Performs a Hermitian rank-2k update.
event her2k(queue &exec_queue, uplo upper_lower, transpose trans, std::int64_t n, std::int64_t k, T alpha, const T* a, std::int64_t lda, const T* b, std::int64_t ldb, T_real beta, T* c, std::int64_t ldc, const vector_class<event> &dependencies = {});
The USM version of her2k supports the following precisions and devices:
T | T_real | Devices Supported |
---|---|---|
std::complex<float> | float | Host, CPU, and GPU |
std::complex<double> | double | Host, CPU, and GPU |
The her2k routines perform a rank-2k update of an n x n Hermitian matrix C by general matrices A and B. If trans = transpose::nontrans. The operation is defined as
C <- alpha*A*BH + conjg(alpha)B*AH + beta*C
where A is n x k and B is k x n.
If trans = transpose::conjtrans, the operation is defined as:
C <- alpha*AH*B + conjg(alpha)*BH*A + beta*C
where A is k x n and B is n x k.
In both cases:
alpha is a complex scalar and beta is a real scalar.
C is a Hermitian matrix and A, B are general matrices.
The inner dimension of both matrix multiplications is k.
The queue where the routine should be executed.
Specifies whether A's data is stored in its upper or lower triangle. See Data Types for more details.
Specifies the operation to apply, as described above. Supported operations are transpose::nontrans and transpose::conjtrans.
The number of rows and columns in C. The value of n must be at least zero.
The inner dimension of matrix multiplications. The value of k must be at least equal to zero.
Complex scaling factor for the rank-2k update.
Pointer to input matrix A. If trans = transpose::nontrans, A is an n-by-k matrix so the array a must have size at least lda*k (respectively, lda*n) if column (respectively, row) major layout is used to store matrices. Otherwise, A is an k-by-n matrix so the array a must have size at least lda*n (respectively, lda*k) if column (respectively, row) major layout is used to store matrices. See Matrix and Vector Storage for more details.
Leading dimension of A. If matrices are stored using column major layout, lda must be at least n if trans=transpose::nontrans, and at least k otherwise. If matrices are stored using row major layout, lda must be at least k if trans=transpose::nontrans, and at least n otherwise. Must be positive.
Real scaling factor for matrix C.
Pointer to input matrix B. If trans = transpose::nontrans, B is an k-by-n matrix so the array b must have size at least ldb*n (respectively, ldb*k) if column (respectively, row) major layout is used to store matrices. Otherwise, B is an n-by-k matrix so the array b must have size at least ldb*k (respectively, ldb*n) if column (respectively, row) major layout is used to store matrices. See Matrix and Vector Storage for more details.
Leading dimension of B. If matrices are stored using column major layout, ldb must be at least k if trans = transpose::nontrans, and at least n otherwise. If matrices are stored using row major layout, ldb must be at least n if trans = transpose::nontrans, and at least k otherwise. Must be positive.
Pointer to input/output matrix C. Must have size at least ldc*n. See Matrix and Vector Storage for more details.
Leading dimension of C. Must be positive and at least n.
List of events to wait for before starting computation, if any. If omitted, defaults to no dependencies.
Pointer to the output matrix, overwritten by the updated C matrix.
Output event to wait on to ensure computation is complete.