Skip to content
XFELPP

MPIThreadedDistribution::IterationState

IterationState

Section titled “IterationState”
#include <sbio/execution/mpi_threaded.hh>
class IterationState

Defined in src/lib/sbio/execution/mpi_threaded.hh:318

Friends

Section titled “Friends”
Name Description
MPIThreadedDistribution

MPIThreadedDistribution

Section titled “MPIThreadedDistribution”
friend class MPIThreadedDistribution

Defined in src/lib/sbio/execution/mpi_threaded.hh:364

Public Methods

Section titled “Public Methods”
Return Name Description
IterationState inline
IterationState inline noexcept
IterationState & operator= inline noexcept

IterationState

Section titled “IterationState”

inline

inline IterationState()

Defined in src/lib/sbio/execution/mpi_threaded.hh:320

IterationState

Section titled “IterationState”

inline noexcept

inline IterationState(IterationState && other) noexcept

Defined in src/lib/sbio/execution/mpi_threaded.hh:325

operator=

Section titled “operator=”

inline noexcept

inline IterationState & operator=(IterationState && other) noexcept

Defined in src/lib/sbio/execution/mpi_threaded.hh:342

Private Attributes

Section titled “Private Attributes”
Return Name Description
std::atomic< std::size_t > m_shared_capacity Intra-rank capacity store for inter-thread synchronization of indices.
std::atomic< bool > m_exhausted Latch for if the rank has exhausted all indices.
std::atomic< std::size_t > m_in_flight The number of steps handed out by next() and not yet finished processing.
std::mutex m_trigger_mutex Mutex for thread synchronization on reindexing
SegmentState< bool > m_holding_step Whether this thread is holding a step.
std::atomic< std::size_t > m_batch_round The round counter for the rank.
std::atomic< std::size_t > m_batch_round_base The round when the capacity window started (was refilled).
std::atomic< std::size_t > m_batch_window_start The first step of the window.
std::atomic< std::size_t > m_batch_blocks_before The number dealt in the window before the current one.
std::atomic< std::size_t > m_window_version Odd while a thread changes the window above (with m_shared_capacity).
std::atomic< std::size_t > m_active_units The number of this rank’s threads currently iterating with this state.
SegmentState< bool > m_participating Whether this thread is iterating.
std::atomic< bool > m_departed Latch for if this rank left the iteration early (all its threads left).
std::vector< char > m_node_departed Which node ranks left the iteration early.
int m_tag Message tag of this iteration (in the order states are created on each rank).

m_shared_capacity

Section titled “m_shared_capacity”
std::atomic< std::size_t > m_shared_capacity { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:369

Intra-rank capacity store for inter-thread synchronization of indices.

m_exhausted

Section titled “m_exhausted”
std::atomic< bool > m_exhausted { false }

Defined in src/lib/sbio/execution/mpi_threaded.hh:373

Latch for if the rank has exhausted all indices.

m_in_flight

Section titled “m_in_flight”
std::atomic< std::size_t > m_in_flight { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:380

The number of steps handed out by next() and not yet finished processing.

A thread returns a step at the start of each next() call.

m_trigger_mutex

Section titled “m_trigger_mutex”
std::mutex m_trigger_mutex

Defined in src/lib/sbio/execution/mpi_threaded.hh:385

Mutex for thread synchronization on reindexing

m_holding_step

Section titled “m_holding_step”
SegmentState< bool > m_holding_step

Type: SegmentState< bool >

Defined in src/lib/sbio/execution/mpi_threaded.hh:387

Whether this thread is holding a step.

m_batch_round

Section titled “m_batch_round”
std::atomic< std::size_t > m_batch_round { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:392

The round counter for the rank.

m_batch_round_base

Section titled “m_batch_round_base”
std::atomic< std::size_t > m_batch_round_base { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:396

The round when the capacity window started (was refilled).

m_batch_window_start

Section titled “m_batch_window_start”
std::atomic< std::size_t > m_batch_window_start { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:400

The first step of the window.

m_batch_blocks_before

Section titled “m_batch_blocks_before”
std::atomic< std::size_t > m_batch_blocks_before { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:404

The number dealt in the window before the current one.

m_window_version

Section titled “m_window_version”
std::atomic< std::size_t > m_window_version { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:408

Odd while a thread changes the window above (with m_shared_capacity).

m_active_units

Section titled “m_active_units”
std::atomic< std::size_t > m_active_units { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:413

The number of this rank’s threads currently iterating with this state.

m_participating

Section titled “m_participating”
SegmentState< bool > m_participating

Type: SegmentState< bool >

Defined in src/lib/sbio/execution/mpi_threaded.hh:414

Whether this thread is iterating.

m_departed

Section titled “m_departed”
std::atomic< bool > m_departed { false }

Defined in src/lib/sbio/execution/mpi_threaded.hh:419

Latch for if this rank left the iteration early (all its threads left).

m_node_departed

Section titled “m_node_departed”
std::vector< char > m_node_departed {}

Defined in src/lib/sbio/execution/mpi_threaded.hh:426

Which node ranks left the iteration early.

m_tag

Section titled “m_tag”
int m_tag { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:431

Message tag of this iteration (in the order states are created on each rank).

Private Static Attributes

Section titled “Private Static Attributes”
Return Name Description
std::atomic< int > next_tag static

next_tag

Section titled “next_tag”

static

std::atomic< int > next_tag { 0 }

Defined in src/lib/sbio/execution/mpi_threaded.hh:432