batmat::linalg::micro_kernels::small_potrf Namespace Reference

Classes
struct	KernelConfig

Typedefs
template<class T>
using	scalar_view = uview<T, datapar::scalar_abi<std::remove_const_t<T>>, StorageOrder::ColMajor>

Functions
template<class T, index_t NC>
void	potrf_trsm_microkernel (index_t k, scalar_view< const T > A, scalar_view< T > L) noexcept
template<class T, index_t RowsReg, index_t ColsReg>
void	potrf_syrk_microkernel (index_t k, scalar_view< const T > L21, scalar_view< const T > A22, scalar_view< T > L22) noexcept
	Outer product for updating the bottom right tail during Cholesky factorization.
template<class T, index_t RowsReg = 4>
void	small_potrf (view< const T, datapar::scalar_abi< T > > A, view< T, datapar::scalar_abi< T > > L, index_t n=-1) noexcept
template<class T, KernelConfig Conf, index_t NC, index_t NR>
void	syrk_potrf_trsm_microkernel (index_t m, index_t k0, scalar_view< const T > L20, index_t k, scalar_view< const T > L21, scalar_view< const T > A22, scalar_view< T > L22) noexcept
	Left-looking variant of small_potrf, which updates the current block with the outer product of the previously computed part L21.
template<class T, KernelConfig Conf, index_t RowsReg = 4, index_t S = 8>
void	small_potrf_left (view< const T, datapar::scalar_abi< T > > A, view< const T, datapar::scalar_abi< T > > C, view< T, datapar::scalar_abi< T > > D) noexcept

---

Class Documentation

batmat::linalg::micro_kernels::small_potrf::KernelConfig

struct batmat::linalg::micro_kernels::small_potrf::KernelConfig

Class Members
bool	negate_A = false	Whether to compute chol(C - AAᵀ) instead of chol(C + AAᵀ).

Typedef Documentation

scalar_view

template<class T>

using batmat::linalg::micro_kernels::small_potrf::scalar_view = uview<T, datapar::scalar_abi<std::remove_const_t<T>>, StorageOrder::ColMajor>

Definition at line 15 of file small-potrf.hpp.

Function Documentation

potrf_trsm_microkernel()

template<class T, index_t NC>

void batmat::linalg::micro_kernels::small_potrf::potrf_trsm_microkernel ( index_t k, scalar_view< const T > A, scalar_view< T > L )

noexcept

Definition at line 18 of file small-potrf.tpp.

potrf_syrk_microkernel()

template<class T, index_t RowsReg, index_t ColsReg>

void batmat::linalg::micro_kernels::small_potrf::potrf_syrk_microkernel ( index_t k, scalar_view< const T > L21, scalar_view< const T > A22, scalar_view< T > L22 )

noexcept

Outer product for updating the bottom right tail during Cholesky factorization.

Parameters

A21	rows×ColsReg
A22	rows×RowsReg

Definition at line 72 of file small-potrf.tpp.

small_potrf()

template<class T, index_t RowsReg = 4>

void batmat::linalg::micro_kernels::small_potrf::small_potrf ( view< const T, datapar::scalar_abi< T > > A, view< T, datapar::scalar_abi< T > > L, index_t n = -1 )

noexcept

Definition at line 118 of file small-potrf.tpp.

syrk_potrf_trsm_microkernel()

template<class T, KernelConfig Conf, index_t NC, index_t NR>

void batmat::linalg::micro_kernels::small_potrf::syrk_potrf_trsm_microkernel ( index_t m, index_t k0, scalar_view< const T > L20, index_t k, scalar_view< const T > L21, scalar_view< const T > A22, scalar_view< T > L22 )

noexcept

Left-looking variant of small_potrf, which updates the current block with the outer product of the previously computed part L21.

An additional part L20 can also be added to the update.

Parameters

L20	m×k0
L21	m×k
A22	m×NC
L22	m×NC

Definition at line 210 of file small-potrf.tpp.

small_potrf_left()

template<class T, KernelConfig Conf, index_t RowsReg = 4, index_t S = 8>

void batmat::linalg::micro_kernels::small_potrf::small_potrf_left ( view< const T, datapar::scalar_abi< T > > A, view< const T, datapar::scalar_abi< T > > C, view< T, datapar::scalar_abi< T > > D )

noexcept

Definition at line 297 of file small-potrf.tpp.