batmat
0.0.16
Batched linear algebra routines
Loading...
Searching...
No Matches
generic.hpp
Go to the documentation of this file.
1
#pragma once
2
3
#include <batmat/config.hpp>
4
5
namespace
batmat
{
6
namespace
linalg::micro_kernels::gemm
{
7
8
/// Register block size of the matrix-matrix multiplication micro-kernels.
9
/// Assumes that the platform has at least 16 vector registers, we use 9 registers for a 3×3
10
/// accumulator block of matrix C (leaving some registers for loading A and B):
11
template
<
class
T,
class
Abi>
12
inline
constexpr
index_t
RowsReg
= 3;
13
14
}
// namespace linalg::micro_kernels::gemm
15
namespace
ops
{
16
17
template
<
class
T>
18
inline
constexpr
index_t
RowsRegTranspose
= 4;
19
template
<
class
T>
20
inline
constexpr
index_t
ColsRegTranspose
= 4;
21
22
}
// namespace ops
23
}
// namespace batmat
guanaqo::linalg::index_t
std::ptrdiff_t index_t
batmat::linalg::micro_kernels::gemm
Definition
gemm.hpp:9
batmat::linalg::micro_kernels::gemm::RowsReg
constexpr index_t RowsReg
Register block size of the matrix-matrix multiplication micro-kernels.
Definition
avx-512.hpp:13
batmat::ops
Definition
cneg.hpp:11
batmat::ops::RowsRegTranspose
constexpr index_t RowsRegTranspose
Definition
avx-512.hpp:23
batmat::ops::ColsRegTranspose
constexpr index_t ColsRegTranspose
Definition
avx-512.hpp:25
batmat
Definition
dtypes.hpp:9
batmat
include
batmat
platform
generic.hpp
Generated on
for batmat by
1.16.1