Struct pargraph::executors::multi_thread::MultiThreadExecutor

pub struct MultiThreadExecutor { /* private fields */ }

Execute graph operators on multiple threads.

This executor will create clones of the operator for each worker thread. Therefore, the operators should usually be cheap to clone. For example they can include large shared data as a reference. Cloning the operators also allows them to keep thread-local data, if necessary.

Implementations

impl MultiThreadExecutor

pub fn new() -> Self

Create a new multi-threaded executor. Automatically determine the number of threads.

pub fn with_num_threads(self, num_threads: NonZeroUsize) -> Self

Set the number of threads to be used.

pub fn run_node_labelling<Wl, Op, G>( &self, initial_worklist: Wl, operator: Op, graph: G, )

where Wl: Worklist<Op::WorkItem>, Wl::Channel: Send, Op: LabellingOperator<G> + Clone + Send, G: GraphBase + DataMap + Sync, G::NodeWeight: BorrowDataCell<UserData = Op::NodeWeight>, Op::WorkItem: Copy,

Run an algorithm which modifies node labels but keeps the topology unchanged.

pub fn run_readonly<Wl, Op, G>( &self, initial_worklist: Wl, operator: Op, graph: G, )

where Wl: Worklist<Op::WorkItem>, Wl::Channel: Send, Op: ReadonlyOperator<G> + Clone + Send, G: GraphBase + Sync,

Run an algorithm which does not need explicit mutable access to the graph elements. It can, of course, modify node and edge weights by using interior mutability together with an appropriate locking mechansim or atomic operations.

Trait Implementations

impl Default for MultiThreadExecutor

fn default() -> Self

Returns the “default value” for a type.