|
Aleph-w 3.0
A C++ Library for Data Structures and Algorithms
|
Configuration for Parallel_Synchronous_Engine.
More...
#include <tpl_ca_parallel_engine.H>
Public Attributes | |
| ThreadPool * | pool = nullptr |
Thread pool to schedule on. nullptr means Aleph::default_pool(). | |
| std::size_t | num_partitions = 0 |
| Number of partitions per step. | |
| std::size_t | min_parallel_cells = 4096 |
| Workloads with fewer cells than this are executed sequentially without touching the pool. | |
Configuration for Parallel_Synchronous_Engine.
Every field has a sensible default, so the structure is mostly useful for tuning. Pass it to the engine constructor or update the configuration in place via set_config().
Definition at line 144 of file tpl_ca_parallel_engine.H.
| std::size_t Aleph::CA::Parallel_Engine_Config::min_parallel_cells = 4096 |
Workloads with fewer cells than this are executed sequentially without touching the pool.
The default is sized so that the per-step task overhead never dominates the cell update cost.
Definition at line 158 of file tpl_ca_parallel_engine.H.
Referenced by Aleph::CA::Parallel_Synchronous_Engine< Lattice, Rule, Neighborhood, Order >::effective_partitions().
| std::size_t Aleph::CA::Parallel_Engine_Config::num_partitions = 0 |
Number of partitions per step.
0 means "use `pool->num_threads()`". Setting this larger than the pool size oversubscribes the pool; setting it smaller reduces task overhead at the cost of using fewer cores.
Definition at line 153 of file tpl_ca_parallel_engine.H.
Referenced by Aleph::CA::Parallel_Synchronous_Engine< Lattice, Rule, Neighborhood, Order >::effective_partitions(), expect_engine_equivalence(), main(), TEST(), TEST(), TEST(), TEST(), TEST(), TEST(), and TEST().
| ThreadPool* Aleph::CA::Parallel_Engine_Config::pool = nullptr |
Thread pool to schedule on. nullptr means Aleph::default_pool().
Definition at line 147 of file tpl_ca_parallel_engine.H.
Referenced by Aleph::CA::Parallel_Synchronous_Engine< Lattice, Rule, Neighborhood, Order >::step(), and TEST().