batmat::linalg::micro_kernels::gemv Namespace Reference

Classes
struct	KernelConfig

Functions
template<class T, class Abi, KernelConfig Conf, index_t RowsReg, StorageOrder OA>
void	gemv_copy_microkernel (const uview< const T, Abi, OA > A, const uview< const T, Abi, StorageOrder::ColMajor > B, const std::optional< uview< const T, Abi, StorageOrder::ColMajor > > C, const uview< T, Abi, StorageOrder::ColMajor > D, const index_t k) noexcept
	Generalized matrix-vector multiplication d = c ± A⁽ᵀ⁾ b. Single register block.
template<class T, class Abi, KernelConfig Conf, StorageOrder OA>
void	gemv_copy_register (const view< const T, Abi, OA > A, const view< const T, Abi > B, const std::optional< view< const T, Abi > > C, const view< T, Abi > D) noexcept
	Generalized matrix multiplication d = c ± A⁽ᵀ⁾ b. Using register blocking.

Variables
template<class T, class Abi>
constexpr index_t	RowsReg = 2 * gemm::RowsReg<T, Abi>
template<class T, class Abi, KernelConfig Conf, StorageOrder OA>
const constinit auto	gemv_copy_lut

---

Class Documentation

batmat::linalg::micro_kernels::gemv::KernelConfig

struct batmat::linalg::micro_kernels::gemv::KernelConfig

Class Members
bool	negate = false
int	shift_A = 0
int	shift_B = 0
int	rotate_C = 0
int	rotate_D = rotate_C
int	mask_D = rotate_D

Function Documentation

gemv_copy_microkernel()

template<class T, class Abi, KernelConfig Conf, index_t RowsReg, StorageOrder OA>

void batmat::linalg::micro_kernels::gemv::gemv_copy_microkernel ( uview< const T, Abi, OA > A, uview< const T, Abi, StorageOrder::ColMajor > B, std::optional< uview< const T, Abi, StorageOrder::ColMajor > > C, uview< T, Abi, StorageOrder::ColMajor > D, index_t k )

noexcept

Generalized matrix-vector multiplication d = c ± A⁽ᵀ⁾ b. Single register block.

Definition at line 16 of file gemv.tpp.

gemv_copy_register()

template<class T, class Abi, KernelConfig Conf, StorageOrder OA>

void batmat::linalg::micro_kernels::gemv::gemv_copy_register ( view< const T, Abi, OA > A, view< const T, Abi > B, std::optional< view< const T, Abi > > C, view< T, Abi > D )

noexcept

Generalized matrix multiplication d = c ± A⁽ᵀ⁾ b. Using register blocking.

Definition at line 79 of file gemv.tpp.

Variable Documentation

RowsReg

template<class T, class Abi>

index_t batmat::linalg::micro_kernels::gemv::RowsReg = 2 * gemm::RowsReg<T, Abi>

constexpr

Definition at line 30 of file gemv.hpp.

gemv_copy_lut

template<class T, class Abi, KernelConfig Conf, StorageOrder OA>

const constinit auto batmat::linalg::micro_kernels::gemv::gemv_copy_lut

inlineconstinit

Initial value:

                                      =
make_1d_lut<RowsReg<T, Abi>>([]<index_t Row>(index_constant<Row>) {
    return gemv_copy_microkernel<T, Abi, Conf, Row + 1, OA>;
})

Definition at line 33 of file gemv.hpp.