HERK - Hermitian rank-k update to a matrix

Functions
clblasStatus	clblasCherk (clblasOrder order, clblasUplo uplo, clblasTranspose transA, size_t N, size_t K, float alpha, const cl_mem A, size_t offa, size_t lda, float beta, cl_mem C, size_t offc, size_t ldc, cl_uint numCommandQueues, cl_command_queue *commandQueues, cl_uint numEventsInWaitList, const cl_event *eventWaitList, cl_event *events)
	Rank-k update of a hermitian matrix with float-complex elements. More...
clblasStatus	clblasZherk (clblasOrder order, clblasUplo uplo, clblasTranspose transA, size_t N, size_t K, double alpha, const cl_mem A, size_t offa, size_t lda, double beta, cl_mem C, size_t offc, size_t ldc, cl_uint numCommandQueues, cl_command_queue *commandQueues, cl_uint numEventsInWaitList, const cl_event *eventWaitList, cl_event *events)
	Rank-k update of a hermitian matrix with double-complex elements. More...

Detailed Description

Function Documentation

clblasStatus clblasCherk	(	clblasOrder	order,
		clblasUplo	uplo,
		clblasTranspose	transA,
		size_t	N,
		size_t	K,
		float	alpha,
		const cl_mem	A,
		size_t	offa,
		size_t	lda,
		float	beta,
		cl_mem	C,
		size_t	offc,
		size_t	ldc,
		cl_uint	numCommandQueues,
		cl_command_queue *	commandQueues,
		cl_uint	numEventsInWaitList,
		const cl_event *	eventWaitList,
		cl_event *	events
	)

Rank-k update of a hermitian matrix with float-complex elements.

Rank-k updates:

• \( C \leftarrow \alpha A A^H + \beta C \)
• \( C \leftarrow \alpha A^H A + \beta C \)

where C is a hermitian matrix.

Parameters

[in]	order	Row/column order.
[in]	uplo	The triangle in matrix C being referenced.
[in]	transA	How matrix A is to be transposed.
[in]	N	Number of rows and columns in matrix C.
[in]	K	Number of columns of the matrix A if it is not transposed, and number of rows otherwise.
[in]	alpha	The factor of matrix A.
[in]	A	Buffer object storing the matrix A.
[in]	offa	Offset in number of elements for the first element in matrix A.
[in]	lda	Leading dimension of matrix A. It cannot be less than K if A is in the row-major format, and less than N otherwise.
[in]	beta	The factor of the matrix C.
[out]	C	Buffer object storing matrix C.
[in]	offc	Offset in number of elements for the first element in matrix C.
[in]	ldc	Leading dimension of matric C. It cannot be less than N.
[in]	numCommandQueues	Number of OpenCL command queues in which the task is to be performed.
[in]	commandQueues	OpenCL command queues.
[in]	numEventsInWaitList	Number of events in the event wait list.
[in]	eventWaitList	Event wait list.
[in]	events	Event objects per each command queue that identify a particular kernel execution instance.

Returns

• clblasSuccess on success;
• clblasNotInitialized if clblasSetup() was not called;
• clblasInvalidValue if invalid parameters are passed:
  • either N or K is zero, or
  • any of the leading dimensions is invalid;
  • the matrix sizes lead to accessing outsize of any of the buffers;
• clblasInvalidMemObject if either A or C object is invalid, or an image object rather than the buffer one;
• clblasOutOfHostMemory if the library can't allocate memory for internal structures;
• clblasInvalidCommandQueue if the passed command queue is invalid;
• clblasInvalidContext if a context a passed command queue belongs to was released.

Examples:

example_cherk.cpp.

clblasStatus clblasZherk	(	clblasOrder	order,
		clblasUplo	uplo,
		clblasTranspose	transA,
		size_t	N,
		size_t	K,
		double	alpha,
		const cl_mem	A,
		size_t	offa,
		size_t	lda,
		double	beta,
		cl_mem	C,
		size_t	offc,
		size_t	ldc,
		cl_uint	numCommandQueues,
		cl_command_queue *	commandQueues,
		cl_uint	numEventsInWaitList,
		const cl_event *	eventWaitList,
		cl_event *	events
	)

Rank-k update of a hermitian matrix with double-complex elements.

Rank-k updates:

• \( C \leftarrow \alpha A A^H + \beta C \)
• \( C \leftarrow \alpha A^H A + \beta C \)

where C is a hermitian matrix.

Parameters

[in]	order	Row/column order.
[in]	uplo	The triangle in matrix C being referenced.
[in]	transA	How matrix A is to be transposed.
[in]	N	Number of rows and columns in matrix C.
[in]	K	Number of columns of the matrix A if it is not transposed, and number of rows otherwise.
[in]	alpha	The factor of matrix A.
[in]	A	Buffer object storing the matrix A.
[in]	offa	Offset in number of elements for the first element in matrix A.
[in]	lda	Leading dimension of matrix A. It cannot be less than K if A is in the row-major format, and less than N otherwise.
[in]	beta	The factor of the matrix C.
[out]	C	Buffer object storing matrix C.
[in]	offc	Offset in number of elements for the first element in matrix C.
[in]	ldc	Leading dimension of matric C. It cannot be less than N.
[in]	numCommandQueues	Number of OpenCL command queues in which the task is to be performed.
[in]	commandQueues	OpenCL command queues.
[in]	numEventsInWaitList	Number of events in the event wait list.
[in]	eventWaitList	Event wait list.
[in]	events	Event objects per each command queue that identify a particular kernel execution instance.

Returns

• clblasSuccess on success;
• clblasInvalidDevice if a target device does not support floating point arithmetic with double precision;
• the same error codes as the clblasCherk() function otherwise.