What’s New in Intel® Threading Building Blocks 3.0

By Terry Wilmarth (Intel) (10 posts) on May 4, 2010 at 6:20 am

After much hard work, extensive debate, prioritization, and efforts to eke out the last few percentages of performance gains, Intel® Threading Building Blocks (Intel® TBB) version 3.0 has arrived! With it come several new features and improvements, available now for download.

In the upcoming weeks, look for detailed blogs from the TBB developers on several of these topics, starting with Arch Robison who will discuss making the quick and painless transition to Intel® TBB 3.0. Intel® TBB 3.0 fully preserves binary compatibility and thus simple library substitution, even without re-linking, should work for most applications. Arch will discuss the very few exceptions to this rule in his blog.

Here’s a brief summary of the major changes since the release of version 2.2:

Extended compatibility and interoperability support
• Added support for Microsoft* Visual Studio* 2010
• Intel® TBB 3.0 can run on top of Microsoft* Concurrency Runtime on Microsoft* Windows* 7
• Added Microsoft* Parallel Patterns Library*-compatible critical_section and reader_writer_lock
• Apple* Snow Leopard* support
• C++0x feature support: Intel® TBB 3.0 now provides std::thread, std::lock_guard, std::unique_lock, and most of std::condition_variable

New and improved parallel pipeline
• Elegant new parallel_pipeline function provides a strongly-typed lambda-friendly pipeline interface

New and improved data containers
• New concurrent_unordered_map, an associative container that permits concurrent insertion and traversal with no visible locking; similar to C++0x std::unordered_map; based on a prototype developed at Microsoft for a future version of PPL
• Faster enumerable_thread_specific and combinable

Enhanced task scheduler features
• Fire and forget tasks: To start worker tasks from a GUI, process tasks in FIFO-like fashion and increase starvation-resistance of tasks, see the new task::enqueue method
• Independent task scheduling for foreign threads: the scheduler maps tasks to threads in a way that avoids causing a user-created thread to block directly or indirectly on another user-created thread
• A task_group_context can now be created and destroyed in different threads

Enhanced scalable memory allocator
• More reliable dynamic function substitution on Windows*
• Improved performance for large object allocation

New Design Patterns Manual
• Get the most out of Intel® TBB: This excellent resource, for everyone from the novice to the expert, illustrates the power of Intel® TBB for implementing common parallel programming design patterns

Categories: Parallel Programming, Software Tools
Tags: , , , ,

For more complete information about compiler optimizations, see our Optimization Notice.

Comments (5)

May 4, 2010 7:38 AM PDT

Terry Wilmarth (Intel)
Terry Wilmarth (Intel)Total Points:
1,260
Brown Belt
 Also check out James Reinders' excellent blog detailing the history of TBB leading up to the current release! It's here: http://software.intel.com/en-us/blogs/2010/05/04/tbb-30-new-..... ng-blocks/
May 4, 2010 9:18 AM PDT

Terry Wilmarth (Intel)
Terry Wilmarth (Intel)Total Points:
1,260
Brown Belt
 Arch's blog on Transitioning to TBB 3.0 is now available here: http://software.intel.com/en-us/blogs/2010/05/04/transitioning-to -tbb-30/
May 10, 2010 2:03 PM PDT


Bob Scheier 		Hello, Terry; my name is Bob Scheier, a free-lance writer doing a story for InfoWorld on what developers can do to get higher performance out of client-side applications (such as on Windows 7) from today's multicore chips and multithreading operating systems. TBB seems like a key enabling technology our audience should know about -- was wondering if we could schedule a phone chat or if I could send you some questions via email? Thanks in advance;

Bob
May 12, 2010 1:47 PM PDT

rolandr
rolandrTotal Points:
105
Registered User
 Reading your changes about this new task scheduler strategy I wonder whether my initially reported problem that
Robert Reed later followed up on in his blog: "Under the hood: Learning more about task scheduling" will now behave the way I'd prefer to have it. Or did I interpret this wrong?
May 14, 2010 4:32 PM PDT

Alexey Kukanov (Intel)
Alexey Kukanov (Intel)Total Points:
22,588
Black Belt
 ronaldr, I suggest you ask the question on the TBB forum; it's just more convenient place for discussions than blog comments. I would be glad to explain in more details how new fire-and-forget tasks are scheduled and executed, and help to match it to your use case if necessary.

Trackbacks (5)


Leave a comment  

To obtain technical support, please go to Software Support.
Name (required)*

Email (required; will not be displayed on this page)*

Your URL (optional)


Comment*