hemm: Hermitian matrix multiply

Functions
template<Target target = Target::HostTask, typename scalar_t >
void	slate::internal::hemm (Side side, scalar_t alpha, HermitianMatrix< scalar_t > &&A, Matrix< scalar_t > &&B, scalar_t beta, Matrix< scalar_t > &&C, int priority)
	Hermitian matrix multiply to update trailing matrix, where A is a single tile.
template<typename scalar_t >
void	slate::internal::hemm (internal::TargetType< Target::HostTask >, Side side, scalar_t alpha, HermitianMatrix< scalar_t > &A, Matrix< scalar_t > &B, scalar_t beta, Matrix< scalar_t > &C, int priority)
	Hermitian matrix multiply to update trailing matrix.
template<typename scalar_t >
void	slate::internal::hemm (internal::TargetType< Target::HostNest >, Side side, scalar_t alpha, HermitianMatrix< scalar_t > &A, Matrix< scalar_t > &B, scalar_t beta, Matrix< scalar_t > &C, int priority)
	Hermitian matrix multiply to update trailing matrix.

Detailed Description

Function Documentation

hemm() [1/3]

template<typename scalar_t >

void slate::internal::hemm	(	internal::TargetType< Target::HostNest >	,
		Side	side,
		scalar_t	alpha,
		HermitianMatrix< scalar_t > &	A,
		Matrix< scalar_t > &	B,
		scalar_t	beta,
		Matrix< scalar_t > &	C,
		int	priority
	)

Hermitian matrix multiply to update trailing matrix.

Host nested OpenMP implementation.

hemm() [2/3]

template<typename scalar_t >

void slate::internal::hemm	(	internal::TargetType< Target::HostTask >	,
		Side	side,
		scalar_t	alpha,
		HermitianMatrix< scalar_t > &	A,
		Matrix< scalar_t > &	B,
		scalar_t	beta,
		Matrix< scalar_t > &	C,
		int	priority
	)

Hermitian matrix multiply to update trailing matrix.

Host OpenMP task implementation.

hemm() [3/3]

template<Target target = Target::HostTask, typename scalar_t >

void slate::internal::hemm	(	Side	side,
		scalar_t	alpha,
		HermitianMatrix< scalar_t > &&	A,
		Matrix< scalar_t > &&	B,
		scalar_t	beta,
		Matrix< scalar_t > &&	C,
		int	priority
	)

Hermitian matrix multiply to update trailing matrix, where A is a single tile.

If side = left, B and C are each a single block row; if side = right, B and C are each a single block col. Unlike most BLAS operations, here op(B) and op(C) must be both the same, either both NoTrans or both ConjTrans. Dispatches to target implementations.