Enum ExecutionStrategyType

Defines the type of execution strategy for kernel and pipeline execution.

public enum ExecutionStrategyType

Fields

Adaptive = 6: Adaptive execution - strategy chosen based on workload characteristics.
Batch = 8: Batch execution - processes data in batches for efficiency.
DataParallel = 2: Data parallel execution - data is split across multiple execution units.
Heterogeneous = 11: Heterogeneous CPU+GPU execution. Coordinates execution across different types of computing devices (CPUs and GPUs). Maximizes utilization of all available computing resources.
Hybrid = 5: Hybrid execution - combines multiple strategies adaptively.
ModelParallel = 3: Model parallel execution - model is split across multiple devices.
Parallel = 1: Parallel execution - tasks run concurrently where possible.
PipelineParallel = 4: Pipeline parallel execution - different stages run on different devices.
Sequential = 0: Sequential execution - tasks run one after another.
Single = 9: Single device execution strategy. Executes the entire workload on a single computing device without parallelization. Best for small workloads or when device resources are limited.
Streaming = 7: Streaming execution - continuous processing of data streams.
WorkStealing = 10: Dynamic work stealing for load balancing. Allows idle devices to steal work from busy devices to maintain optimal load distribution. Best for workloads with irregular or unpredictable execution times.