CHANGES

------------------------------------------------------------------------
The list of most significant changes made over time in
Intel(R) Threading Building Blocks (Intel(R) TBB).

Intel TBB 2017 Update 3
TBB_INTERFACE_VERSION == 9103

Changes (w.r.t. Intel TBB 2017 Update 2):

- Added support for Android* 7.0 and Android* NDK r13, r13b.

Preview Features:

- Added template class gfx_factory to the flow graph API. It implements
    the Factory concept for streaming_node to offload computations to
    Intel(R) processor graphics.

Bugs fixed:

- Fixed a possible deadlock caused by missed wakeup signals in
    task_arena::execute().

------------------------------------------------------------------------
Intel TBB 2017 Update 2
TBB_INTERFACE_VERSION == 9102

Changes (w.r.t. Intel TBB 2017 Update 1):

- Removed the long-outdated support for Xbox* consoles.

Bugs fixed:

- Fixed the issue with task_arena::execute() not being processed when
    the calling thread cannot join the arena.
- Fixed dynamic memory allocation replacement failure on macOS* 10.12.

------------------------------------------------------------------------
Intel TBB 2017 Update 1
TBB_INTERFACE_VERSION == 9101

Changes (w.r.t. Intel TBB 2017):

Bugs fixed:

- Fixed dynamic memory allocation replacement failures on Windows* 10
    Anniversary Update.
- Fixed emplace() method of concurrent unordered containers to not
    require a copy constructor.

------------------------------------------------------------------------
Intel TBB 2017
TBB_INTERFACE_VERSION == 9100

Changes (w.r.t. Intel TBB 4.4 Update 5):

- static_partitioner class is now a fully supported feature.
- async_node class is now a fully supported feature.
- Improved dynamic memory allocation replacement on Windows* OS to skip
    DLLs for which replacement cannot be done, instead of aborting.
- Intel TBB no longer performs dynamic memory allocation replacement
    for Microsoft* Visual Studio* 2008.
- For 64-bit platforms, quadrupled the worst-case limit on the amount
    of memory the Intel TBB allocator can handle.
- Added TBB_USE_GLIBCXX_VERSION macro to specify the version of GNU
    libstdc++ when it cannot be properly recognized, e.g. when used
    with Clang on Linux* OS. Inspired by a contribution from David A.
- Added graph/stereo example to demostrate tbb::flow::async_msg.
- Removed a few cases of excessive user data copying in the flow graph.
- Reworked split_node to eliminate unnecessary overheads.
- Added support for C++11 move semantics to the argument of
    tbb::parallel_do_feeder::add() method.
- Added C++11 move constructor and assignment operator to
    tbb::combinable template class.
- Added tbb::this_task_arena::max_concurrency() function and
    max_concurrency() method of class task_arena returning the maximal
    number of threads that can work inside an arena.
- Deprecated tbb::task_arena::current_thread_index() static method;
    use tbb::this_task_arena::current_thread_index() function instead.
- All examples for commercial version of library moved online:
    https://software.intel.com/en-us/product-code-samples. Examples are
    available as a standalone package or as a part of Intel(R) Parallel
    Studio XE or Intel(R) System Studio Online Samples packages.

Changes affecting backward compatibility:

- Renamed following methods and types in async_node class:
    Old                   New
    async_gateway_type => gateway_type
    async_gateway()    => gateway()
    async_try_put()    => try_put()
    async_reserve()    => reserve_wait()
    async_commit()     => release_wait()
- Internal layout of some flow graph nodes has changed; recompilation
    is recommended for all binaries that use the flow graph.

Preview Features:

- Added template class streaming_node to the flow graph API. It allows
    a flow graph to offload computations to other devices through
    streaming or offloading APIs.
- Template class opencl_node reimplemented as a specialization of
    streaming_node that works with OpenCL*.
- Added tbb::this_task_arena::isolate() function to isolate execution
    of a group of tasks or an algorithm from other tasks submitted
    to the scheduler.

Bugs fixed:

- Added a workaround for GCC bug #62258 in std::rethrow_exception()
    to prevent possible problems in case of exception propagation.
- Fixed parallel_scan to provide correct result if the initial value
    of an accumulator is not the operation identity value.
- Fixed a memory corruption in the memory allocator when it meets
    internal limits.
- Fixed the memory allocator on 64-bit platforms to align memory
    to 16 bytes by default for all allocations bigger than 8 bytes.
- As a workaround for crashes in the Intel TBB library compiled with
    GCC 6, added -flifetime-dse=1 to compilation options on Linux* OS.
- Fixed a race in the flow graph implementation.

Open-source contributions integrated:

- Enabling use of C++11 'override' keyword by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.4 Update 6
TBB_INTERFACE_VERSION == 9006

Changes (w.r.t. Intel TBB 4.4 Update 5):

- For 64-bit platforms, quadrupled the worst-case limit on the amount
    of memory the Intel TBB allocator can handle.

Bugs fixed:

- Fixed a memory corruption in the memory allocator when it meets
    internal limits.
- Fixed the memory allocator on 64-bit platforms to align memory
    to 16 bytes by default for all allocations bigger than 8 bytes.
- Fixed parallel_scan to provide correct result if the initial value
    of an accumulator is not the operation identity value.
- As a workaround for crashes in the Intel TBB library compiled with
    GCC 6, added -flifetime-dse=1 to compilation options on Linux* OS.

------------------------------------------------------------------------
Intel TBB 4.4 Update 5
TBB_INTERFACE_VERSION == 9005

Changes (w.r.t. Intel TBB 4.4 Update 4):

- Modified graph/fgbzip2 example to remove unnecessary data queuing.

Preview Features:

- Added a Python* module which is able to replace Python's thread pool
    class with the implementation based on Intel TBB task scheduler.

Bugs fixed:

- Fixed the implementation of 64-bit tbb::atomic for IA-32 architecture
    to work correctly with GCC 5.2 in C++11/14 mode.
- Fixed a possible crash when tasks with affinity (e.g. specified via
    affinity_partitioner) are used simultaneously with task priority
    changes.

------------------------------------------------------------------------
Intel TBB 4.4 Update 4
TBB_INTERFACE_VERSION == 9004

Changes (w.r.t. Intel TBB 4.4 Update 3):

- Removed a few cases of excessive user data copying in the flow graph.
- Improved robustness of concurrent_bounded_queue::abort() in case of
    simultaneous push and pop operations.

Preview Features:

- Added tbb::flow::async_msg, a special message type to support
    communications between the flow graph and external asynchronous
    activities.
- async_node modified to support use with C++03 compilers.

Bugs fixed:

- Fixed a bug in dynamic memory allocation replacement for Windows* OS.
- Fixed excessive memory consumption on Linux* OS caused by enabling
    zero-copy realloc.
- Fixed performance regression on Intel(R) Xeon Phi(tm) coprocessor with
    auto_partitioner.

------------------------------------------------------------------------
Intel TBB 4.4 Update 3
TBB_INTERFACE_VERSION == 9003

Changes (w.r.t. Intel TBB 4.4 Update 2):

- Modified parallel_sort to not require a default constructor for values
    and to use iter_swap() for value swapping.
- Added support for creating or initializing a task_arena instance that
    is connected to the arena currently used by the thread.
- graph/binpack example modified to use multifunction_node.
- For performance analysis, use Intel(R) VTune(TM) Amplifier XE 2015
    and higher; older versions are no longer supported.
- Improved support for compilation with disabled RTTI, by omitting its use
    in auxiliary code, such as assertions. However some functionality,
    particularly the flow graph, does not work if RTTI is disabled.
- The tachyon example for Android* can be built using Android Studio 1.5
    and higher with experimental Gradle plugin 0.4.0.

Preview Features:

- Added class opencl_subbufer that allows using OpenCL* sub-buffer
    objects with opencl_node.
- Class global_control supports the value of 1 for
    max_allowed_parallelism.

Bugs fixed:

- Fixed a race causing "TBB Warning: setaffinity syscall failed" message.
- Fixed a compilation issue on OS X* with Intel(R) C++ Compiler 15.0.
- Fixed a bug in queuing_rw_mutex::downgrade() that could temporarily
    block new readers.
- Fixed speculative_spin_rw_mutex to stop using the lazy subscription
    technique due to its known flaws.
- Fixed memory leaks in the tool support code.

------------------------------------------------------------------------
Intel TBB 4.4 Update 2
TBB_INTERFACE_VERSION == 9002

Changes (w.r.t. Intel TBB 4.4 Update 1):

- Improved interoperability with Intel(R) OpenMP RTL (libiomp) on Linux:
    OpenMP affinity settings do not affect the default number of threads
    used in the task scheduler. Intel(R) C++ Compiler 16.0 Update 1
    or later is required.
- Added a new flow graph example with different implementations of the
    Cholesky Factorization algorithm.

Preview Features:

- Added template class opencl_node to the flow graph API. It allows a
    flow graph to offload computations to OpenCL* devices.
- Extended join_node to use type-specified message keys. It simplifies
    the API of the node by obtaining message keys via functions
    associated with the message type (instead of node ports).
- Added static_partitioner that minimizes overhead of parallel_for and
    parallel_reduce for well-balanced workloads.
- Improved template class async_node in the flow graph API to support
    user settable concurrency limits.

Bugs fixed:

- Fixed a possible crash in the GUI layer for library examples on Linux.

------------------------------------------------------------------------
Intel TBB 4.4 Update 1
TBB_INTERFACE_VERSION == 9001

Changes (w.r.t. Intel TBB 4.4):

- Added support for Microsoft* Visual Studio* 2015.
- Intel TBB no longer performs dynamic replacement of memory allocation
    functions for Microsoft Visual Studio 2005 and earlier versions.
- For GCC 4.7 and higher, the intrinsics-based platform isolation layer
    uses __atomic_* built-ins instead of the legacy __sync_* ones.
    This change is inspired by a contribution from Mathieu Malaterre.
- Improvements in task_arena:
    Several application threads may join a task_arena and execute tasks
    simultaneously. The amount of concurrency reserved for application
    threads at task_arena construction can be set to any value between
    0 and the arena concurrency limit.
- The fractal example was modified to demonstrate class task_arena
    and moved to examples/task_arena/fractal.

Bugs fixed:

- Fixed a deadlock during destruction of task_scheduler_init objects
    when one of destructors is set to wait for worker threads.
- Added a workaround for a possible crash on OS X* when dynamic memory
    allocator replacement (libtbbmalloc_proxy) is used and memory is
    released during application startup.
- Usage of mutable functors with task_group::run_and_wait() and
    task_arena::enqueue() is disabled. An attempt to pass a functor
    which operator()() is not const will produce compilation errors.
- Makefiles and environment scripts now properly recognize GCC 5.0 and
    higher.

Open-source contributions integrated:

- Improved performance of parallel_for_each for inputs allowing random
    access, by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.4
TBB_INTERFACE_VERSION == 9000

Changes (w.r.t. Intel TBB 4.3 Update 6):

- The following features are now fully supported:
    tbb::flow::composite_node;
    additional policies of tbb::flow::graph_node::reset().
- Platform abstraction layer for Windows* OS updated to use compiler
    intrinsics for most atomic operations.
- The tbb/compat/thread header updated to automatically include
    C++11 <thread> where available.
- Fixes and refactoring in the task scheduler and class task_arena.
- Added key_matching policy to tbb::flow::join_node, which removes
    the restriction on the type that can be compared-against.
- For tag_matching join_node, tag_value is redefined to be 64 bits
    wide on all architectures.
- Expanded the documentation for the flow graph with details about
    node semantics and behavior.
- Added dynamic replacement of C11 standard function aligned_alloc()
    under Linux* OS.
- Added C++11 move constructors and assignment operators to
    tbb::enumerable_thread_specific container.
- Added hashing support for tbb::tbb_thread::id.
- On OS X*, binaries that depend on libstdc++ are not provided anymore.
    In the makefiles, libc++ is now used by default; for building with
    libstdc++, specify stdlib=libstdc++ in the make command line.

Preview Features:

- Added a new example, graph/fgbzip2, that shows usage of
    tbb::flow::async_node.
- Modification to the low-level API for memory pools:
    added a function for finding a memory pool by an object allocated
    from that pool.
- tbb::memory_pool now does not request memory till the first allocation
    from the pool.

Changes affecting backward compatibility:

- Internal layout of flow graph nodes has changed; recompilation is
    recommended for all binaries that use the flow graph.
- Resetting a tbb::flow::source_node will immediately activate it,
    unless it was created in inactive state.

Bugs fixed:

- Failure at creation of a memory pool will not cause process
    termination anymore.

Open-source contributions integrated:

- Supported building TBB with Clang on AArch64 with use of built-in
    intrinsics by David A.

------------------------------------------------------------------------
Intel TBB 4.3 Update 6
TBB_INTERFACE_VERSION == 8006

Changes (w.r.t. Intel TBB 4.3 Update 5):

- Supported zero-copy realloc for objects >1MB under Linux* via
    mremap system call.
- C++11 move-aware insert and emplace methods have been added to
    concurrent_hash_map container.
- install_name is set to @rpath/<library name> on OS X*.

Preview Features:

- Added template class async_node to the flow graph API. It allows a
    flow graph to communicate with an external activity managed by
    the user or another runtime.
- Improved speed of flow::graph::reset() clearing graph edges.
    rf_extract flag has been renamed rf_clear_edges.
- extract() method of graph nodes now takes no arguments.

Bugs fixed:

- concurrent_unordered_{set,map} behaves correctly for degenerate
    hashes.
- Fixed a race condition in the memory allocator that may lead to
    excessive memory consumption under high multithreading load.

------------------------------------------------------------------------
Intel TBB 4.3 Update 5
TBB_INTERFACE_VERSION == 8005

Changes (w.r.t. Intel TBB 4.3 Update 4):

- Added add_ref_count() method of class tbb::task.

Preview Features:

- Added class global_control for application-wide control of allowed
    parallelism and thread stack size.
- memory_pool_allocator now throws the std::bad_alloc exception on
    allocation failure.
- Exceptions thrown for by memory pool constructors changed from
    std::bad_alloc to std::invalid_argument and std::runtime_error.

Bugs fixed:

- scalable_allocator now throws the std::bad_alloc exception on
    allocation failure.
- Fixed a race condition in the memory allocator that may lead to
    excessive memory consumption under high multithreading load.
- A new scheduler created right after destruction of the previous one
    might be unable to modify the number of worker threads.

Open-source contributions integrated:

- (Added but not enabled) push_front() method of class tbb::task_list
    by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.3 Update 4
TBB_INTERFACE_VERSION == 8004

Changes (w.r.t. Intel TBB 4.3 Update 3):

- Added a C++11 variadic constructor for enumerable_thread_specific.
    The arguments from this constructor are used to construct
    thread-local values.
- Improved exception safety for enumerable_thread_specific.
- Added documentation for tbb::flow::tagged_msg class and
    tbb::flow::output_port function.
- Fixed build errors for systems that do not support dynamic linking.
- C++11 move-aware insert and emplace methods have been added to
    concurrent unordered containers.

Preview Features:

- Interface-breaking change: typedefs changed for node predecessor and
    successor lists, affecting copy_predecessors and copy_successors
    methods.
- Added template class composite_node to the flow graph API. It packages
    a subgraph to represent it as a first-class flow graph node.
- make_edge and remove_edge now accept multiport nodes as arguments,
    automatically using the node port with index 0 for an edge.

Open-source contributions integrated:

- Draft code for enumerable_thread_specific constructor with multiple
    arguments (see above) by Adrien Guinet.
- Fix for GCC invocation on IBM* Blue Gene*
    by Jeff Hammond and Raf Schietekat.
- Extended testing with smart pointers for Clang & libc++
    by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.3 Update 3
TBB_INTERFACE_VERSION == 8003

Changes (w.r.t. Intel TBB 4.3 Update 2):

- Move constructor and assignment operator were added to unique_lock.

Preview Features:

- Time overhead for memory pool destruction was reduced.

Open-source contributions integrated:

- Build error fix for iOS* by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.3 Update 2
TBB_INTERFACE_VERSION == 8002

Changes (w.r.t. Intel TBB 4.3 Update 1):

- Binary files for 64-bit Android* applications were added as part of the
    Linux* OS package.
- Exact exception propagation is enabled for Intel C++ Compiler on OS X*.
- concurrent_vector::shrink_to_fit was optimized for types that support
    C++11 move semantics.

Bugs fixed:

- Fixed concurrent unordered containers to insert elements much faster
    in debug mode.
- Fixed concurrent priority queue to support types that do not have
    copy constructors.
- Fixed enumerable_thread_specific to forbid copying from an instance
    with a different value type.

Open-source contributions integrated:

- Support for PathScale* EKOPath* Compiler by Erik Lindahl.

------------------------------------------------------------------------
Intel TBB 4.3 Update 1
TBB_INTERFACE_VERSION == 8001

Changes (w.r.t. Intel TBB 4.3):

- The ability to split blocked_ranges in a proportion, used by
    affinity_partitioner since version 4.2 Update 4, became a formal
    extension of the Range concept.
- More checks for an incorrect address to release added to the debug
    version of the memory allocator.
- Different kind of solutions for each TBB example were merged.

Preview Features:

- Task priorities are re-enabled in preview binaries.

Bugs fixed:

- Fixed a duplicate symbol when TBB_PREVIEW_VARIADIC_PARALLEL_INVOKE is
    used in multiple compilation units.
- Fixed a crash in __itt_fini_ittlib seen on Ubuntu 14.04.
- Fixed a crash in memory release after dynamic replacement of the
    OS X* memory allocator.
- Fixed incorrect indexing of arrays in seismic example.
- Fixed a data race in lazy initialization of task_arena.

Open-source contributions integrated:

- Fix for dumping information about gcc and clang compiler versions
    by Misty De Meo.

------------------------------------------------------------------------
Intel TBB 4.3
TBB_INTERFACE_VERSION == 8000

Changes (w.r.t. Intel TBB 4.2 Update 5):

- The following features are now fully supported: flow::indexer_node,
    task_arena, speculative_spin_rw_mutex.
- Compatibility with C++11 standard improved for tbb/compat/thread
    and tbb::mutex.
- C++11 move constructors have been added to concurrent_queue and
    concurrent_bounded_queue.
- C++11 move constructors and assignment operators have been added to
    concurrent_vector, concurrent_hash_map, concurrent_priority_queue,
    concurrent_unordered_{set,multiset,map,multimap}.
- C++11 move-aware emplace/push/pop methods have been added to
    concurrent_vector, concurrent_queue, concurrent_bounded_queue,
    concurrent_priority_queue.
- Methods to insert a C++11 initializer list have been added:
    concurrent_vector::grow_by(), concurrent_hash_map::insert(),
    concurrent_unordered_{set,multiset,map,multimap}::insert().
- Testing for compatibility of containers with some C++11 standard
    library types has been added.
- Dynamic replacement of standard memory allocation routines has been
    added for OS X*.
- Microsoft* Visual Studio* projects for Intel TBB examples updated
    to VS 2010.
- For open-source packages, debugging information (line numbers) in
    precompiled binaries now matches the source code.
- Debug information was added to release builds for OS X*, Solaris*,
    FreeBSD* operating systems and MinGW*.
- Various improvements in documentation, debug diagnostics and examples.

Preview Features:

- Additional actions on reset of graphs, and extraction of individual
    nodes from a graph (TBB_PREVIEW_FLOW_GRAPH_FEATURES).
- Support for an arbitrary number of arguments in parallel_invoke
   (TBB_PREVIEW_VARIADIC_PARALLEL_INVOKE).

Changes affecting backward compatibility:

- For compatibility with C++11 standard, copy and move constructors and
    assignment operators are disabled for all mutex classes. To allow
    the old behavior, use TBB_DEPRECATED_MUTEX_COPYING macro.
- flow::sequencer_node rejects messages with repeating sequence numbers.
- Changed internal interface between tbbmalloc and tbbmalloc_proxy.
- Following deprecated functionality has been removed:
    old debugging macros TBB_DO_ASSERT & TBB_DO_THREADING_TOOLS;
    no-op depth-related methods in class task;
    tbb::deprecated::concurrent_queue;
    deprecated variants of concurrent_vector methods.
- register_successor() and remove_successor() are deprecated as methods
    to add and remove edges in flow::graph; use make_edge() and
    remove_edge() instead.

Bugs fixed:

- Fixed incorrect scalable_msize() implementation for aligned objects.
- Flow graph buffering nodes now destroy their copy of forwarded items.
- Multiple fixes in task_arena implementation, including for:
    inconsistent task scheduler state inside executed functions;
    incorrect floating-point settings and exception propagation;
    possible stalls in concurrent invocations of execute().
- Fixed floating-point settings propagation when the same instance of
    task_group_context is used in different arenas.
- Fixed compilation error in pipeline.h with Intel Compiler on OS X*.
- Added missed headers for individual components to tbb.h.

Open-source contributions integrated:

- Range interface addition to parallel_do, parallel_for_each and
    parallel_sort by Stephan Dollberg.
- Variadic template implementation of parallel_invoke
    by Kizza George Mbidde (see Preview Features).
- Improvement in Seismic example for MacBook Pro* with Retina* display
    by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.2 Update 5
TBB_INTERFACE_VERSION == 7005

Changes (w.r.t. Intel TBB 4.2 Update 4):

- The second template argument of class aligned_space<T,N> now is set
    to 1 by default.

Preview Features:

- Better support for exception safety, task priorities and floating
    point settings in class task_arena.
- task_arena::current_slot() has been renamed to
    task_arena::current_thread_index().

Bugs fixed:

- Task priority change possibly ignored by a worker thread entering
    a nested parallel construct.
- Memory leaks inside the task scheduler when running on
    Intel(R) Xeon Phi(tm) coprocessor.

Open-source contributions integrated:

- Improved detection of X Window support for Intel TBB examples
    and other feedback by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.2 Update 4
TBB_INTERFACE_VERSION == 7004

Changes (w.r.t. Intel TBB 4.2 Update 3):

- Added possibility to specify floating-point settings at invocation
    of most parallel algorithms (including flow::graph) via
    task_group_context.
- Added dynamic replacement of malloc_usable_size() under
    Linux*/Android* and dlmalloc_usable_size() under Android*.
- Added new methods to concurrent_vector:
    grow_by() that appends a sequence between two given iterators;
    grow_to_at_least() that initializes new elements with a given value.
- Improved affinity_partitioner for better performance on balanced
    workloads.
- Improvements in the task scheduler, including better scalability
    when threads search for a task arena, and better diagnostics.
- Improved allocation performance for workloads that do intensive
    allocation/releasing of same-size objects larger than ~8KB from
    multiple threads.
- Exception support is enabled by default for 32-bit MinGW compilers.
- The tachyon example for Android* can be built for all targets
    supported by the installed NDK.
- Added Windows Store* version of the tachyon example.
- GettingStarted/sub_string_finder example ported to offload execution
    on Windows* for Intel(R) Many Integrated Core Architecture.

Preview Features:

- Removed task_scheduler_observer::on_scheduler_leaving() callback.
- Added task_scheduler_observer::may_sleep() callback.
- The CPF or_node has been renamed indexer_node. The input to
    indexer_node is now a list of types. The output of indexer_node is
    a tagged_msg type composed of a tag and a value. For indexer_node,
    the tag is a size_t.

Bugs fixed:

- Fixed data races in preview extensions of task_scheduler_observer.
- Added noexcept(false) for destructor of task_group_base to avoid
    crash on cancellation of structured task group in C++11.

Open-source contributions integrated:

- Improved concurrency detection for BG/Q, and other improvements
    by Raf Schietekat.
- Fix for crashes in enumerable_thread_specific in case if a contained
    object is too big to be constructed on the stack by Adrien Guinet.

------------------------------------------------------------------------
Intel TBB 4.2 Update 3
TBB_INTERFACE_VERSION == 7003

Changes (w.r.t. Intel TBB 4.2 Update 2):

- Added support for Microsoft* Visual Studio* 2013.
- Improved Microsoft* PPL-compatible form of parallel_for for better
    support of auto-vectorization.
- Added a new example for cancellation and reset in the flow graph:
    Kohonen self-organizing map (examples/graph/som).
- Various improvements in source code, tests, and makefiles.

Bugs fixed:

- Added dynamic replacement of _aligned_msize() previously missed.
- Fixed task_group::run_and_wait() to throw invalid_multiple_scheduling
    exception if the specified task handle is already scheduled.

Open-source contributions integrated:

- A fix for ARM* processors by Steve Capper.
- Improvements in std::swap calls by Robert Maynard.

------------------------------------------------------------------------
Intel TBB 4.2 Update 2
TBB_INTERFACE_VERSION == 7002

Changes (w.r.t. Intel TBB 4.2 Update 1):

- Enable C++11 features for Microsoft* Visual Studio* 2013 Preview.
- Added a test for compatibility of TBB containers with C++11
    range-based for loop.

Changes affecting backward compatibility:

- Internal layout changed for class tbb::flow::limiter_node.

Preview Features:

- Added speculative_spin_rw_mutex, a read-write lock class which uses
    Intel(R) Transactional Synchronization Extensions.

Bugs fixed:

- When building for Intel(R) Xeon Phi(tm) coprocessor, TBB programs
    no longer require explicit linking with librt and libpthread.

Open-source contributions integrated:

- Fixes for ARM* processors by Steve Capper, Leif Lindholm
    and Steven Noonan.
- Support for Clang on Linux by Raf Schietekat.
- Typo correction in scheduler.cpp by Julien Schueller.

------------------------------------------------------------------------
Intel TBB 4.2 Update 1
TBB_INTERFACE_VERSION == 7001

Changes (w.r.t. Intel TBB 4.2):

- Added project files for Microsoft* Visual Studio* 2010.
- Initial support of Microsoft* Visual Studio* 2013 Preview.
- Enable C++11 features available in Intel(R) C++ Compiler 14.0.
- scalable_allocation_mode(TBBMALLOC_SET_SOFT_HEAP_LIMIT, <size>) can be
    used to urge releasing memory from tbbmalloc internal buffers when
    the given limit is exceeded.

Preview Features:

- Class task_arena no longer requires linking with a preview library,
    though still remains a community preview feature.
- The method task_arena::wait_until_empty() is removed.
- The method task_arena::current_slot() now returns -1 if
    the task scheduler is not initialized in the thread.

Changes affecting backward compatibility:

- Because of changes in internal layout of graph nodes, the namespace
    interface number of flow::graph has been incremented from 6 to 7.

Bugs fixed:

- Fixed a race in lazy initialization of task_arena.
- Fixed flow::graph::reset() to prevent situations where tasks would be
    spawned in the process of resetting the graph to its initial state.
- Fixed decrement bug in limiter_node.
- Fixed a race in arc deletion in the flow graph.

Open-source contributions integrated:

- Improved support for IBM* Blue Gene* by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.2
TBB_INTERFACE_VERSION == 7000

Changes (w.r.t. Intel TBB 4.1 Update 4):

- Added speculative_spin_mutex, which uses Intel(R) Transactional
    Synchronization Extensions when they are supported by hardware.
- Binary files linked with libc++ (the C++ standard library in Clang)
    were added on OS X*.
- For OS X* exact exception propagation is supported with Clang;
    it requires use of libc++ and corresponding Intel TBB binaries.
- Support for C++11 initializer lists in constructor and assigment
    has been added to concurrent_hash_map, concurrent_unordered_set,
    concurrent_unordered_multiset, concurrent_unordered_map,
    concurrent_unordered_multimap.
- The memory allocator may now clean its per-thread memory caches
    when it cannot get more memory.
- Added the scalable_allocation_command() function for on-demand
    cleaning of internal memory caches.
- Reduced the time overhead for freeing memory objects smaller than ~8K.
- Simplified linking with the debug library for applications that use
    Intel TBB in code offloaded to Intel(R) Xeon Phi(tm) coprocessors.
    See an example in
    examples/GettingStarted/sub_string_finder/Makefile.
- Various improvements in source code, scripts and makefiles.

Changes affecting backward compatibility:

- tbb::flow::graph has been modified to spawn its tasks;
    the old behaviour (task enqueuing) is deprecated. This change may
    impact applications that expected a flow graph to make progress
    without calling wait_for_all(), which is no longer guaranteed. See
    the documentation for more details.
- Changed the return values of the scalable_allocation_mode() function.

Bugs fixed:

- Fixed a leak of parallel_reduce body objects when execution is
    cancelled or an exception is thrown, as suggested by Darcy Harrison.
- Fixed a race in the task scheduler which can lower the effective
    priority despite the existence of higher priority tasks.
- On Linux an error during destruction of the internal thread local
    storage no longer results in an exception.

Open-source contributions integrated:

- Fixed task_group_context state propagation to unrelated context trees
    by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.1 Update 4
TBB_INTERFACE_VERSION == 6105

Changes (w.r.t. Intel TBB 4.1 Update 3):

- Use /volatile:iso option with VS 2012 to disable extended
    semantics for volatile variables.
- Various improvements in affinity_partitioner, scheduler,
    tests, examples, makefiles.
- Concurrent_priority_queue class now supports initialization/assignment
    via C++11 initializer list feature (std::initializer_list<T>).

Bugs fixed:

- Fixed more possible stalls in concurrent invocations of
    task_arena::execute(), especially waiting for enqueued tasks.
- Fixed requested number of workers for task_arena(P,0).
- Fixed interoperability with Intel(R) VTune(TM) Amplifier XE in
    case of using task_arena::enqueue() from a terminating thread.

Open-source contributions integrated:

- Type fixes, cleanups, and code beautification by Raf Schietekat.
- Improvements in atomic operations for big endian platforms
    by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 4.1 Update 3
TBB_INTERFACE_VERSION == 6103

Changes (w.r.t. Intel TBB 4.1 Update 2):

- Binary files for Android* applications were added to the Linux* OS
    package.
- Binary files for Windows Store* applications were added to the
    Windows* OS package.
- Exact exception propagation (exception_ptr) support on Linux OS is
    now turned on by default for GCC 4.4 and higher.
- Stopped implicit use of large memory pages by tbbmalloc (Linux-only).
    Now use of large pages must be explicitly enabled with
    scalable_allocation_mode() function or TBB_MALLOC_USE_HUGE_PAGES
    environment variable.

Community Preview Features:

- Extended class task_arena constructor and method initialize() to
    allow some concurrency to be reserved strictly for application
    threads.
- New methods terminate() and is_active() were added to class
    task_arena.

Bugs fixed:

- Fixed initialization of hashing helper constant in the hash
    containers.
- Fixed possible stalls in concurrent invocations of
    task_arena::execute() when no worker thread is available to make
    progress.
- Fixed incorrect calculation of hardware concurrency in the presence
    of inactive processor groups, particularly on systems running
    Windows* 8 and Windows* Server 2012.

Open-source contributions integrated:

- The fix for the GUI examples on OS X* systems by Raf Schietekat.
- Moved some power-of-2 calculations to functions to improve readability
    by Raf Schietekat.
- C++11/Clang support improvements by arcata.
- ARM* platform isolation layer by Steve Capper, Leif Lindholm, Leo Lara
    (ARM).

------------------------------------------------------------------------
Intel TBB 4.1 Update 2
TBB_INTERFACE_VERSION == 6102

Changes (w.r.t. Intel TBB 4.1 Update 1):

- Objects up to 128 MB are now cached by the tbbmalloc. Previously
    the threshold was 8MB. Objects larger than 128 MB are still
    processed by direct OS calls.
- concurrent_unordered_multiset and concurrent_unordered_multimap
    have been added, based on Microsoft* PPL prototype.
- Ability to value-initialize a tbb::atomic<T> variable on construction
    in C++11, with const expressions properly supported.

Community Preview Features:

- Added a possibility to wait until all worker threads terminate.
    This is necessary before calling fork() from an application.

Bugs fixed:

- Fixed data race in tbbmalloc that might lead to memory leaks
    for large object allocations.
- Fixed task_arena::enqueue() to use task_group_context of target arena.
- Improved implementation of 64 bit atomics on ia32.

------------------------------------------------------------------------
Intel TBB 4.1 Update 1
TBB_INTERFACE_VERSION == 6101

Changes (w.r.t. Intel TBB 4.1):

- concurrent_vector class now supports initialization/assignment
    via C++11 initializer list feature (std::initializer_list<T>)
- Added implementation of the platform isolation layer based on
    Intel compiler atomic built-ins; it is supposed to work on
    any platform supported by compiler version 12.1 and newer.
- Using GetNativeSystemInfo() instead of GetSystemInfo() to support
    more than 32 processors for 32-bit applications under WOW64.
- The following form of parallel_for:
    parallel_for(first, last, [step,] f[, context]) now accepts an
    optional partitioner parameter after the function f.

Backward-incompatible API changes:

- The library no longer injects tuple in to namespace std.
    In previous releases, tuple was injected into namespace std by
    flow_graph.h when std::tuple was not available.  In this release,
    flow_graph.h now uses tbb::flow::tuple.  On platforms where
    std::tuple is available, tbb::flow::tuple is typedef'ed to
    std::tuple.  On all other platforms, tbb::flow::tuple provides
    a subset of the functionality defined by std::tuple. Users of
    flow_graph.h may need to change their uses of std::tuple to
    tbb::flow::tuple to ensure compatibility with non-C++11 compliant
    compilers.

Bugs fixed:

- Fixed local observer to be able to override propagated CPU state and
    to provide correct value of task_arena::current_slot() in callbacks.

------------------------------------------------------------------------
Intel TBB 4.1
TBB_INTERFACE_VERSION == 6100

Changes (w.r.t. Intel TBB 4.0 Update 5):

- _WIN32_WINNT must be set to 0x0501 or greater in order to use TBB
    on Microsoft* Windows*.
- parallel_deterministic_reduce template function is fully supported.
- TBB headers can be used with C++0x/C++11 mode (-std=c++0x) of GCC
    and Intel(R) Compiler.
- C++11 std::make_exception_ptr is used where available, instead of
    std::copy_exception from earlier C++0x implementations.
- Improvements in the TBB allocator to reduce extra memory consumption.
- Partial refactoring of the task scheduler data structures.
- TBB examples allow more flexible specification of the thread number,
    including arithmetic and geometric progression.

Bugs fixed:

- On Linux & OS X*, pre-built TBB binaries do not yet support exact
    exception propagation via C++11 exception_ptr. To prevent run time
    errors, by default TBB headers disable exact exception propagation
    even if the C++ implementation provides exception_ptr.

Community Preview Features:

- Added: class task_arena, for work submission by multiple application
    threads with thread-independent control of concurrency level.
- Added: task_scheduler_observer can be created as local to a master
    thread, to observe threads that work on behalf of that master.
    Local observers may have new on_scheduler_leaving() callback.

------------------------------------------------------------------------
Intel TBB 4.0 Update 5
TBB_INTERFACE_VERSION == 6005

Changes (w.r.t. Intel TBB 4.0 Update 4):

- Parallel pipeline optimization (directly storing small objects in the
    interstage data buffers) limited to trivially-copyable types for
    C++11 and a short list of types for earlier compilers.
- _VARIADIC_MAX switch is honored for TBB tuple implementation
    and flow::graph nodes based on tuple.
- Support of Cocoa framework was added to the GUI examples on OS X*
    systems.

Bugs fixed:

- Fixed a tv_nsec overflow bug in condition_variable::wait_for.
- Fixed execution order of enqueued tasks with different priorities.
- Fixed a bug with task priority changes causing lack of progress
    for fire-and-forget tasks when TBB was initialized to use 1 thread.
- Fixed duplicate symbol problem when linking multiple compilation
    units that include flow_graph.h on VC 10.

------------------------------------------------------------------------
Intel TBB 4.0 Update 4
TBB_INTERFACE_VERSION == 6004

Changes (w.r.t. Intel TBB 4.0 Update 3):

- The TBB memory allocator transparently supports large pages on Linux.
- A new flow_graph example, logic_sim, was added.
- Support for DirectX* 9 was added to GUI examples.

Community Preview Features:

- Added: aggregator, a new concurrency control mechanism.

Bugs fixed:

- The abort operation on concurrent_bounded_queue now leaves the queue
    in a reusable state. If a bad_alloc or bad_last_alloc exception is
    thrown while the queue is recovering from an abort, that exception
    will be reported instead of user_abort on the thread on which it
    occurred, and the queue will not be reusable.
- Steal limiting heuristic fixed to avoid premature stealing disabling
    when large amount of __thread data is allocated on thread stack.
- Fixed a low-probability leak of arenas in the task scheduler.
- In STL-compatible allocator classes, the method construct() was fixed
    to comply with C++11 requirements.
- Fixed a bug that prevented creation of fixed-size memory pools
    smaller than 2M.
- Significantly reduced the amount of warnings from various compilers.

Open-source contributions integrated:

- Multiple improvements by Raf Schietekat.
- Basic support for Clang on OS X* by Blas Rodriguez Somoza.
- Fixes for warnings and corner-case bugs by Blas Rodriguez Somoza
    and Edward Lam.

------------------------------------------------------------------------
Intel TBB 4.0 Update 3
TBB_INTERFACE_VERSION == 6003

Changes (w.r.t. Intel TBB 4.0 Update 2):

- Modifications to the low-level API for memory pools:
    added support for aligned allocations;
    pool policies reworked to allow backward-compatible extensions;
    added a policy to not return memory space till destruction;
    pool_reset() does not return memory space anymore.
- Class tbb::flow::graph_iterator added to iterate over all nodes
    registered with a graph instance.
- multioutput_function_node has been renamed multifunction_node.
    multifunction_node and split_node are now fully-supported features.
- For the tagged join node, the policy for try_put of an item with
    already existing tag has been defined: the item will be rejected.
- Matching the behavior on Windows, on other platforms the optional
    shared libraries (libtbbmalloc, libirml) now are also searched
    only in the directory where libtbb is located.
- The platform isolation layer based on GCC built-ins is extended.

Backward-incompatible API changes:

- a graph reference parameter is now required to be passed to the
    constructors of the following flow graph nodes: overwrite_node,
    write_once_node, broadcast_node, and the CPF or_node.
- the following tbb::flow node methods and typedefs have been renamed:
       Old                             New
    join_node and or_node:
       inputs()                 ->     input_ports()
       input_ports_tuple_type   ->     input_ports_type
    multifunction_node and split_node:
       ports_type               ->     output_ports_type

Bugs fixed:

- Not all logical processors were utilized on systems with more than
    64 cores split by Windows into several processor groups.

------------------------------------------------------------------------
Intel TBB 4.0 Update 2 commercial-aligned release
TBB_INTERFACE_VERSION == 6002

Changes (w.r.t. Intel TBB 4.0 Update 1 commercial-aligned release):

- concurrent_bounded_queue now has an abort() operation that releases
    threads involved in pending push or pop operations. The released
    threads will receive a tbb::user_abort exception.
- Added Community Preview Feature:  concurrent_lru_cache container,
    a concurrent implementation of LRU (least-recently-used) cache.

Bugs fixed:

- fixed a race condition in the TBB scalable allocator.
- concurrent_queue counter wraparound bug was fixed, which occurred when
    the number of push and pop operations exceeded ~>4 billion on IA32.
- fixed races in the TBB scheduler that could put workers asleep too
    early, especially in presence of affinitized tasks.

------------------------------------------------------------------------
Intel TBB 4.0 Update 1 commercial-aligned release
TBB_INTERFACE_VERSION == 6000 (forgotten to increment)

Changes (w.r.t. Intel TBB 4.0 commercial-aligned release):

- Memory leaks fixed in binpack example.
- Improvements and fixes in the TBB allocator.

------------------------------------------------------------------------
Intel TBB 4.0 commercial-aligned release
TBB_INTERFACE_VERSION == 6000

Changes (w.r.t. Intel TBB 3.0 Update 8 commercial-aligned release):

- concurrent_priority_queue is now a fully supported feature.
    Capacity control methods were removed.
- Flow graph is now a fully supported feature.
- A new memory backend has been implemented in the TBB allocator.
    It can reuse freed memory for both small and large objects, and
    returns unused memory blocks to the OS more actively.
- Improved partitioning algorithms for parallel_for and parallel_reduce
    to better handle load imbalance.
- The convex_hull example has been refactored for reproducible
    performance results.
- The major interface version has changed from 5 to 6.
    Deprecated interfaces might be removed in future releases.

Community Preview Features:

- Added: serial subset, i.e. sequential implementations of TBB generic
    algorithms (currently, only provided for parallel_for).
- Preview of new flow graph nodes:
    or_node (accepts multiple inputs, forwards each input separately
      to all successors),
    split_node (accepts tuples, and forwards each element of a tuple
      to a corresponding successor), and
    multioutput_function_node (accepts one input, and passes the input
    and a tuple of output ports to the function body to support outputs
    to multiple successors).
- Added: memory pools for more control on memory source, grouping,
    and collective deallocation.

------------------------------------------------------------------------
Intel TBB 3.0 Update 8 commercial-aligned release
TBB_INTERFACE_VERSION == 5008

Changes (w.r.t. Intel TBB 3.0 Update 7 commercial-aligned release):

- Task priorities become an official feature of TBB,
    not community preview as before.
- Atomics API extended, and implementation refactored.
- Added task::set_parent() method.
- Added concurrent_unordered_set container.

Open-source contributions integrated:

- PowerPC support by Raf Schietekat.
- Fix of potential task pool overrun and other improvements
    in the task scheduler by Raf Schietekat.
- Fix in parallel_for_each to work with std::set in Visual* C++ 2010.

Community Preview Features:

- Graph community preview feature was renamed to flow graph.
    Multiple improvements in the implementation.
    Binpack example was added for the feature.
- A number of improvements to concurrent_priority_queue.
    Shortpath example was added for the feature.
- TBB runtime loaded functionality was added (Windows*-only).
    It allows to specify which versions of TBB should be used,
    as well as to set directories for the library search.
- parallel_deterministic_reduce template function was added.

------------------------------------------------------------------------
Intel TBB 3.0 Update 7 commercial-aligned release
TBB_INTERFACE_VERSION == 5006 (forgotten to increment)

Changes (w.r.t. Intel TBB 3.0 Update 6 commercial-aligned release):

- Added implementation of the platform isolation layer based on
    GCC atomic built-ins; it is supposed to work on any platform
    where GCC has these built-ins.

Community Preview Features:

- Graph's dining_philosophers example added.
- A number of improvements to graph and concurrent_priority_queue.


------------------------------------------------------------------------
Intel TBB 3.0 Update 6 commercial-aligned release
TBB_INTERFACE_VERSION == 5006

Changes (w.r.t. Intel TBB 3.0 Update 5 commercial-aligned release):

- Added Community Preview feature: task and task group priority, and
    Fractal example demonstrating it.
- parallel_pipeline optimized for data items of small and large sizes.
- Graph's join_node is now parametrized with a tuple of up to 10 types.
- Improved performance of concurrent_priority_queue.

Open-source contributions integrated:

- Initial NetBSD support by Aleksej Saushev.

Bugs fixed:

- Failure to enable interoperability with Intel(R) Cilk(tm) Plus runtime
    library, and a crash caused by invoking the interoperability layer
    after one of the libraries was unloaded.
- Data race that could result in concurrent_unordered_map structure
    corruption after call to clear() method.
- Stack corruption caused by PIC version of 64-bit CAS compiled by Intel
    compiler on Linux.
- Inconsistency of exception propagation mode possible when application
    built with Microsoft* Visual Studio* 2008 or earlier uses TBB built
    with Microsoft* Visual Studio* 2010.
- Affinitizing master thread to a subset of available CPUs after TBB
    scheduler was initialized tied all worker threads to the same CPUs.
- Method is_stolen_task() always returned 'false' for affinitized tasks.
- write_once_node and overwrite_node did not immediately send buffered
    items to successors

------------------------------------------------------------------------
Intel TBB 3.0 Update 5 commercial-aligned release
TBB_INTERFACE_VERSION == 5005

Changes (w.r.t. Intel TBB 3.0 Update 4 commercial-aligned release):

- Added Community Preview feature: graph.
- Added automatic propagation of master thread FPU settings to
    TBB worker threads.
- Added a public function to perform a sequentially consistent full
    memory fence: tbb::atomic_fence() in tbb/atomic.h.

Bugs fixed:

- Data race that could result in scheduler data structures corruption
    when using fire-and-forget tasks.
- Potential referencing of destroyed concurrent_hash_map element after
    using erase(accessor&A) method with A acquired as const_accessor.
- Fixed a correctness bug in the convex hull example.

Open-source contributions integrated:

- Patch for calls to internal::atomic_do_once() by Andrey Semashev.

------------------------------------------------------------------------
Intel TBB 3.0 Update 4 commercial-aligned release
TBB_INTERFACE_VERSION == 5004

Changes (w.r.t. Intel TBB 3.0 Update 3 commercial-aligned release):

- Added Community Preview feature: concurrent_priority_queue.
- Fixed library loading to avoid possibility for remote code execution,
    see http://www.microsoft.com/technet/security/advisory/2269637.mspx.
- Added support of more than 64 cores for appropriate Microsoft*
    Windows* versions. For more details, see
    http://msdn.microsoft.com/en-us/library/dd405503.aspx.
- Default number of worker threads is adjusted in accordance with
    process affinity mask.

Bugs fixed:

- Calls of scalable_* functions from inside the allocator library
    caused issues if the functions were overridden by another module.
- A crash occurred if methods run() and wait() were called concurrently
    for an empty tbb::task_group (1736).
- The tachyon example exhibited build problems associated with
    bug 554339 on Microsoft* Visual Studio* 2010. Project files were
    modified as a partial workaround to overcome the problem. See
    http://connect.microsoft.com/VisualStudio/feedback/details/554339.

------------------------------------------------------------------------
Intel TBB 3.0 Update 3 commercial-aligned release
TBB_INTERFACE_VERSION == 5003

Changes (w.r.t. Intel TBB 3.0 Update 2 commercial-aligned release):

- cache_aligned_allocator class reworked to use scalable_aligned_malloc.
- Improved performance of count() and equal_range() methods
    in concurrent_unordered_map.
- Improved implementation of 64-bit atomic loads and stores on 32-bit
    platforms, including compilation with VC 7.1.
- Added implementation of atomic operations on top of OSAtomic API
    provided by OS X*.
- Removed gratuitous try/catch blocks surrounding thread function calls
  in tbb_thread.
- Xcode* projects were added for sudoku and game_of_life examples.
- Xcode* projects were updated to work without TBB framework.

Bugs fixed:

- Fixed a data race in task scheduler destruction that on rare occasion
    could result in memory corruption.
- Fixed idle spinning in thread bound filters in tbb::pipeline (1670).

Open-source contributions integrated:

- MinGW-64 basic support by brsomoza (partially).
- Patch for atomic.h by Andrey Semashev.
- Support for AIX & GCC on PowerPC by Giannis Papadopoulos.
- Various improvements by Raf Schietekat.

------------------------------------------------------------------------
Intel TBB 3.0 Update 2 commercial-aligned release
TBB_INTERFACE_VERSION == 5002

Changes (w.r.t. Intel TBB 3.0 Update 1 commercial-aligned release):

- Destructor of tbb::task_group class throws missing_wait exception
    if there are tasks running when it is invoked.
- Interoperability layer with Intel Cilk Plus runtime library added
    to protect TBB TLS in case of nested usage with Intel Cilk Plus.
- Compilation fix for dependent template names in concurrent_queue.
- Memory allocator code refactored to ease development and maintenance.

Bugs fixed:

- Improved interoperability with other Intel software tools on Linux in
    case of dynamic replacement of memory allocator (1700)
- Fixed install issues that prevented installation on
    Mac OS* X 10.6.4 (1711).

------------------------------------------------------------------------
Intel TBB 3.0 Update 1 commercial-aligned release
TBB_INTERFACE_VERSION == 5000 (forgotten to increment)

Changes (w.r.t. Intel TBB 3.0 commercial-aligned release):

- Decreased memory fragmentation by allocations bigger than 8K.
- Lazily allocate worker threads, to avoid creating unnecessary stacks.

Bugs fixed:

- TBB allocator used much more memory than malloc (1703) - see above.
- Deadlocks happened in some specific initialization scenarios
    of the TBB allocator (1701, 1704).
- Regression in enumerable_thread_specific: excessive requirements
    for object constructors.
- A bug in construction of parallel_pipeline filters when body instance
    was a temporary object.
- Incorrect usage of memory fences on PowerPC and XBOX360 platforms.
- A subtle issue in task group context binding that could result
    in cancellation signal being missed by nested task groups.
- Incorrect construction of concurrent_unordered_map if specified
    number of buckets is not power of two.
- Broken count() and equal_range() of concurrent_unordered_map.
- Return type of postfix form of operator++ for hash map's iterators.

------------------------------------------------------------------------
Intel TBB 3.0 commercial-aligned release
TBB_INTERFACE_VERSION == 5000

Changes (w.r.t. Intel TBB 2.2 Update 3 commercial-aligned release):

- All open-source-release changes down to TBB 2.2 U3 below
    were incorporated into this release.

------------------------------------------------------------------------
20100406 open-source release

Changes (w.r.t. 20100310 open-source release):

- Added support for Microsoft* Visual Studio* 2010, including binaries.
- Added a PDF file with recommended Design Patterns for TBB.
- Added parallel_pipeline function and companion classes and functions
    that provide a strongly typed lambda-friendly pipeline interface.
- Reworked enumerable_thread_specific to use a custom implementation of
    hash map that is more efficient for ETS usage models.
- Added example for class task_group; see examples/task_group/sudoku.
- Removed two examples, as they were long outdated and superceded:
    pipeline/text_filter (use pipeline/square);
    parallel_while/parallel_preorder (use parallel_do/parallel_preorder).
- PDF documentation updated.
- Other fixes and changes in code, tests, and examples.

Bugs fixed:

- Eliminated build errors with MinGW32.
- Fixed post-build step and other issues in VS projects for examples.
- Fixed discrepancy between scalable_realloc and scalable_msize that
    caused crashes with malloc replacement on Windows.

------------------------------------------------------------------------
20100310 open-source release

Changes (w.r.t. Intel TBB 2.2 Update 3 commercial-aligned release):

- Version macros changed in anticipation of a future release.
- Directory structure aligned with Intel(R) C++ Compiler;
    now TBB binaries reside in <arch>/<os_key>/[bin|lib]
    (in TBB 2.x, it was [bin|lib]/<arch>/<os_key>).
- Visual Studio projects changed for examples: instead of separate set
    of files for each VS version, now there is single 'msvs' directory
    that contains workspaces for MS C++ compiler (<example>_cl.sln) and
    Intel C++ compiler (<example>_icl.sln). Works with VS 2005 and above.
- The name versioning scheme for backward compatibility was improved;
    now compatibility-breaking changes are done in a separate namespace.
- Added concurrent_unordered_map implementation based on a prototype
    developed in Microsoft for a future version of PPL.
- Added PPL-compatible writer-preference RW lock (reader_writer_lock).
- Added TBB_IMPLEMENT_CPP0X macro to control injection of C++0x names
    implemented in TBB into namespace std.
- Added almost-C++0x-compatible std::condition_variable, plus a bunch
    of other C++0x classes required by condition_variable.
- With TBB_IMPLEMENT_CPP0X, tbb_thread can be also used as std::thread.
- task.cpp was split into several translation units to structure
    TBB scheduler sources layout. Static data layout and library
    initialization logic were also updated.
- TBB scheduler reworked to prevent master threads from stealing
    work belonging to other masters.
- Class task was extended with enqueue() method, and slightly changed
    semantics of methods spawn() and destroy(). For exact semantics,
    refer to TBB Reference manual.
- task_group_context now allows for destruction by non-owner threads.
- Added TBB_USE_EXCEPTIONS macro to control use of exceptions in TBB
    headers. It turns off (i.e. sets to 0) automatically if specified
    compiler options disable exception handling.
- TBB is enabled to run on top of Microsoft's Concurrency Runtime
    on Windows* 7 (via our worker dispatcher known as RML).
- Removed old unused busy-waiting code in concurrent_queue.
- Described the advanced build & test options in src/index.html.
- Warning level for GCC raised with -Wextra and a few other options.
- Multiple fixes and improvements in code, tests, examples, and docs.

Open-source contributions integrated:

- Xbox support by Roman Lut (Deep Shadows), though further changes are
    required to make it working; e.g. post-2.1 entry points are missing.
- "Eventcount" by Dmitry Vyukov evolved into concurrent_monitor,
    an internal class used in the implementation of concurrent_queue.

------------------------------------------------------------------------
Intel TBB 2.2 Update 3 commercial-aligned release
TBB_INTERFACE_VERSION == 4003

Changes (w.r.t. Intel TBB 2.2 Update 2 commercial-aligned release):

- PDF documentation updated.

Bugs fixed:

- concurrent_hash_map compatibility issue exposed on Linux in case
    two versions of the container were used by different modules.
- enforce 16 byte stack alignment for consistence with GCC; required
    to work correctly with 128-bit variables processed by SSE.
- construct() methods of allocator classes now use global operator new.

------------------------------------------------------------------------
Intel TBB 2.2 Update 2 commercial-aligned release
TBB_INTERFACE_VERSION == 4002

Changes (w.r.t. Intel TBB 2.2 Update 1 commercial-aligned release):

- parallel_invoke and parallel_for_each now take function objects
    by const reference, not by value.
- Building TBB with /MT is supported, to avoid dependency on particular
    versions of Visual C++* runtime DLLs. TBB DLLs built with /MT
    are located in vc_mt directory.
- Class critical_section introduced.
- Improvements in exception support: new exception classes introduced,
    all exceptions are thrown via an out-of-line internal method.
- Improvements and fixes in the TBB allocator and malloc replacement,
    including robust memory identification, and more reliable dynamic
    function substitution on Windows*.
- Method swap() added to class tbb_thread.
- Methods rehash() and bucket_count() added to concurrent_hash_map.
- Added support for Visual Studio* 2010 Beta2. No special binaries
    provided, but CRT-independent DLLs (vc_mt) should work.
- Other fixes and improvements in code, tests, examples, and docs.

Open-source contributions integrated:

- The fix to build 32-bit TBB on Mac OS* X 10.6.
- GCC-based port for SPARC Solaris by Michailo Matijkiw, with use of
    earlier work by Raf Schietekat.

Bugs fixed:

- 159 - TBB build for PowerPC* running Mac OS* X.
- 160 - IBM* Java segfault if used with TBB allocator.
- crash in concurrent_queue<char> (1616).

------------------------------------------------------------------------
Intel TBB 2.2 Update 1 commercial-aligned release
TBB_INTERFACE_VERSION == 4001

Changes (w.r.t. Intel TBB 2.2 commercial-aligned release):

- Incorporates all changes from open-source releases below.
- Documentation was updated.
- TBB scheduler auto-initialization now covers all possible use cases.
- concurrent_queue: made argument types of sizeof used in paddings
  consistent with those actually used.
- Memory allocator was improved: supported corner case of user's malloc
    calling scalable_malloc (non-Windows), corrected processing of
    memory allocation requests during tbb memory allocator startup
    (Linux).
- Windows malloc replacement has got better support for static objects.
- In pipeline setups that do not allow actual parallelism, execution
    by a single thread is guaranteed, idle spinning eliminated, and
    performance improved.
- RML refactoring and clean-up.
- New constructor for concurrent_hash_map allows reserving space for
    a number of items.
- Operator delete() added to the TBB exception classes.
- Lambda support was improved in parallel_reduce.
- gcc 4.3 warnings were fixed for concurrent_queue.
- Fixed possible initialization deadlock in modules using TBB entities
    during construction of global static objects.
- Copy constructor in concurrent_hash_map was fixed.
- Fixed a couple of rare crashes in the scheduler possible before
    in very specific use cases.
- Fixed a rare crash in the TBB allocator running out of memory.
- New tests were implemented, including test_lambda.cpp that checks
    support for lambda expressions.
- A few other small changes in code, tests, and documentation.

------------------------------------------------------------------------
20090809 open-source release

Changes (w.r.t. Intel TBB 2.2 commercial-aligned release):

- Fixed known exception safety issues in concurrent_vector.
- Better concurrency of simultaneous grow requests in concurrent_vector.
- TBB allocator further improves performance of large object allocation.
- Problem with source of text relocations was fixed on Linux
- Fixed bugs related to malloc replacement under Windows
- A few other small changes in code and documentation.

------------------------------------------------------------------------
Intel TBB 2.2 commercial-aligned release
TBB_INTERFACE_VERSION == 4000

Changes (w.r.t. Intel TBB 2.1 U4 commercial-aligned release):

- Incorporates all changes from open-source releases below.
- Architecture folders renamed from em64t to intel64 and from itanium
    to ia64.
- Major Interface version changed from 3 to 4. Deprecated interfaces
    might be removed in future releases.
- Parallel algorithms that use partitioners have switched to use
    the auto_partitioner by default.
- Improved memory allocator performance for allocations bigger than 8K.
- Added new thread-bound filters functionality for pipeline.
- New implementation of concurrent_hash_map that improves performance
    significantly.
- A few other small changes in code and documentation.

------------------------------------------------------------------------
20090511 open-source release

Changes (w.r.t. previous open-source release):

- Basic support for MinGW32 development kit.
- Added tbb::zero_allocator class that initializes memory with zeros.
    It can be used as an adaptor to any STL-compatible allocator class.
- Added tbb::parallel_for_each template function as alias to parallel_do.
- Added more overloads for tbb::parallel_for.
- Added support for exact exception propagation (can only be used with
    compilers that support C++0x std::exception_ptr).
- tbb::atomic template class can be used with enumerations.
- mutex, recursive_mutex, spin_mutex, spin_rw_mutex classes extended
    with explicit lock/unlock methods.
- Fixed size() and grow_to_at_least() methods of tbb::concurrent_vector
    to provide space allocation guarantees. More methods added for
    compatibility with std::vector, including some from C++0x.
- Preview of a lambda-friendly interface for low-level use of tasks.
- scalable_msize function added to the scalable allocator (Windows only).
- Rationalized internal auxiliary functions for spin-waiting and backoff.
- Several tests undergo decent refactoring.

Changes affecting backward compatibility:

- Improvements in concurrent_queue, including limited API changes.
    The previous version is deprecated; its functionality is accessible
    via methods of the new tbb::concurrent_bounded_queue class.
- grow* and push_back methods of concurrent_vector changed to return
    iterators; old semantics is deprecated.

------------------------------------------------------------------------
Intel TBB 2.1 Update 4 commercial-aligned release
TBB_INTERFACE_VERSION == 3016

Changes (w.r.t. Intel TBB 2.1 U3 commercial-aligned release):

- Added tests for aligned memory allocations and malloc replacement.
- Several improvements for better bundling with Intel(R) C++ Compiler.
- A few other small changes in code and documentaion.

Bugs fixed:

- 150 - request to build TBB examples with debug info in release mode.
- backward compatibility issue with concurrent_queue on Windows.
- dependency on VS 2005 SP1 runtime libraries removed.
- compilation of GUI examples under Xcode* 3.1 (1577).
- On Windows, TBB allocator classes can be instantiated with const types
    for compatibility with MS implementation of STL containers (1566).

------------------------------------------------------------------------
20090313 open-source release

Changes (w.r.t. 20081109 open-source release):

- Includes all changes introduced in TBB 2.1 Update 2 & Update 3
    commercial-aligned releases (see below for details).
- Added tbb::parallel_invoke template function. It runs up to 10
    user-defined functions in parallel and waits for them to complete.
- Added a special library providing ability to replace the standard
    memory allocation routines in Microsoft* C/C++ RTL (malloc/free,
    global new/delete, etc.) with the TBB memory allocator.
    Usage details are described in include/tbb/tbbmalloc_proxy.h file.
- Task scheduler switched to use new implementation of its core
    functionality (deque based task pool, new structure of arena slots).
- Preview of Microsoft* Visual Studio* 2005 project files for
    building the library is available in build/vsproject folder.
- Added tests for aligned memory allocations and malloc replacement.
- Added parallel_for/game_of_life.net example (for Windows only)
    showing TBB usage in a .NET application.
- A number of other fixes and improvements to code, tests, makefiles,
    examples and documents.

Bugs fixed:

- The same list as in TBB 2.1 Update 4 right above.

------------------------------------------------------------------------
Intel TBB 2.1 Update 3 commercial-aligned release
TBB_INTERFACE_VERSION == 3015

Changes (w.r.t. Intel TBB 2.1 U2 commercial-aligned release):

- Added support for aligned allocations to the TBB memory allocator.
- Added a special library to use with LD_PRELOAD on Linux* in order to
    replace the standard memory allocation routines in C/C++ with the
    TBB memory allocator.
- Added null_mutex and null_rw_mutex: no-op classes interface-compliant
    to other TBB mutexes.
- Improved performance of parallel_sort, to close most of the serial gap
    with std::sort, and beat it on 2 and more cores.
- A few other small changes.

Bugs fixed:

- the problem where parallel_for hanged after exception throw
    if affinity_partitioner was used (1556).
- get rid of VS warnings about mbstowcs deprecation (1560),
    as well as some other warnings.
- operator== for concurrent_vector::iterator fixed to work correctly
    with different vector instances.

------------------------------------------------------------------------
Intel TBB 2.1 Update 2 commercial-aligned release
TBB_INTERFACE_VERSION == 3014

Changes (w.r.t. Intel TBB 2.1 U1 commercial-aligned release):

- Incorporates all open-source-release changes down to TBB 2.1 U1,
    except for:
    - 20081019 addition of enumerable_thread_specific;
- Warning level for Microsoft* Visual C++* compiler raised to /W4 /Wp64;
    warnings found on this level were cleaned or suppressed.
- Added TBB_runtime_interface_version API function.
- Added new example: pipeline/square.
- Added exception handling and cancellation support
    for parallel_do and pipeline.
- Added copy constructor and [begin,end) constructor to concurrent_queue.
- Added some support for beta version of Intel(R) Parallel Amplifier.
- Added scripts to set environment for cross-compilation of 32-bit
    applications on 64-bit Linux with Intel(R) C++ Compiler.
- Fixed semantics of concurrent_vector::clear() to not deallocate
    internal arrays. Fixed compact() to perform such deallocation later.
- Fixed the issue with atomic<T*> when T is incomplete type.
- Improved support for PowerPC* Macintosh*, including the fix
    for a bug in masked compare-and-swap reported by a customer.
- As usual, a number of other improvements everywhere.

------------------------------------------------------------------------
20081109 open-source release

Changes (w.r.t. previous open-source release):

- Added new serial out of order filter for tbb::pipeline.
- Fixed the issue with atomic<T*>::operator= reported at the forum.
- Fixed the issue with using tbb::task::self() in task destructor
    reported at the forum.
- A number of other improvements to code, tests, makefiles, examples
    and documents.

Open-source contributions integrated:
- Changes in the memory allocator were partially integrated.

------------------------------------------------------------------------
20081019 open-source release

Changes (w.r.t. previous open-source release):

- Introduced enumerable_thread_specific<T>.  This new class provides a
    wrapper around native thread local storage as well as iterators and
    ranges for accessing the thread local copies (1533).
- Improved support for Intel(R) Threading Analysis Tools
    on Intel(R) 64 architecture.
- Dependency from Microsoft* CRT was integrated to the libraries using
    manifests, to avoid issues if called from code that uses different
    version of Visual C++* runtime than the library.
- Introduced new defines TBB_USE_ASSERT, TBB_USE_DEBUG,
    TBB_USE_PERFORMANCE_WARNINGS, TBB_USE_THREADING_TOOLS.
- A number of other improvements to code, tests, makefiles, examples
    and documents.

Open-source contributions integrated:

- linker optimization: /incremental:no .

------------------------------------------------------------------------
20080925 open-source release

Changes (w.r.t. previous open-source release):

- Same fix for a memory leak in the memory allocator as in TBB 2.1 U1.
- Improved support for lambda functions.
- Fixed more concurrent_queue issues reported at the forum.
- A number of other improvements to code, tests, makefiles, examples
    and documents.

------------------------------------------------------------------------
Intel TBB 2.1 Update 1 commercial-aligned release
TBB_INTERFACE_VERSION == 3013

Changes (w.r.t. Intel TBB 2.1 commercial-aligned release):

- Fixed small memory leak in the memory allocator.
- Incorporates all open-source-release changes since TBB 2.1,
    except for:
    - 20080825 changes for parallel_do;

------------------------------------------------------------------------
20080825 open-source release

Changes (w.r.t. previous open-source release):

- Added exception handling and cancellation support for parallel_do.
- Added default HashCompare template argument for concurrent_hash_map.
- Fixed concurrent_queue.clear() issues due to incorrect assumption
    about clear() being private method.
- Added the possibility to use TBB in applications that change
    default calling conventions (Windows* only).
- Many improvements to code, tests, examples, makefiles and documents.

Bugs fixed:

- 120, 130 - memset declaration missed in concurrent_hash_map.h

------------------------------------------------------------------------
20080724 open-source release

Changes (w.r.t. previous open-source release):

- Inline assembly for atomic operations improved for gcc 4.3
- A few more improvements to the code.

------------------------------------------------------------------------
20080709 open-source release

Changes (w.r.t. previous open-source release):

- operator=() was added to the tbb_thread class according to
    the current working draft for std::thread.
- Recognizing SPARC* in makefiles for Linux* and Sun Solaris*.

Bugs fixed:

- 127 - concurrent_hash_map::range fixed to split correctly.

Open-source contributions integrated:

- fix_set_midpoint.diff by jyasskin
- SPARC* support in makefiles by Raf Schietekat

------------------------------------------------------------------------
20080622 open-source release

Changes (w.r.t. previous open-source release):

- Fixed a hang that rarely happened on Linux
    during deinitialization of the TBB scheduler.
- Improved support for Intel(R) Thread Checker.
- A few more improvements to the code.

------------------------------------------------------------------------
Intel TBB 2.1 commercial-aligned release
TBB_INTERFACE_VERSION == 3011

Changes (w.r.t. Intel TBB 2.0 U3 commercial-aligned release):

- All open-source-release changes down to, and including, TBB 2.0 below,
    were incorporated into this release.

------------------------------------------------------------------------
20080605 open-source release

Changes (w.r.t. previous open-source release):

- Explicit control of exported symbols by version scripts added on Linux.
- Interfaces polished for exception handling & algorithm cancellation.
- Cache behavior improvements in the scalable allocator.
- Improvements in text_filter, polygon_overlay, and other examples.
- A lot of other stability improvements in code, tests, and makefiles.
- First release where binary packages include headers/docs/examples, so
    binary packages are now self-sufficient for using TBB.

Open-source contributions integrated:

- atomics patch (partially).
- tick_count warning patch.

Bugs fixed:

- 118 - fix for boost compatibility.
- 123 - fix for tbb_machine.h.

------------------------------------------------------------------------
20080512 open-source release

Changes (w.r.t. previous open-source release):

- Fixed a problem with backward binary compatibility
    of debug Linux builds.
- Sun* Studio* support added.
- soname support added on Linux via linker script. To restore backward
    binary compatibility, *.so -> *.so.2 softlinks should be created.
- concurrent_hash_map improvements - added few new forms of insert()
    method and fixed precondition and guarantees of erase() methods.
    Added runtime warning reporting about bad hash function used for
    the container. Various improvements for performance and concurrency.
- Cancellation mechanism reworked so that it does not hurt scalability.
- Algorithm parallel_do reworked. Requirement for Body::argument_type
    definition removed, and work item argument type can be arbitrarily
    cv-qualified.
- polygon_overlay example added.
- A few more improvements to code, tests, examples and Makefiles.

Open-source contributions integrated:

- Soname support patch for Bugzilla #112.

Bugs fixed:

- 112 - fix for soname support.

------------------------------------------------------------------------
Intel TBB 2.0 U3 commercial-aligned release (package 017, April 20, 2008)

Corresponds to commercial 019 (for Linux*, 020; for Mac OS* X, 018)
packages.

Changes (w.r.t. Intel TBB 2.0 U2 commercial-aligned release):

- Does not contain open-source-release changes below; this release is
    only a minor update of TBB 2.0 U2.
- Removed spin-waiting in pipeline and concurrent_queue.
- A few more small bug fixes from open-source releases below.

------------------------------------------------------------------------
20080408 open-source release

Changes (w.r.t. previous open-source release):

- count_strings example reworked: new word generator implemented, hash
    function replaced, and tbb_allocator is used with std::string class.
- Static methods of spin_rw_mutex were replaced by normal member
    functions, and the class name was versioned.
- tacheon example was renamed to tachyon.
- Improved support for Intel(R) Thread Checker.
- A few more minor improvements.

Open-source contributions integrated:

- Two sets of Sun patches for IA Solaris support.

------------------------------------------------------------------------
20080402 open-source release

Changes (w.r.t. previous open-source release):

- Exception handling and cancellation support for tasks and algorithms
    fully enabled.
- Exception safety guaranties defined and fixed for all concurrent
    containers.
- User-defined memory allocator support added to all concurrent
    containers.
- Performance improvement of concurrent_hash_map, spin_rw_mutex.
- Critical fix for a rare race condition during scheduler
    initialization/de-initialization.
- New methods added for concurrent containers to be closer to STL,
    as well as automatic filters removal from pipeline
    and __TBB_AtomicAND function.
- The volatile keyword dropped from where it is not really needed.
- A few more minor improvements.

------------------------------------------------------------------------
20080319 open-source release

Changes (w.r.t. previous open-source release):

- Support for gcc version 4.3 was added.
- tbb_thread class, near compatible with std::thread expected in C++0x,
    was added.

Bugs fixed:

- 116 - fix for compilation issues with gcc version 4.2.1.
- 120 - fix for compilation issues with gcc version 4.3.

------------------------------------------------------------------------
20080311 open-source release

Changes (w.r.t. previous open-source release):

- An enumerator added for pipeline filter types (serial vs. parallel).
- New task_scheduler_observer class introduced, to observe when
    threads start and finish interacting with the TBB task scheduler.
- task_scheduler_init reverted to not use internal versioned class;
    binary compatibility guaranteed with stable releases only.
- Various improvements to code, tests, examples and Makefiles.

------------------------------------------------------------------------
20080304 open-source release

Changes (w.r.t. previous open-source release):

- Task-to-thread affinity support, previously kept under a macro,
    now fully legalized.
- Work-in-progress on cache_aligned_allocator improvements.
- Pipeline really supports parallel input stage; it's no more serialized.
- Various improvements to code, tests, examples and Makefiles.

Bugs fixed:

- 119 - fix for scalable_malloc sometimes failing to return a big block.
- TR575 - fixed a deadlock occurring on Windows in startup/shutdown
    under some conditions.

------------------------------------------------------------------------
20080226 open-source release

Changes (w.r.t. previous open-source release):

- Introduced tbb_allocator to select between standard allocator and
    tbb::scalable_allocator when available.
- Removed spin-waiting in pipeline and concurrent_queue.
- Improved performance of concurrent_hash_map by using tbb_allocator.
- Improved support for Intel(R) Thread Checker.
- Various improvements to code, tests, examples and Makefiles.

------------------------------------------------------------------------
Intel TBB 2.0 U2 commercial-aligned release (package 017, February 14, 2008)

Corresponds to commercial 017 (for Linux*, 018; for Mac OS* X, 016)
packages.

Changes (w.r.t. Intel TBB 2.0 U1 commercial-aligned release):

- Does not contain open-source-release changes below; this release is
    only a minor update of TBB 2.0 U1.
- Add support for Microsoft* Visual Studio* 2008, including binary
    libraries and VS2008 projects for examples.
- Use SwitchToThread() not Sleep() to yield threads on Windows*.
- Enhancements to Doxygen-readable comments in source code.
- A few more small bug fixes from open-source releases below.

Bugs fixed:

- TR569 - Memory leak in concurrent_queue.

------------------------------------------------------------------------
20080207 open-source release

Changes (w.r.t. previous open-source release):

- Improvements and minor fixes in VS2008 projects for examples.
- Improvements in code for gating worker threads that wait for work,
  previously consolidated under #if IMPROVED_GATING, now legalized.
- Cosmetic changes in code, examples, tests.

Bugs fixed:

- 113 - Iterators and ranges should be convertible to their const
    counterparts.
- TR569 - Memory leak in concurrent_queue.

------------------------------------------------------------------------
20080122 open-source release

Changes (w.r.t. previous open-source release):

- Updated examples/parallel_for/seismic to improve the visuals and to
    use the affinity_partitioner (20071127 and forward) for better
    performance.
- Minor improvements to unittests and performance tests.

------------------------------------------------------------------------
20080115 open-source release

Changes (w.r.t. previous open-source release):

- Cleanup, simplifications and enhancements to the Makefiles for
    building the libraries (see build/index.html for high-level
    changes) and the examples.
- Use SwitchToThread() not Sleep() to yield threads on Windows*.
- Engineering work-in-progress on exception safety/support.
- Engineering work-in-progress on affinity_partitioner for
    parallel_reduce.
- Engineering work-in-progress on improved gating for worker threads
    (idle workers now block in the OS instead of spinning).
- Enhancements to Doxygen-readable comments in source code.

Bugs fixed:

- 102 - Support for parallel build with gmake -j
- 114 - /Wp64 build warning on Windows*.

------------------------------------------------------------------------
20071218 open-source release

Changes (w.r.t. previous open-source release):

- Full support for Microsoft* Visual Studio* 2008 in open-source.
    Binaries for vc9/ will be available in future stable releases.
- New recursive_mutex class.
- Full support for 32-bit PowerMac including export files for builds.
- Improvements to parallel_do.

------------------------------------------------------------------------
20071206 open-source release

Changes (w.r.t. previous open-source release):

- Support for Microsoft* Visual Studio* 2008 in building libraries
    from source as well as in vc9/ projects for examples.
- Small fixes to the affinity_partitioner first introduced in 20071127.
- Small fixes to the thread-stack size hook first introduced in 20071127.
- Engineering work in progress on concurrent_vector.
- Engineering work in progress on exception behavior.
- Unittest improvements.

------------------------------------------------------------------------
20071127 open-source release

Changes (w.r.t. previous open-source release):

- Task-to-thread affinity support (affinity partitioner) first appears.
- More work on concurrent_vector.
- New parallel_do algorithm (function-style version of parallel while)
    and parallel_do/parallel_preorder example.
- New task_scheduler_init() hooks for getting default_num_threads() and
    for setting thread stack size.
- Support for weak memory consistency models in the code base.
- Futex usage in the task scheduler (Linux).
- Started adding 32-bit PowerMac support.
- Intel(R) 9.1 compilers are now the base supported Intel(R) compiler
    version.
- TBB libraries added to link line automatically on Microsoft Windows*
    systems via #pragma comment linker directives.

Open-source contributions integrated:

- FreeBSD platform support patches.
- AIX weak memory model patch.

Bugs fixed:

- 108 - Removed broken affinity.h reference.
- 101 - Does not build on Debian Lenny (replaced arch with uname -m).

------------------------------------------------------------------------
20071030 open-source release

Changes (w.r.t. previous open-source release):

- More work on concurrent_vector.
- Better support for building with -Wall -Werror (or not) as desired.
- A few fixes to eliminate extraneous warnings.
- Begin introduction of versioning hooks so that the internal/API
    version is tracked via TBB_INTERFACE_VERSION.  The newest binary
    libraries should always work with previously-compiled code when-
    ever possible.
- Engineering work in progress on using futex inside the mutexes (Linux).
- Engineering work in progress on exception behavior.
- Engineering work in progress on a new parallel_do algorithm.
- Unittest improvements.

------------------------------------------------------------------------
20070927 open-source release

Changes (w.r.t. Intel TBB 2.0 U1 commercial-aligned release):

- Minor update to TBB 2.0 U1 below.
- Begin introduction of new concurrent_vector interfaces not released
    with TBB 2.0 U1.

------------------------------------------------------------------------
Intel TBB 2.0 U1 commercial-aligned release (package 014, October 1, 2007)

Corresponds to commercial 014 (for Linux*, 016) packages.

Changes (w.r.t. Intel TBB 2.0 commercial-aligned release):

- All open-source-release changes down to, and including, TBB 2.0
    below, were incorporated into this release.
- Made a number of changes to the officially supported OS list:
    Added Linux* OSs:
	Asianux* 3, Debian* 4.0, Fedora Core* 6, Fedora* 7,
	Turbo Linux* 11, Ubuntu* 7.04;
    Dropped Linux* OSs:
	Asianux* 2, Fedora Core* 4, Haansoft* Linux 2006 Server,
	Mandriva/Mandrake* 10.1, Miracle Linux* 4.0,
	Red Flag* DC Server 5.0;
    Only Mac OS* X 10.4.9 (and forward) and Xcode* tool suite 2.4.1 (and
	forward) are now supported.
- Commercial installers on Linux* fixed to recommend the correct
    binaries to use in more cases, with less unnecessary warnings.
- Changes to eliminate spurious build warnings.

Open-source contributions integrated:

- Two small header guard macro patches; it also fixed bug #94.
- New blocked_range3d class.

Bugs fixed:

- 93 - Removed misleading comments in task.h.
- 94 - See above.

------------------------------------------------------------------------
20070815 open-source release

Changes:

- Changes to eliminate spurious build warnings.
- Engineering work in progress on concurrent_vector allocator behavior.
- Added hooks to use the Intel(R) compiler code coverage tools.

Open-source contributions integrated:

- Mac OS* X build warning patch.

Bugs fixed:

- 88 - Fixed TBB compilation errors if both VS2005 and Windows SDK are
    installed.

------------------------------------------------------------------------
20070719 open-source release

Changes:

- Minor update to TBB 2.0 commercial-aligned release below.
- Changes to eliminate spurious build warnings.

------------------------------------------------------------------------
Intel TBB 2.0 commercial-aligned release (package 010, July 19, 2007)

Corresponds to commercial 010 (for Linux*, 012) packages.

- TBB open-source debut release.

------------------------------------------------------------------------
Intel TBB 1.1 commercial release (April 10, 2007)

Changes (w.r.t. Intel TBB 1.0 commercial release):

- auto_partitioner which offered an automatic alternative to specifying
    a grain size parameter to estimate the best granularity for tasks.
- The release was added to the Intel(R) C++ Compiler 10.0 Pro.

------------------------------------------------------------------------
Intel TBB 1.0 Update 2 commercial release

Changes (w.r.t. Intel TBB 1.0 Update 1 commercial release):

- Mac OS* X 64-bit support added.
- Source packages for commercial releases introduced.

------------------------------------------------------------------------
Intel TBB 1.0 Update 1 commercial-aligned release

Changes (w.r.t. Intel TBB 1.0 commercial release):

- Fix for critical package issue on Mac OS* X.

------------------------------------------------------------------------
Intel TBB 1.0 commercial release (August 29, 2006)

Changes (w.r.t. Intel TBB 1.0 beta commercial release):

- New namespace (and compatibility headers for old namespace).
    Namespaces are tbb and tbb::internal and all classes are in the
    underscore_style not the WindowsStyle.
- New class: scalable_allocator (and cache_aligned_allocator using that
    if it exists).
- Added parallel_for/tacheon example.
- Removed C-style casts from headers for better C++ compliance.
- Bug fixes.
- Documentation improvements.
- Improved performance of the concurrent_hash_map class.
- Upgraded parallel_sort() to support STL-style random-access iterators
    instead of just pointers.
- The Windows vs7_1 directories renamed to vs7.1 in examples.
- New class: spin version of reader-writer lock.
- Added push_back() interface to concurrent_vector().

------------------------------------------------------------------------
Intel TBB 1.0 beta commercial release

Initial release.

Features / APIs:

- Concurrent containers: ConcurrentHashTable, ConcurrentVector,
    ConcurrentQueue.
- Parallel algorithms: ParallelFor, ParallelReduce, ParallelScan,
    ParallelWhile, Pipeline, ParallelSort.
- Support: AlignedSpace, BlockedRange (i.e., 1D), BlockedRange2D
- Task scheduler with multi-master support.
- Atomics: read, write, fetch-and-store, fetch-and-add, compare-and-swap.
- Locks: spin, reader-writer, queuing, OS-wrapper.
- Memory allocation: STL-style memory allocator that avoids false
    sharing.
- Timers.

Tools Support:
- Intel(R) Thread Checker 3.0.
- Intel(R) Thread Profiler 3.0.

Documentation:
- First Use Documents: README.txt, INSTALL.txt, Release_Notes.txt,
    Doc_Index.html, Getting_Started.pdf, Tutorial.pdf, Reference.pdf.
- Class hierarchy HTML pages (Doxygen).
- Tree of index.html pages for navigating the installed package, esp.
    for the examples.

Examples:
- One for each of these TBB features: ConcurrentHashTable, ParallelFor,
    ParallelReduce, ParallelWhile, Pipeline, Task.
- Live copies of examples from Getting_Started.pdf.
- TestAll example that exercises every class and header in the package
    (i.e., a "liveness test").
- Compilers: see Release_Notes.txt.
- APIs: OpenMP, WinThreads, Pthreads.

Packaging:
- Package for Windows installs IA-32 and EM64T bits.
- Package for Linux installs IA-32, EM64T and IPF bits.
- Package for Mac OS* X installs IA-32 bits.
- All packages support Intel(R) software setup assistant (ISSA) and
    install-time FLEXlm license checking.
- ISSA support allows license file to be specified directly in case of
    no Internet connection or problems with IRC or serial #s.
- Linux installer allows root or non-root, RPM or non-RPM installs.
- FLEXlm license servers (for those who need floating/counted licenses)
    are provided separately on Intel(R) Premier.

------------------------------------------------------------------------
Intel and Cilk are registered trademarks or trademarks of Intel Corporation or its
subsidiaries in the United States and other countries.

* Other names and brands may be claimed as the property of others.