Accelerate

Buffer Protocols 4

Protocols that describe the contiguous and matrix-shaped storage Accelerate routines read from and write to.

• Pr
  AccelerateBufferiOS 13.0+

  An object composed of count elements that are stored contiguously in memory.
• Pr
  AccelerateMatrixBufferiOS 18.0+

  A `rowCount` x `columnCount` matrix.
• Pr
  AccelerateMutableBufferiOS 13.0+

  A mutable object composed of count elements that are stored contiguously
• Pr
  AccelerateMutableMatrixBufferiOS 18.0+

  A mutable `rowCount` x `columnCount` matrix.

Image Buffers and Conversion 8

The vImage buffer, converters, and format descriptors for processing and converting pixel data.

• St
  vImage_Buffer

  @typedef vImage_Buffer
• Cl
  vImageConverter

  @class vImageConverterRef
• Cl
  vImageCVImageFormat

  @class vImageCVImageFormatRef
• Cl
  vImageConstCVImageFormat

  An immutable object that describes the pixel format of a Core Video image buffer.
• St
  vImage_CGImageFormat

  @struct vImage_CGImageFormat
• Pr
  InitializableFromCGImageiOS 16.0+

  A pixel format that supports initialization from a `CGImage`.
• St
  vImageMDTableUsageHint

  Constants that hint how a vImage multidimensional lookup table will be used.
• St
  vImage_InterpolationMethod

  Constants that specify the interpolation method used when sampling pixels.

Pixel Formats 7

Protocols and structures that describe single-plane, multi-plane, and channel pixel layouts.

• Pr
  PixelFormatiOS 16.0+

  A protocol describing the layout and channel composition of a vImage pixel format.
• Pr
  StaticPixelFormatiOS 16.0+

  A pixel format that is known at compile time.
• Pr
  SinglePlanePixelFormatiOS 16.0+

  A pixel format that contains a single underlying vImage buffer.
• Pr
  MultiplePlanePixelFormatiOS 16.0+

  A pixel format that contains multiple homogeneous planes represented by multiple underlying vImage buffers.
• St
  vImageChannelDescription

  @typedef vImageChannelDescription
• St
  vImageARGBType

  @typedef vImageARGBType
• St
  vImageYpCbCrType

  @typedef vImageYpCbCrType

Color and Geometry Transforms 11

Color-space matrices, white points, transfer functions, and affine and perspective geometry for vImage.

• St
  vImageTransferFunction

  @functiongroup Low level colorspace initialization
• St
  vImageRGBPrimaries

  A structure that specifies the RGB chromaticity primaries of a color space.
• St
  vImageWhitePoint

  A structure that specifies the reference white point of a color space.
• St
  vImage_AffineTransform

  @typedef vImage_AffineTransform
• St
  vImage_AffineTransform_Double

  A structure that represents a double-precision affine geometric transform for images.
• St
  vImage_PerpsectiveTransform

  @typedef vImage_PerpsectiveTransform
• St
  vImage_YpCbCrToARGBMatrix

  @struct vImage_YpCbCrToARGBMatrix
• St
  vImage_YpCbCrToARGB

  @typedef vImage_YpCbCrToARGB
• St
  vImage_ARGBToYpCbCrMatrix

  @typedef vImage_ARGBToYpCbCrMatrix
• St
  vImage_ARGBToYpCbCr

  @typedef vImage_ARGBToYpCbCr
• St
  vImage_YpCbCrPixelRange

  @typedef vImage_YpCbCrPixelRange

Digital Signal Processing 14

Protocols and complex types for Fourier transforms, biquad filtering, and DSP conversions.

• Pr
  vDSP_BiquadFunctionsiOS 13.0+

  A protocol that defines the operations for a vDSP biquad infinite impulse response filter.
• Pr
  vDSP_DFTFunctions

  A protocol that defines the discrete Fourier transform operations available in vDSP.
• Pr
  vDSP_DiscreteFourierTransformableiOS 15.0+

  A protocol for element types that can undergo a vDSP discrete Fourier transform.
• Pr
  vDSP_DiscreteTransformLifecycleFunctionsiOS 15.0+

  A protocol that defines setup and teardown operations for vDSP discrete transforms.
• Pr
  vDSP_FloatingPointBiquadFilterableiOS 13.0+

  A protocol for floating-point element types that vDSP biquad filters can process.
• Pr
  vDSP_FloatingPointConvertableiOS 13.0+

  A protocol for floating-point types that vDSP can convert between representations.
• Pr
  vDSP_FloatingPointDiscreteFourierTransformable

  A protocol for floating-point types that support vDSP discrete Fourier transforms.
• Pr
  vDSP_FloatingPointGeneratableiOS 13.0+

  Types that support vectorized window generation.
• Pr
  vDSP_FourierTransformFunctionsiOS 13.0+

  A protocol that defines the fast Fourier transform operations available in vDSP.
• Pr
  vDSP_FourierTransformableiOS 13.0+

  A protocol for element types that can undergo a vDSP fast Fourier transform.
• Pr
  vDSP_IntegerConvertableiOS 13.0+

  A protocol for integer types that vDSP can convert to and from other representations.
• St
  DSPComplex

  A structure that represents a single-precision complex number for DSP operations.
• St
  DSPDoubleComplex

  A structure that represents a double-precision complex number for DSP operations.
• St
  DSPSplitComplex

  A structure that represents a complex signal as separate real and imaginary buffers.

Neural Network Building Blocks 13

BNNS function enumerations that define activations, pooling, loss, and arithmetic operations.

• St
  BNNSActivationFunction

  @abstract Activation layer function
• St
  BNNSActivation

  @abstract Common activation function parameters
• St
  BNNSPoolingFunction

  @abstract Pooling layer function
• St
  BNNSLossFunction

  @abstract Loss layer function
• St
  BNNSLossReductionFunction

  @abstract Loss layer reduction function
• St
  BNNSArithmeticFunction

  @abstract Arithmetic function
• St
  BNNSArithmeticUnary

  @abstract Arithmetic struct that hold 1 input and an output descriptor
• St
  BNNSArithmeticBinary

  @abstract Arithmetic struct that holds 2 input descriptors and an output descriptor
• St
  BNNSArithmeticTernary

  @abstract Arithmetic struct that holds 3 input descriptors and an output descriptor
• St
  BNNSReduceFunction

  @abstract Reduction layer functions
• St
  BNNSNormType

  @abstract Type of norm
• St
  BNNSQuantizerFunction

  @abstract Quantizer function
• St
  BNNSRandomGeneratorMethod

  @abstract Random number generation methods

Neural Network Layer Parameters 28

Descriptors that configure individual BNNS layers such as convolution, pooling, and attention.

• St
  BNNSLayerParametersConvolution

  @abstract Convolution parameters
• St
  BNNSLayerParametersFullyConnected

  @abstract Fully connected layer parameters
• St
  BNNSLayerParametersPooling

  @abstract Pooling layer parameters
• St
  BNNSLayerParametersActivation

  @abstract Activation/Conversion layer parameters
• St
  BNNSLayerParametersNormalization

  @abstract Normalization layer parameters
• St
  BNNSLayerParametersDropout

  @abstract Dropout Layer Fields
• St
  BNNSLayerParametersLSTM

  @abstract LSTM layer parameters
• St
  BNNSLayerParametersArithmetic

  @abstract Arithmetic layer parameters
• St
  BNNSLayerParametersPermute

  @abstract Permute Layer Parameters
• St
  BNNSLayerParametersTensorContraction

  @abstract Tensor Contraction Layer Parameters
• St
  BNNSLayerParametersGram

  @abstract Gram Matrix Layer Parameters
• St
  BNNSLayerParametersResize

  @abstract Resize Layer Parameters
• St
  BNNSLayerParametersCropResize

  @abstract Crop and Resize Layer Parameters
• St
  BNNSLayerParametersBroadcastMatMul

  @abstract Broadcast Matrix Multiplication Layer Parameters
• St
  BNNSLayerParametersMultiheadAttention

  @abstract Multihead Attention Parameters
• St
  BNNSLayerParametersReduction

  @discussion computing reduction function across selected dimensions
• St
  BNNSLayerParametersPadding

  @abstract Padding Layer Fields
• St
  BNNSLayerParametersEmbedding

  @abstract Embedding Layer Fields
• St
  BNNSLayerParametersQuantization

  @abstract Quantization layer parameters
• St
  BNNSLayerParametersLossBase

  @abstract Loss layer parameters (basic version)
• St
  BNNSLayerParametersLossSoftmaxCrossEntropy

  @abstract Loss layer parameters (Softmax Cross Entropy version)
• St
  BNNSLayerParametersLossSigmoidCrossEntropy

  @abstract Loss layer parameters (Sigmoid Cross Entropy version)
• St
  BNNSLayerParametersLossHuber

  @abstract Loss layer parameters (Huber version)
• St
  BNNSLayerParametersLossYolo

  @abstract Loss layer parameters (YoLo version)
• St
  BNNSMHAProjectionParameters

  @abstract Multihead Attention Projection Parameters
• St
  BNNSConvolutionLayerParameters

  @abstract Convolution parameters (DEPRECATED, Use BNNSLayerParametersConvolution)
• St
  BNNSFullyConnectedLayerParameters

  @abstract Fully connected layer parameters (DEPRECATED, Use BNNSLayerParametersFullyConnected)
• St
  BNNSPoolingLayerParameters

  @abstract Pooling layer parameters (DEPRECATED, Use BNNSLayerParametersPooling)

Neural Network Optimizers 11

Optimizer functions and their field descriptors for training BNNS networks.

• Pr
  BNNSOptimizer

  ===----------------------------------------------------------------------===//
• St
  BNNSOptimizerFunction

  @abstract Optimizer layer function
• St
  BNNSOptimizerRegularizationFunction

  @abstract Optimizer regularization function
• St
  BNNSOptimizerSGDMomentumVariant

  @abstract Optimizer SGD with momentum variant
• St
  BNNSOptimizerClippingFunction

  @abstract Optimizer clipping function
• St
  BNNSOptimizerSGDMomentumFields

  @abstract SGD with Momentum Optimizer Fields
• St
  BNNSOptimizerSGDMomentumWithClippingFields

  @abstract SGD with Momentum and gradient clipping Optimizer Fields
• St
  BNNSOptimizerAdamFields

  @abstract Adam and AdamW Fields
• St
  BNNSOptimizerAdamWithClippingFields

  @abstract Adam and AdamW with gradient clipping Fields
• St
  BNNSOptimizerRMSPropFields

  @abstract RMSProp Fields
• St
  BNNSOptimizerRMSPropWithClippingFields

  @abstract RMSProp with gradient clipping Fields

Tensors and Data Descriptors 11

Tensor, array, and data structures that describe the shape, type, and layout of BNNS data.

• St
  BNNSTensor

  @abstract Generic Tensor
• St
  BNNSNDArrayDescriptor

  @abstract Generic N-dimensional array
• St
  BNNSImageStackDescriptor

  @abstract Image stack descriptor (DEPRECATED, Use BNNSNDArrayDescriptor)
• St
  BNNSVectorDescriptor

  @abstract Vector format descriptor (DEPRECATED, Use BNNSNDArrayDescriptor)
• St
  BNNSLayerData

  @abstract Common layer parameters (DEPRECATED, Use BNNSNDArrayDescriptor)
• St
  BNNSLSTMGateDescriptor

  @abstract LSTM Gate Layer Descriptor
• St
  BNNSLSTMDataDescriptor

  @abstract contains NDarrat descriptors for inputs and outputs of LSTM
• St
  BNNSSparsityParametersiOS 16.0+

  @abstract parameters to describe sparsity attributes
• St
  BNNSFilterParameters

  @abstract Common filter parameters
• Pr
  FusableLayerParameters

  A protocol for layer parameter types whose operations BNNS can fuse for efficiency.
• Pr
  BNNSScalariOS 14.0+

  A type that can be used as an element in a BNNS n-dimensional array.

Neural Network Enumerations 17

Constants that specify data types, layouts, flags, and modes used throughout BNNS.

• St
  BNNSDataType

  @abstract Storage data type
• St
  BNNSDataLayout

  @abstract Data layout and dimension
• St
  BNNSFlags

  @abstract Filter flags
• St
  BNNSLayerFlags

  @discussion LSTM layer flags
• St
  BNNSDescriptorType

  @abstract BNNS Descriptor Data type
• St
  BNNSFilterType

  @abstract filter type
• St
  BNNSInterpolationMethod

  @abstract Interpolation methods for resize.
• St
  BNNSLinearSamplingMode

  @abstract Interpolation sampling mode.
• St
  BNNSBoxCoordinateMode

  @abstract
• St
  BNNSPaddingMode

  @abstract padding modes for padding.
• St
  BNNSRelationalOperator

  @abstract Relational layer operation type
• St
  BNNSPointerSpecifier

  @abstract Pointer extraction specifiers
• St
  BNNSNDArrayFlags

  @abstract Flags to control behaviors on BNNSNDArrayDescriptors
• St
  BNNSEmbeddingFlags

  @abstract Flags to control behavior of Embedding layers
• St
  BNNSSparsityType

  @constant BNNSSparsityTypeUnstructured - no special pattern in the sparsity
• St
  BNNSTargetSystem

  @abstract Target System values for BNNSGetBestDataLayout* functions
• St
  BNNSShuffleType

  @abstract Shuffle type for BNNSShuffle function

Neural Network Graphs 10

Graph handles, contexts, and options for compiling and executing BNNS computation graphs.

• St
  bnns_graph_tiOS 18.0+

  The compiled version of a graph that can be executed by BNNS.
• St
  bnns_graph_context_tiOS 18.0+

  Wrapper around a `bnns_graph_t` object that adds mutable data storage.
• St
  bnns_graph_compile_options_tiOS 18.0+

  Compilation options used to compile source model to a `bnns_graph_t` object.
• St
  bnns_graph_shape_tiOS 18.0+

  Specifies the shape of an argument
• St
  bnns_graph_argument_t

  Describes data associated with an input or output argument
• St
  bnns_user_message_data_tiOS 18.0+

  Additional user-defined logging argument for user to pass into the message logging callbacks
• St
  BNNSGraphMessageLevel

  Message Level Enum
• St
  BNNSGraphOptimizationPreference

  BNNS Graph Compilation Optimization Preferences
• St
  BNNSGraphArgumentIntent

  Describes the intent of an argument
• St
  BNNSGraphArgumentType

  Used to specify the interpretation of the `arguments` argument to `BNNSGraphContextExecute()`.

Numerical Integration 4

Types that configure and report the status of quadrature-based numerical integration.

• St
  quadrature_status

  @enum quadrature_status
• St
  quadrature_integrator

  @enum quadrature_integrator
• St
  quadrature_integrate_function

  @abstract Function to integrate
• St
  quadrature_integrate_options

  @abstract Integration options

Sparse Matrices and Vectors 21

Structures that represent sparse and dense matrices and vectors and describe their properties.

• St
  sparse_status

  @abstract The type reflecting the status of an operations.
• St
  sparse_matrix_property

  @abstract The matrix property type
• St
  sparse_norm

  @abstract The norm specifier
• St
  SparseKind_t

  A flag to describe the type of matrix represented.
• St
  SparseTriangle_t

  A flag to indicate which triangle of a matrix is used.
• St
  SparseAttributes_t

  A type representing the attributes of a matrix.
• St
  SparseAttributesComplex_t

  A type representing the attributes of a matrix.
• St
  SparseMatrixStructure

  A type representing the sparsity structure of a sparse matrix.
• St
  SparseMatrixStructureComplex

  A type representing the sparsity structure of a sparse complex matrix.
• St
  SparseMatrix_Double

  A type representing a sparse matrix.
• St
  SparseMatrix_Float

  A type representing a sparse matrix.
• St
  SparseMatrix_Complex_Double

  A type representing a sparse complex matrix.
• St
  SparseMatrix_Complex_Float

  A type representing a sparse complex matrix.
• St
  DenseVector_Double

  Contains a dense vector of double values.
• St
  DenseVector_Float

  Contains a dense vector of float values.
• St
  DenseVector_Complex_Double

  Contains a dense vector of double complex values.
• St
  DenseVector_Complex_Float

  Contains a dense vector of float complex values.
• St
  DenseMatrix_Double

  Contains a dense `rowCount` x `columnCount` matrix of double
• St
  DenseMatrix_Float

  Contains a dense `rowCount` x `columnCount` matrix of float
• St
  DenseMatrix_Complex_Double

  Contains a dense `rowCount` x `columnCount` matrix of complex double
• St
  DenseMatrix_Complex_Float

  Contains a dense `rowCount` x `columnCount` matrix of complex float

Sparse Factorization 18

Options and opaque results for direct sparse factorization, subfactoring, and updates.

• St
  SparseStatus_t

  Status field for a factorization.
• St
  SparseFactorization_t

  Types of factorization than can be performed.
• St
  SparseControl_t

  Control flags for matrix factorization.
• St
  SparseOrder_t

  Specifies type of fill-reducing ordering.
• St
  SparseScaling_t

  Specifies type of scaling to be performed.
• St
  SparseSymbolicFactorOptions

  SparseSymbolicFactorOptions
• St
  SparseNumericFactorOptions

  Options that affect the numerical stage of a sparse factorization.
• St
  SparseOpaqueSymbolicFactorization

  A semi-opaque type representing symbolic matrix factorization.
• St
  SparseOpaqueFactorization_Double

  A semi-opaque type representing a matrix factorization in double.
• St
  SparseOpaqueFactorization_Float

  A semi-opaque type representing a matrix factorization in float.
• St
  SparseOpaqueFactorization_Complex_Double

  A semi-opaque type representing a matrix factorization in complex double.
• St
  SparseOpaqueFactorization_Complex_Float

  A semi-opaque type representing a matrix factorization in complex float.
• St
  SparseSubfactor_t

  Types of sub-factor object.
• St
  SparseOpaqueSubfactor_Double

  Represents a sub-factor of the factorization (for example, `L` from `LDL^T`).
• St
  SparseOpaqueSubfactor_Float

  Represents a sub-factor of the factorization (for example, `L` from `LDL^T`).
• St
  SparseOpaqueSubfactor_Complex_Double

  Represents a sub-factor of the factorization (for example, `L` from `LDL^T`).
• St
  SparseOpaqueSubfactor_Complex_Float

  Represents a sub-factor of the factorization (for example, `L` from `LDL^T`).
• St
  SparseUpdate_t

  Low-rank update algorithm selector

Sparse Iterative Solvers 14

Preconditioners, methods, and options for solving sparse systems with iterative algorithms.

• St
  SparsePreconditioner_t

  Types of preconditioner.
• St
  SparseOpaquePreconditioner_Double

  Represents a preconditioner for matrices of double values .
• St
  SparseOpaquePreconditioner_Float

  Represents a preconditioner for matrices of float values .
• St
  SparseOpaquePreconditioner_Complex_Double

  Represents a preconditioner for matrices of complex double values .
• St
  SparseOpaquePreconditioner_Complex_Float

  Represents a preconditioner for matrices of complex float values .
• St
  SparseIterativeStatus_t

  Exit status definitions for sparse iterative methods.
• St
  _SparseIterativeMethodBaseOptions

  @internal This type required for implementation use only.
• St
  SparseCGOptions

  Conjugate Gradient Options.
• St
  SparseGMRESVariant_t

  Specify different variants of GMRES that can be used.
• St
  SparseGMRESOptions

  Right-preconditioned (F/DQ)GMRES Parameters Options.
• St
  SparseLSMRConvergenceTest_t

  Available types of convergence test for LSMR.
• St
  SparseLSMROptions

  LSMR is MINRES specialised for solving least squares.
• St
  SparseIterativeMethod

  General description object for all iterative methods.
• St
  _SparseIterativeMethod_t

  A structure that identifies the underlying sparse iterative method.

BLAS and LAPACK Interop 9

Constants and complex types that bridge to the C BLAS and LAPACK linear-algebra interfaces.

• St
  CBLAS_ORDER

  Constants that specify row-major or column-major storage order for BLAS routines.
• St
  CBLAS_TRANSPOSE

  Constants that specify whether a matrix is transposed in a BLAS operation.
• St
  CBLAS_UPLO

  Constants that specify whether the upper or lower triangle of a matrix is referenced.
• St
  CBLAS_DIAG

  Constants that specify whether a triangular matrix has a unit diagonal.
• St
  CBLAS_SIDE

  Constants that specify whether a matrix multiplies from the left or right in BLAS.
• St
  BLAS_THREADING

  Constants that specify the threading behavior of BLAS routines.
• St
  __CLPK_complex

  A structure that represents a single-precision complex number for LAPACK routines.
• St
  __CLPK_doublecomplex

  A structure that represents a double-precision complex number for LAPACK routines.
• Pr
  OS_la_object

  A protocol representing an opaque linear algebra object used by the Accelerate sparse and dense routines.

Structures 7

• St
  DSPDoubleSplitComplex
• St
  vDSP_uint24
• St
  vDSP_int24
• St
  BLASmacOS 15.0+
• St
  QuadratureiOS 13.0+

  A structure that approximates the definite integral of a function over a finite interval.
• St
  vDSP_SplitComplexDoubleiOS 13.0+
• St
  vDSP_SplitComplexFloatiOS 13.0+

Enumerations 9

• En
  vDSP_DFT_Direction
• En
  vDSP_DFT_RealtoComplex
• En
  vDSP_DCT_Type
• En
  AccelerateMatrixOrderiOS 18.0+

  The ordering of the matrix.
• En
  BNNSiOS 14.0+

  An enum that acts as a namespace for Swift overlays to BNNS.
• En
  BNNSGraphiOS 18.0+
• En
  vDSPiOS 13.0+

  An enum that acts as a namespace for Swift overlays to vDSP based functions.
• En
  vForceiOS 13.0+

  An enum that acts as a namespace for Swift overlays to vForce based functions.
• En
  vImageiOS 13.0+

  An enum that acts as a namespace for Swift overlays to vImage option sets and enums..

Type Aliases 91

• Ty
  vImage_MultidimensionalTable
• Ty
  vImageCVImageFormatError

  @typedef vImageCVImageFormatError
• Ty
  vImageMatrixType

  @typedef vImageMatrixType
• Ty
  vImagePixelCount

  @typedef vImagePixelCount
• Ty
  vImage_CGAffineTransform
• Ty
  vImage_WarpInterpolation

  @enum vImage_WarpInterpolation
• Ty
  Pixel_8

  @typedef Pixel_8
• Ty
  Pixel_F

  @typedef Pixel_F
• Ty
  Pixel_88

  @typedef Pixel_88
• Ty
  Pixel_8888

  @typedef Pixel_8888
• Ty
  Pixel_FFFF

  @typedef Pixel_FFFF
• Ty
  Pixel_16U

  @typedef Pixel_16U
• Ty
  Pixel_16S

  @typedef Pixel_16S
• Ty
  Pixel_16Q12

  @typedef Pixel_16Q12
• Ty
  Pixel_16U16U

  @typedef Pixel_16U16U
• Ty
  Pixel_16S16S

  @typedef Pixel_16S16S

Extends 24