• September 2011
    M T W T F S S

Intel Code Lights Road to Many-Core Future

EE Times (09/15/11) Rick Merritt

 Intel’s release of open source code for a data-parallel version of Javascript seeks to help mainstream programmers who use scripting languages tap the power of multicore processors. Intel’s Justin Rattner says in an interview that there will be multiple programming models, and Parallel JS encompasses one such model. The language enhances performance for data-intensive, browser-based apps such as photo and video editing and three-dimensional gaming running on Intel chips. Rattner describes Parallel JS as “a pretty important step that gets us beyond the prevailing view that once you are beyond a few cores, multicore chips are only for technical apps.” A later iteration of Parallel JS also will exploit the graphics cores currently incorporated into Intel’s latest processors. In addition, Intel is working on ways to enhance modern data-parallel tools that operate general-purpose programs on graphics processors, and those tools could be issued next year, Rattner says. He notes that beyond that, data-parallel methods require a more basic change to boost power efficiency by becoming more asynchronous.


Parallel Programming Skills Crisis Could Stall Server Evolution

Computerworld Australia (09/12/11) Sandra Van Dijk

The lack of parallel programming expertise worldwide will become a major issue for the information technology (IT) industry over the next 10 years, warns a RMIT University report. Advanced parallel programming skills are hard to find among professional programmers because it is taught only as an advanced elective in most computer science curricula, says RMIT professor James Harland. The report warns that without a massive injection in parallel programming talent in the IT sector, new server platforms will stall, especially in areas such as fabric-based computing. The lack of programming skill can lead to frozen states commonly known as a deadly embrace, but techniques to deal with frozen states are only taught briefly in computer science and IT programs, according to Harland. “Hence, the only real way to overcome such obstacles is to spend more time on parallel programming techniques in these courses,” he notes. Parallel programming represents a different paradigm of software development in that it introduces the notion of complexity into software development, similar to learning a new programming language, says RMIT’s John Lenarcic. “Only a minority of developers have sufficient training to handle parallel programs, and only a fraction have enough experience to do it well,” says Gartner analyst Carl Claunch.