batmat 0.0.18
Batched linear algebra routines
Loading...
Searching...
No Matches
batmat::linalg::micro_kernels::gemm::detail Namespace Reference

Variables

template<class T, class Abi, KernelConfig Conf, StorageOrder OA, StorageOrder OB, StorageOrder OC, StorageOrder OD>
constexpr auto gemm_copy_lut

Variable Documentation

◆ gemm_copy_lut

template<class T, class Abi, KernelConfig Conf, StorageOrder OA, StorageOrder OB, StorageOrder OC, StorageOrder OD>
auto batmat::linalg::micro_kernels::gemm::detail::gemm_copy_lut
constexpr
Initial value:
= make_2d_lut<RowsReg<T, Abi>, ColsReg<T, Abi>>(
[]<index_t Row, index_t Col>(index_constant<Row>, index_constant<Col>) {
return gemm_copy_microkernel<T, Abi, Conf, Row + 1, Col + 1, OA, OB, OC, OD>;
})
batmat::make_2d_lut
consteval auto make_2d_lut(F f)
Returns a 2D array of the form:
Definition lut.hpp:25
batmat::linalg::micro_kernels::gemm::gemm_copy_microkernel
void gemm_copy_microkernel(uview< const T, Abi, OA > A, uview< const T, Abi, OB > B, std::optional< uview< const T, Abi, OC > > C, uview< T, Abi, OD > D, index_t k) noexcept
Generalized matrix multiplication D = C ± A⁽ᵀ⁾ B⁽ᵀ⁾. Single register block.
Definition gemm.tpp:36
batmat::linalg::micro_kernels::gemm::ColsReg
constexpr index_t ColsReg
Definition gemm.hpp:38
batmat::index_constant
std::integral_constant< index_t, I > index_constant
Definition lut.hpp:10

Definition at line 47 of file gemm.hpp.