informa
/
2 min read
Commentary

Developing Parallel Apps in MATLAB Got Easier Today

You may not appreciate my pointing it out, but the biggest roadblock to advancement in high-performance computing is you.
That is, software development. Parallel programming is just plain hard. And while eleven and a half billion dollars was spent on HPC hardware last year, IDC analyst Jie Wu says that 'software development in this area will continue to be the number-one ...
You may not appreciate my pointing it out, but the biggest roadblock to advancement in high-performance computing is you.
That is, software development. Parallel programming is just plain hard. And while eleven and a half billion dollars was spent on HPC hardware last year, IDC analyst Jie Wu says that 'software development in this area will continue to be the number-one roadblock to further adoption of this  advanced hardware.'
So it's good news whenever somebody makes parallel software development easier.
Today, The MathWorks announced the inclusion of parallel computing capabilities in two of its MATLAB optimization toolboxes.
This should be a boon to scientists and engineers doing computationally-intensive software development using MATLAB. The new tools largely shield developers from the difficulties of parallel code development in these toolboxes. The toolboxes affected are the Optimization Toolbox and the Genetic Algorithm and Direct Search Toolbox; these have been integrated with the Parallel Computing Toolbox. Using the new capabilities, developers of parallel MATLAB apps for multicore computers and clusters can now solve computationally-intensive optimization problems without having to make significant changes to their code, the company said.
The Parallel Computing Toolbox extends the MATLAB language with a number of high-level parallel constructs, including parallel for-loops, distributed arrays, message-passing functions, and parallel numerical algorithms.