batmat main
Batched linear algebra routines
Loading...
Searching...
No Matches
gemv.hpp File Reference
#include <batmat/linalg/structure.hpp>
#include <batmat/linalg/uview.hpp>
#include <batmat/micro-kernels/gemv/export.h>
#include <batmat/platform/platform.hpp>
#include <optional>

Go to the source code of this file.

Classes

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

Namespaces

namespace  batmat
namespace  batmat::linalg
namespace  batmat::linalg::micro_kernels
namespace  batmat::linalg::micro_kernels::gemv

Functions

template<class T, class Abi, KernelConfig Conf, index_t RowsReg, StorageOrder OA>
void batmat::linalg::micro_kernels::gemv::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 batmat::linalg::micro_kernels::gemv::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 batmat::linalg::micro_kernels::gemv::RowsReg = 2 * gemm::RowsReg<T, Abi>

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 rotate_B = 0
int rotate_C = 0
int rotate_D = rotate_C
int mask_D = rotate_D