---
title: mma
description: "矩阵乘法累加接口，计算A*B+C。"
url: https://www.hikunpeng.com/document/detail/zh/kunpenghpcs/hpckit/devg/KunpengHPCKit_developer_112.html
sourcePath: /source/zh/kunpenghpcs/hpckit/devg/KunpengHPCKit_developer_112.html
indexId: 64f6a816638febbc05ee9b2a1df08e454cedb6e16940ad800712a37a158db44b67
---
# mma

矩阵乘法累加接口，计算A*B+C。

涉及TiledMma策略和Tensor D/A/B/C矩阵输入，其中Tensor D/A/B/C的形状Shape和跨度步长Stride必须与TiledMma策略定义一致。

#### 接口定义

template<typename TiledMma,

typename dtypeD, typename LayoutD,

typename dtypeA, typename LayoutA,

typename dtypeB, typename LayoutB,

typename dtypeC, typename LayoutC>

void mma(TiledMma tiled_mma, Tensor<dtypeD, LayoutD> D, Tensor<dtypeA, LayoutA> A, Tensor<dtypeB, LayoutB> B, Tensor<dtypeC, LayoutC> C);


#### 模板参数


**表1 模板参数定义**

| 参数名 | 类型 | 描述 |
| --- | --- | --- |
| TiledMma | typename | mma tile策略类型。 |
| dtypeD, dtypeA, dtypeB, dtypeC | typename | D,A,B,C精度类型。 |
| LayoutD, LayoutA, LayoutB, LayoutC | typename | D,A,B,C布局类型。 |


#### 参数


**表2 参数定义**

| 参数名 | 类型 | 描述 | 输入/输出 |
| --- | --- | --- | --- |
| tiled\_mma | TiledMma | mma tile策略。 | 输入 |
| D | Tensor<dtypeD, LayoutD> | 矩阵对象D。 | 输入 |
| A | Tensor<dtypeA, LayoutA> | 矩阵对象A。 | 输入 |
| B | Tensor<dtypeB, LayoutB> | 矩阵对象B。 | 输入 |
| C | Tensor<dtypeC, LayoutC> | 矩阵对象C。 | 输入 |


#### 返回值

void


#### 示例

```
#include "stdlib.h"
#include "kupl_mma.h"
using namespace kupl::tensor;
int main()
{
constexpr int MATRIX_M  = 32;
constexpr int MATRIX_N  = 16;
constexpr int MATRIX_K = 512;
double *data_a = (double *)malloc(sizeof(double) * MATRIX_M * MATRIX_K);
double *data_b = (double *)malloc(sizeof(double) * MATRIX_K * MATRIX_N);
double *data_c = (double *)malloc(sizeof(double) * MATRIX_M * MATRIX_N);
auto shape_a = make_shape(Int<32>{}, Int<512>{});
auto shape_b = make_shape(Int<512>{}, Int<16>{});
auto shape_c = make_shape(Int<32>{}, Int<16>{});
auto stride_a = make_stride(Int<1>{}, Int<32>{});
auto stride_b = make_stride(Int<16>{}, Int<1>{});
auto stride_c = make_stride(Int<16>{}, Int<1>{});
auto layout_a = make_layout(shape_a, stride_a);
auto layout_b = make_layout(shape_b, stride_b);
auto layout_c = make_layout(shape_c, stride_c);
auto mma_atom_shape = make_shape(Int<1>{}, Int<1>{}, Int<1>{});
auto tiled_mma = make_tiled_mma(Ops<KP36_32x16x512_F64F64F64>{}, mma_atom_shape);
auto store_atom_shape = make_shape(Int<1>{}, Int<1>{});
auto tiled_store = make_tiled_store(Ops<KP36_32x16_F64_STORE>{}, store_atom_shape);
auto tensor_a = make_tensor(data_a, layout_a);
auto tensor_b = make_tensor(data_b, layout_b);
auto tensor_c = make_tensor(data_c, layout_c);
mma
(tiled_mma, tensor_c, tensor_a, tensor_b, tensor_c);
store(tiled_store, tensor_c);
free(data_a);
free(data_b);
free(data_c);
return 0;
}
```

上述示例演示了基于32*16*512_F64F64F64矩阵形状的mma流程，其中通过mma接收Tensor对象进行矩阵乘法计算。