batmat 0.0.19
Batched linear algebra routines
Loading...
Searching...
No Matches
symv.hpp
Go to the documentation of this file.
1#pragma once
2
5#include <batmat/micro-kernels/symv/export.h>
7#include <optional>
8
10
11struct BATMAT_LINALG_SYMV_EXPORT KernelConfig {
12 bool negate = false;
14};
15
16template <class T, class Abi, KernelConfig Conf, index_t RowsReg, StorageOrder OA>
19 uview<T, Abi, StorageOrder::ColMajor> D, index_t k) noexcept;
20
21template <class T, class Abi, KernelConfig Conf, StorageOrder OA>
22BATMAT_LINALG_SYMV_EXPORT void symv_copy_register(view<const T, Abi, OA> A, view<const T, Abi> B,
23 std::optional<view<const T, Abi>> C,
24 view<T, Abi> D) noexcept;
25
26template <class T, class Abi>
27constexpr index_t RowsReg = 2 * gemm::RowsReg<T, Abi>;
28
29} // namespace batmat::linalg::micro_kernels::symv
constexpr index_t RowsReg
Register block size of the matrix-matrix multiplication micro-kernels.
Definition avx-512.hpp:13
void symv_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 symv.tpp:72
void symv_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
Symmetric matrix-vector multiplication d = c ± A b. Single register block.
Definition symv.tpp:23
simd_view_types< std::remove_const_t< T >, Abi >::template view< T, Order > view
Definition uview.hpp:70