Enum KernelType

Namespace

DotCompute.Abstractions.Types

Assembly

DotCompute.Abstractions.dll

Enumeration of kernel types for cross-platform kernel categorization and optimization.

public enum KernelType

Fields

ComputeIntensive = 9

Compute-intensive operations with transcendental functions. Optimized for floating-point computation and mathematical operations.

Custom = 10

Custom kernel implementation. User-defined kernel with arbitrary computation pattern.

ElementWise = 1

Element-wise operation kernel. Each thread operates on independent data elements. Examples: vector addition, scalar multiplication, activation functions.

Fused = 11

Fused kernel combining multiple operations. Result of kernel fusion optimization.

Generic = 0

Generic kernel type for unknown or unsupported patterns. Uses fallback implementation with basic optimization.

MatrixMultiply = 5

Matrix multiplication operations using cache-optimized algorithms. Benefits from Tensor Core acceleration on supported hardware.

MemoryIntensive = 8

Memory-intensive operations optimized for bandwidth utilization. Focuses on efficient memory access patterns and data movement.

Reduction = 6

Reduction operations (sum, max, min, etc.) with logarithmic scaling. Combines multiple values into a single result.

Transpose = 7

Matrix transpose kernel. Rearranges matrix data layout in memory.

VectorAdd = 2

Vector addition operations (element-wise addition of two vectors). Optimized with SIMD instructions for parallel processing.

VectorMultiply = 3

Vector multiplication operations (element-wise multiplication of two vectors). Optimized with SIMD instructions for parallel processing.

VectorScale = 4

Vector scaling operations (multiplication of vector by scalar). Optimized with SIMD instructions and vectorized scalar broadcasting.