What's New in TBB 4.0?

Flow Graph - Flexible and convenient API for expressing static and dynamic dependencies between computations. Also extends applicability of Intel(R) TBB to event-driven/reactive programming models.

Task and task group priorities - Provides ability to specify task execution order based on three priority levels (low, normal, and high). Static priorities are supported for enqueued tasks; dynamic priorities are supported for task groups. 

Memory Pools - Enables Intel(R) TBB memory allocator to work on user-provided memory regions. This mechanism allows greater flexibility and performance by getting thread-safe and scalable object allocation in the application-specific memory blocks with custom life-span and growth policies. 

Concurrent Priority Queue - A queue that allows pulling data out in a user-defined priority order. The concurrent priority queue is useful when a certain execution order is enforced by priority relationship between parallel tasks and/or data.

Concurrent Unordered Set - New thread-safe container to store and access user objects using a hash key. Implements a concurrent variation of a standard C++ class std::unordered_set. 

Generic GCC* Atomics Support - Greater library portability which enables the user to develop Intel(R) TBB-based solutions on a broader range of platforms. The user now has a choice to use built-in GCC* atomics routines instead of supplied platform-specific ones. 

New Examples - New examples demonstrate usage of major new features:

• Shortest path (concurrent_priority_queue) 
• Dining philosophers, binpack (flow graph)
• Mandelbrot fractal (task priority)