News
Intel Boosts Parallel Programming
Intel updates Threading Building Blocks library, a developer resource for writing parallelized C++ code for multi-core CPUs.
At the O'Reilly Open Source Convention in July, Intel Corp. released an important update to its Threading Building Blocks (TBB) library, which helps Windows, Mac OS X and Linux developers write parallelized C++ code for advanced multi-core CPUs. For Windows users, a discreet plug-in enables TBB 2.1 to be set up inside any version of Visual Studio.
James Reinders, chief evangelist and director of marketing and sales in the Intel Software Solutions Group, says TBB 2.1 is "a significant improvement" over the previous version that can help developers focus on achieving parallelism.
"A lot of parallel programming in the past starts with thread management as the central concern. This distracts a developer from developing their algorithm and masks their thinking about parallelism," Reinders explains in an e-mail interview. "TBB automates all thread management, offers easy interfaces to express parallelism and then automatically matches the program at run time to the size of the system [or how many cores it has]."
In a statement, Intel touts performance and usability improvements in TBB 2.1. A new "affinity partitioner" class provides for soft task-to-thread affinity, allowing for canceled tasks and created threads independent of task-scheduler worker threads in TBB. Intel says the new class delivers important performance gains in a variety of situations.
Intel also provides container and interface improvements in response to customer requests. The updated Thread Checker promises to produce fewer false positives, according to the company, while the improved Thread Profiler will deliver more useful information for TBB-optimized applications.
About the Author
Michael Desmond is an editor and writer for 1105 Media's Enterprise Computing Group.