming page.

Enhancements for High-Performance Computing

R2008a includes many enhancements to The MathWorks support for high-performance computing (HPC). HPC is the use of multiple commercially available computers with interconnection systems, to solve (typically scientific) large-scale computing problems. With R2008a, MATLAB users get benefits across the spectrum of HPC use, from personal desktop systems to departmental and enterprise cluster systems.

On a desktop machine, MATLAB now enables multithreading by default, choosing the correct number of threads to use based on your processor architecture.

For scaling up problems to “personal HPC systems”, MATLAB users can use the PARFOR command available with the Parallel Computing Toolbox (previously known as the Distributed Computing Toolbox). The PARFOR command is the parallelised for loop.  Opening MATLAB pool jobs allows you to specify the configuration MATLAB is to use, or you can get MATLAB to find the optimal configuration for you.  Then PARFOR can make use of as many jobs as you have started, and return results much faster. For an example, see here.

Cluster users now have full Parallel Computing Toolbox support for PBS Pro® and TORQUE scheduler, or can use MATLAB’s built-in job manager scheduler. The size limits on files and data that can be transmitted between clients and workers have been significantly increased.  The new batch command provides a simple way to run scripts on a cluster, and in the background of multi-core computers. All this is managed by the new Administration GUI, which will guide you through testing your cluster and diagnosing possible failures.
Built-in parallel support has also been added to the following Toolboxes:
Optimisation Toolbox

• Large scale optimisation problems can be broken up and parallelised

• Gradient estimation in non-linear solvers has been significantly accelerated by parallelisation

Genetic Algorithm and Direct Search Toolbox (GADS)

• Genetic algorithms and pattern searches can simultaneously explore the solution space

Model Based Calibration Toolbox

• Fit multiple models to experimental data at the same time with parallel support 
  

New Features in Simulink Product Family

The Simulink Product family includes many incremental improvements since the major update shipped with R2007b, including:

• There is a new library browser supported on all platforms. This included improved performance for browsing and searching.
• The Model Advisor has an enhanced GUI as well as improved reporting functionality. Together with Simulink® Verification and Validation additional standards checks, such as IEC 61508 have been added.
• "What's This?" Context-Sensitive Help for Simulink® Configuration Parameters Dialog allow for quick access to a detailed description of the parameters
• Improvement to rapid accelerator performance for long simulations of small models on the Microsoft® Windows® platform.
• An additional zero crossing algorithm has been added especially useful in systems exhibiting strong chattering behaviour
• Now changing a referenced model that executes in Accelerator mode or is used for code generation triggers rebuilding a parent model only when the change directly affects the referenced model's interface to the parent model.
• Embedded MATLAB™ Function blocks now support nontunable MATLAB structure parameters.
• You can now navigate between a line of generated code and its corresponding line of source code in Embedded MATLAB Function blocks.
• Code checking of Embedded MATLAB™ language subset functions in the M-Lint Code Analyzer and Simulink® Design Verifier™
• With the new auto save option, a backup copy of the model is made before updating or simulating. The model dependencies tools have also been extended.
• The Lookup Table (n-D) block now supports all data types including complex data and non scalar inputs.
  
  

For a full list of new features in the Simulink product family, visit the R2008a page on The MathWorks website.

Events

Upcoming Training Courses

OPTI-NUM solutions regularly holds paid, public training courses covering the use of The MathWorks products, as well as providing technical training on a variety of subjects associated with The MathWorks products. Take advantage of our discounts on multiple training bookings and save on MATLAB and Simulink training! If you purchase more for than 5 days of training, you become eligible for a discount, which increases with the number of training days you purchase. If you are a student or lecturer then you can also receive an additional academic discount.

OPTI-NUM solutions' next 5 standard training courses are shown below.

14 February 2012	MATLAB Fundamentals
07 March 2012	Simulink for System and Algorithm Modelling
08 May 2012	MATLAB Fundamentals
29 May 2012	Advanced MATLAB Programming Techniques
19 June 2012	MATLAB for Building Graphical User Interfaces

View our full training schedule.

Upcoming Seminars

16 February 2012	Cape Town	Speeding up Analysis of Excel Data
14 March 2012	Pretoria	Implementing MATLAB and Simulink Algorithms on FPGAs
16 March 2012	Cape Town	Implementing MATLAB and Simulink Algorithms on FPGAs

 

View all upcoming events.

Upcoming Webinars

07 February 2012	Generating C and C++ Code from MATLAB using MATLAB Coder
09 February 2012	Algorithm Development with MATLAB