EntropyEngine::Core::Concurrency::AdaptiveRankingScheduler::ThreadState

Per-thread state for adaptive scheduling. More…

Public Functions

	Name
void	reset()

Public Attributes

	Name
size_t	rankingUpdateCounter Counts work done since last ranking update.
std::vector< WorkContractGroup * >	rankedGroups Thread-local group priority ordering.
uint64_t	lastSeenGeneration Generation counter for detecting group list changes.
size_t	currentGroupIndex Current position in rankedGroups (thread affinity position).
size_t	consecutiveExecutionCount Number of consecutive executions on current group.

Detailed Description

struct EntropyEngine::Core::Concurrency::AdaptiveRankingScheduler::ThreadState;

Per-thread state for adaptive scheduling.

This structure enables scheduling by maintaining thread-local copies of rankings and affinity state. The design eliminates locks, atomics, and contention between threads. Synchronization occurs only when the group list changes or during periodic rebalancing operations.

The rankedGroups vector represents each thread’s independent priority ordering. Rankings update based on thread-local observations, which may differ slightly between threads. This variance helps prevent thundering herd effects during work distribution.

Public Functions Documentation

function reset

inline void reset()

Public Attributes Documentation

variable rankingUpdateCounter

size_t rankingUpdateCounter = 0;

Counts work done since last ranking update.

variable rankedGroups

std::vector< WorkContractGroup * > rankedGroups;

Thread-local group priority ordering.

variable lastSeenGeneration

uint64_t lastSeenGeneration = 0;

Generation counter for detecting group list changes.

variable currentGroupIndex

size_t currentGroupIndex = 0;

Current position in rankedGroups (thread affinity position).

variable consecutiveExecutionCount

size_t consecutiveExecutionCount = 0;

Number of consecutive executions on current group.

Updated on 2026-01-26 at 17:14:35 -0500