24 #ifndef __ARM_COMPUTE_GRAPH_TYPES_H__ 25 #define __ARM_COMPUTE_GRAPH_TYPES_H__
Affinity at buffer level.
NEON capable target device.
Graph configuration structure Device target types.
Affinity at offset level.
EltwiseOperation
Supported Element-wise operations.
Fast math disabled for Convolution layer.
DataLayoutDimension
Supported tensor data layout dimensions.
Optimized 3x3 direct depthwise convolution.
DimensionRoundingType
Dimension rounding type when down-scaling on CNNs.
Normalization Layer Information class.
Default approach using internal heuristics.
Activation Layer Information class.
Target target
Node target.
This file contains all available output stages for GEMMLowp on OpenCL.
ActivationFunction
Available activation functions.
GLES compute capable target device.
bool use_transition_memory_manager
Use a memory manager to manager transition buffer memory.
bool use_function_memory_manager
Use a memory manager to manage per-funcion auxilary memory.
DepthwiseConvolutionMethod
Supported Depthwise Convolution layer methods.
Generic GEMV based depthwise convolution.
Padding and stride information class.
constexpr EdgeID EmptyEdgeID
Winograd based convolution.
FastMathHint
Enable or disable fast math for Convolution layer.
PoolingType
Available pooling types.
constexpr NodeID EmptyNodeID
Constant EdgeID specifying an equivalent of null edge.
Class for specifying the size of an image or rectangle.
bool use_tuner
Use a tuner in tunable backends.
std::string name
Node name.
ConvolutionMethod
Supported Convolution layer methods.
int num_threads
Number of threads to use (thread capable backends), if 0 the backend will auto-initialize, if -1 the backend will stay as it is.
fixed_point< T > max(fixed_point< T > x, fixed_point< T > y)
constexpr TensorID NullTensorID
Constant NodeID specifying an equivalent of null node.
Arithmetic multiplication.
DataType
Available data types.
OpenCL capable target device.
Pooling Layer Information class.
DataLayout
Supported tensor data layouts.
NormType
The normalization type used for the normalization layer.
MemoryManagerAffinity
Backend Memory Manager affinity.
Fast math enabled for Convolution layer.