hemm: Hermitian matrix multiply

Functions
template<typename scalar_t >
void	blas::impl::hemm (blas::Layout layout, std::vector< blas::Side > const &side, std::vector< blas::Uplo > const &uplo, std::vector< int64_t > const &m, std::vector< int64_t > const &n, std::vector< scalar_t > const &alpha, std::vector< scalar_t * > const &Aarray, std::vector< int64_t > const &lda, std::vector< scalar_t * > const &Barray, std::vector< int64_t > const &ldb, std::vector< scalar_t > const &beta, std::vector< scalar_t * > const &Carray, std::vector< int64_t > const &ldc, size_t batch_size, std::vector< int64_t > &info)
	CPU, variable-size batched version.
template<typename scalar_t >
void	blas::impl::hemm (blas::Layout layout, std::vector< blas::Side > const &side, std::vector< blas::Uplo > const &uplo, std::vector< int64_t > const &m, std::vector< int64_t > const &n, std::vector< scalar_t > const &alpha, std::vector< scalar_t * > const &Aarray, std::vector< int64_t > const &lda, std::vector< scalar_t * > const &Barray, std::vector< int64_t > const &ldb, std::vector< scalar_t > const &beta, std::vector< scalar_t * > const &Carray, std::vector< int64_t > const &ldc, size_t batch_size, std::vector< int64_t > &info, blas::Queue &queue)
	GPU device, variable-size batched version.
template<typename scalar_t >
void	blas::impl::hemm (blas::Layout layout, blas::Side side, blas::Uplo uplo, int64_t m, int64_t n, scalar_t alpha, scalar_t const *A, int64_t lda, scalar_t const *B, int64_t ldb, scalar_t beta, scalar_t *C, int64_t ldc, blas::Queue &queue)
	Mid-level templated wrapper checks and converts arguments, then calls low-level wrapper.
void	blas::internal::hemm (char side, char uplo, blas_int m, blas_int n, std::complex< float > alpha, std::complex< float > const *A, blas_int lda, std::complex< float > const *B, blas_int ldb, std::complex< float > beta, std::complex< float > *C, blas_int ldc)
	Low-level overload wrapper calls Fortran, complex<float> version.
void	blas::internal::hemm (char side, char uplo, blas_int m, blas_int n, std::complex< double > alpha, std::complex< double > const *A, blas_int lda, std::complex< double > const *B, blas_int ldb, std::complex< double > beta, std::complex< double > *C, blas_int ldc)
	Low-level overload wrapper calls Fortran, complex<double> version.
template<typename scalar_t >
void	blas::impl::hemm (blas::Layout layout, blas::Side side, blas::Uplo uplo, int64_t m, int64_t n, scalar_t alpha, scalar_t const *A, int64_t lda, scalar_t const *B, int64_t ldb, scalar_t beta, scalar_t *C, int64_t ldc)
	Mid-level templated wrapper checks and converts arguments, then calls low-level wrapper.

Detailed Description

Function Documentation

hemm() [1/2]

template<typename scalar_t >

void blas::impl::hemm	(	blas::Layout	layout,
		std::vector< blas::Side > const &	side,
		std::vector< blas::Uplo > const &	uplo,
		std::vector< int64_t > const &	m,
		std::vector< int64_t > const &	n,
		std::vector< scalar_t > const &	alpha,
		std::vector< scalar_t * > const &	Aarray,
		std::vector< int64_t > const &	lda,
		std::vector< scalar_t * > const &	Barray,
		std::vector< int64_t > const &	ldb,
		std::vector< scalar_t > const &	beta,
		std::vector< scalar_t * > const &	Carray,
		std::vector< int64_t > const &	ldc,
		size_t	batch_size,
		std::vector< int64_t > &	info
	)

CPU, variable-size batched version.

Mid-level templated wrapper checks and converts arguments, then makes individual routine calls in parallel.

hemm() [2/2]

template<typename scalar_t >

void blas::impl::hemm	(	blas::Layout	layout,
		std::vector< blas::Side > const &	side,
		std::vector< blas::Uplo > const &	uplo,
		std::vector< int64_t > const &	m,
		std::vector< int64_t > const &	n,
		std::vector< scalar_t > const &	alpha,
		std::vector< scalar_t * > const &	Aarray,
		std::vector< int64_t > const &	lda,
		std::vector< scalar_t * > const &	Barray,
		std::vector< int64_t > const &	ldb,
		std::vector< scalar_t > const &	beta,
		std::vector< scalar_t * > const &	Carray,
		std::vector< int64_t > const &	ldc,
		size_t	batch_size,
		std::vector< int64_t > &	info,
		blas::Queue &	queue
	)

GPU device, variable-size batched version.

Mid-level templated wrapper checks and converts arguments, then makes individual routine calls in parallel.