MPSMatrixCopyDescriptor(3) MetalPerformanceShaders.framework MPSMatrixCopyDescriptor(3)NAME
MPSMatrixCopyDescriptor
SYNOPSIS
#import <MPSMatrixCombination.h>
Inherits NSObject.
Instance Methods
(nonnull instancetype) - initWithDevice:count:
(void) - setCopyOperationAtIndex:sourceMatrix:destinationMatrix:offsets:
(nonnull instancetype) - initWithSourceMatrices:destinationMatrices:offsetVector:offset:
(nonnull instancetype) - init
Class Methods
(nonnull instancetype) + descriptorWithSourceMatrix:destinationMatrix:offsets:
Method Documentation
+ (nonnull instancetype) descriptorWithSourceMatrix: (MPSMatrix *__nonnull) sourceMatrix(MPSMatrix *__nonnull)
destinationMatrix(MPSMatrixCopyOffsets) offsets
convenience allocator for single copies
- (nonnull instancetype) init
- (nonnull instancetype) initWithDevice: (nonnull id< MTLDevice >) device(NSUInteger) count
initialize a MPSMatrixCopyDescriptor with default values. Use -setCopyOperationAtIndex:sourceMatrix:destinationMatrix:copyOffsets to
initialize. All indices must be initialized before use.
Parameters:
device The device on which the copy will be performed
count The number of copy operations the object will encode
Returns:
A MPSMatrixCopyDescriptor. It still needs to be initialized with -setCopyOperationAtIndex:sourceMatrix:destinationMatrix:copyOffsets
- (nonnull instancetype) initWithSourceMatrices: (NSArray< MPSMatrix * > *__nonnull) sourceMatrices(NSArray< MPSMatrix * > *__nonnull)
destinationMatrices(MPSVector *__nullable) offsets(NSUInteger) byteOffset
Initialize a MPSMatrixCopyDescriptor using offsets generated on the GPU Use this method when the offsets needed are coming from GPU based
computation.
Parameters:
sourceMatrices A list of matrices from which the matrix data is read
destinationMatrices A list of matrices to which to write the data. The count must match the number of source matrices.
offsets A MPSVector of type MPSDataTypeUInt32 containing the list of offsets, stored as a packed array of MPSMatrixCopyOffsets.
byteOffset A byte offset into the offsets vector where the data starts in 'offsets'. This value must be a multiple of 16.
Returns:
A valid MPSMatrixCopyDescriptor to represent the list of copy operations
- (void) setCopyOperationAtIndex: (NSUInteger) index(MPSMatrix *__nonnull) sourceMatrix(MPSMatrix *__nonnull)
destinationMatrix(MPSMatrixCopyOffsets) offsets
Initialize a MPSMatrixCopyDescriptor using offsets generated on the CPU This is for one at a time intialization of the copy operations
Parameters:
index The index of the copy operation
sourceMatrix The source matrix for this copy operation
destinationMatrix The destination matrix for this copy operation
offsets The offsets to use for the copy operation
Author
Generated automatically by Doxygen for MetalPerformanceShaders.framework from the source code.
Version MetalPerformanceShaders-100 Thu Feb 8 2018 MPSMatrixCopyDescriptor(3)
Check Out this Related Man Page
MPSMatrixDecompositionCholesky(3) MetalPerformanceShaders.framework MPSMatrixDecompositionCholesky(3)NAME
MPSMatrixDecompositionCholesky
SYNOPSIS
#import <MPSMatrixDecomposition.h>
Inherits MPSMatrixUnaryKernel.
Instance Methods
(nonnull instancetype) - initWithDevice:lower:order:
(void) - encodeToCommandBuffer:sourceMatrix:resultMatrix:status:
Additional Inherited Members
Detailed Description
This depends on Metal.framework.
A kernel for computing the Cholesky factorization of a matrix.
A MPSMatrixDecompositionLU object computes one of the following factorizations of a matrix A:
A = L * L**T
A = U**T * U
A is a symmetric positive-definite matrix for which the factorization is to be computed. L and U are lower and upper triangular matrices
respectively.
Method Documentation
- (void) encodeToCommandBuffer: (nonnull id< MTLCommandBuffer >) commandBuffer(MPSMatrix *__nonnull) sourceMatrix(MPSMatrix *__nonnull)
resultMatrix(__nullable id< MTLBuffer >) status
Encode a MPSMatrixDecompositionCholesky kernel into a command Buffer.
Parameters:
commandBuffer A valid MTLCommandBuffer to receive the encoded filter
sourceMatrix A valid MPSMatrix containing the source data. Must have enough space to hold a order x order matrix.
resultMatrix A valid MPSMatrix to contain the result. Must have enough space to hold a order x order matrix.
status A MTLBuffer which indicates the resulting MPSMatrixDecompositionStatus value.
This function encodes the MPSMatrixDecompositionCholesky object to a valid command buffer.
If during the factorization a leading minor of the matrix is found to be not positive definite, MPSMatrixDecompositionNonPositiveDefinite
will be returned in the provided status buffer. Previously computed pivots and the non positive pivot are written to the result, but the
factorization does not complete. The data referenced by the MTLBuffer is not valid until the command buffer has completed execution. If the
matrix return status is not desired NULL may be provided.
If the return status is MPSMatrixDecompositionStatusSuccess, resultMatrix contains the resulting factors in its lower or upper triangular
regions respectively.
This kernel functions either in-place, if the result matrix completely aliases the source matrix, or out-of-place. If there is any partial
overlap between input and output data the results are undefined.
- (nonnull instancetype) initWithDevice: (nonnull id< MTLDevice >) device(BOOL) lower(NSUInteger) order
Initialize an MPSMatrixDecompositionCholesky object on a device
Parameters:
device The device on which the kernel will execute.
lower A boolean value indicating if the lower triangular part of the source matrix is stored. If lower = YES the lower triangular part
will be used and the factor will be written to the lower triangular part of the result, otherwise the upper triangular part will be
used and the factor will be written to the upper triangular part.
order The number of rows and columns in the source matrix.
Returns:
A valid MPSMatrixDecompositionCholesky object or nil, if failure.
Author
Generated automatically by Doxygen for MetalPerformanceShaders.framework from the source code.
Version MetalPerformanceShaders-100 Thu Feb 8 2018 MPSMatrixDecompositionCholesky(3)