Pavel Kumbrasev
Pavel Kumbrasev
### Description PoC with initial separation arena from market. Fixes # - _issue number(s) if exists_ - [x] - git commit message contains an appropriate signed-off-by string _(see [CONTRIBUTING.md](https://github.com/oneapi-src/oneTBB/blob/master/CONTRIBUTING.md#pull-requests) for...
When threads execute `get_task` or `steal_task` they can omit unsuitable tasks (different isolation or affinity). The omitted tasks are virtually extracted from task pool, that makes them temporary invisible for...
Currently market notifies RML about surplus threads at the time of without of work. However, some workers can not leave arena immediately, but counted as workers to join other arenas....
### Description This patch correct the behavior of static_partitioner when it used with custom range that does not support proportional splitting. For example such combination (static_partitioner + custom range) on...
Signed-off-by: pavelkumbrasev ### Description To enable ThreadSanitizer on __APPLE__ (basically on macOS) replaced semaphore_t on dispatch_semaphore_t which supports via ThreadSanitizer. In accordance to scenarios where and how we use semaphore...
### Description task_group has a flat reference counting scheme i.e., there is a central reference counter where all the created tasks should increase/decrease reference during execution. This approach works fine...
### Description Currently constrained arena might be initialized (first touch to fields) on inappropriate constraints because initializing thread doesn't apply constraints during initialization. This PR reuses `binding_observer` to apply constraints...
The lifetime of [theMarket mutex](https://github.com/oneapi-src/oneTBB/blob/c7865c1be098ba2c2a033f67ff74ac0ef4b16374/src/tbb/market.h#L78) is controlled as a lifetime of static variable. However, when main thread call `_exit`, destructor for this mutex might be called while worker threads still...
### Description 1) Introduce per thread `reference_vertex` that should help with scalability problem that single wait_context. 2) `task_group` has a flat reference counting scheme i.e., there is a central reference...
### Description `global_control` we limit concurrency that all the internal arenas can share between each other. Therefore, each particular arena doesn't not know its actual concurrency limit but only one...