基本原理

目前，NN类算子覆盖的范围主要是数学基础运算（如加、减、乘、除等）以及CNN相关的一些操作（如卷积、池化、激活函数）等，具体的API可参见NN类算子接口（aclnn）。

NN类算子API的形式一般定义为“两段式接口”，形如：

aclnnStatus aclnnXxxGetWorkspaceSize(const aclTensor *src,..., aclTensor *out, ..., uint64_t workspaceSize, aclOpExecutor **executor);
aclnnStatus aclnnXxx(void* workspace, int64 workspaceSize, aclOpExecutor* executor, aclrtStream stream);

其中aclnnXxxGetWorkspaceSize为第一段接口，主要用于计算本次API调用计算过程中需要多少的workspace内存。获取到本次API计算需要的workspace大小后，按照workspaceSize大小申请Device侧内存，然后调用第二段接口aclnnXxx执行计算。

示例代码

本节以Add算子调用为例，介绍如何编写算子调用的代码逻辑。其他算子的调用逻辑与Add算子大致一样，请根据实际情况自行修改代码。

Add算子实现了张量加法运算，计算公式为：y=x₁+αxx₂。您可以单击aclnnAdd，从“调用示例”模块获取示例代码，并将文件命名为“test_add.cpp”，代码如下：

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/level2/aclnn_add.h"

#define CHECK_RET(cond, return_expr) \
  do {                               \
    if (!(cond)) {                   \
      return_expr;                   \
    }                                \
  } while (0)

#define LOG_PRINT(message, ...)     \
  do {                              \
    printf(message, ##__VA_ARGS__); \
  } while (0)

int64_t GetShapeSize(const std::vector<int64_t>& shape) {
  int64_t shape_size = 1;
  for (auto i : shape) {
    shape_size *= i;
  }
  return shape_size;
}

int Init(int32_t deviceId, aclrtContext* context, aclrtStream* stream) {
  // 固定写法，acl初始化
  auto ret = aclrtSetDevice(deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateContext(context, deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateContext failed. ERROR: %d\n", ret); return ret);
  ret = aclrtSetCurrentContext(*context);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetCurrentContext failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
  ret = aclInit(nullptr);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
  return 0;
}

template <typename T>
int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr,
                    aclDataType dataType, aclTensor** tensor) {
  auto size = GetShapeSize(shape) * sizeof(T);
  // 调用aclrtMalloc申请device侧内存
  auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
  // 调用aclrtMemcpy将host侧数据拷贝到device侧内存上
  ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);
  // 计算连续tensor的strides
  std::vector<int64_t> strides(shape.size(), 1);
  for (int64_t i = shape.size() - 2; i >= 0; i--) {
    strides[i] = shape[i + 1] * strides[i + 1];
  }
  // 调用aclCreateTensor接口创建aclTensor
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. （固定写法）device/context/stream初始化, 参考acl对外接口列表
  // 根据自己的实际device填写deviceId
  int32_t deviceId = 0;
  aclrtContext context;
  aclrtStream stream;
  auto ret = Init(deviceId, &context, &stream);
  // check根据自己的需要处理
  CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. 构造输入与输出，需要根据API的接口自定义构造
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> otherShape = {4, 2};
  std::vector<int64_t> outShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* otherDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* other = nullptr;
  aclScalar* alpha = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
  std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  float alphaValue = 1.2f;
  // 创建self aclTensor
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建other aclTensor
  ret = CreateAclTensor(otherHostData, otherShape, &otherDeviceAddr, aclDataType::ACL_FLOAT, &other);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建alpha aclScalar
  alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);
  CHECK_RET(alpha != nullptr, return ret);
  // 创建out aclTensor
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. 调用CANN算子库API，需要修改为具体的HostApi
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // 调用aclnnAdd第一段接口
  ret = aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // 根据第一段接口计算出的workspaceSize申请device内存
  void* workspaceAddr = nullptr;
  if (workspaceSize > 0) {
    ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret;);
  }
  // 调用aclnnAdd第二段接口
  ret = aclnnAdd(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAdd failed. ERROR: %d\n", ret); return ret);

  // 4.（ 固定写法）同步等待任务执行结束
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. 获取输出的值，将device侧内存上的结果拷贝至host侧，需要根据具体API的接口定义修改
  auto size = GetShapeSize(outShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
                    ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. 释放aclTensor和aclScalar，需要根据具体API的接口定义修改
  aclDestroyTensor(self);
  aclDestroyTensor(other);
  aclDestroyScalar(alpha);
  aclDestroyTensor(out);

  // 7. 释放device资源，需要根据具体API的接口定义修改
  aclrtFree(selfDeviceAddr);
  aclrtFree(otherDeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtDestroyContext(context);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}

CMakeLists文件

本节以Add算子编译脚本为例，介绍如何编写算子编译脚本CMakeLists.txt。其他算子的编译脚本与Add算子大致一样，请根据实际情况自行修改脚本。

# Copyright (c) Huawei Technologies Co., Ltd. 2019. All rights reserved.

# CMake lowest version requirement
cmake_minimum_required(VERSION 3.14)

# project information
project(ACLNN_EXAMPLE)

# Compile options
add_compile_options(-std=c++11)

# 根据实际情况，设置编译选项
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY  "./bin")    
set(CMAKE_CXX_FLAGS_DEBUG "-fPIC -O0 -g -Wall")
set(CMAKE_CXX_FLAGS_RELEASE "-fPIC -O2 -Wall")

# 根据实际情况，设置ASCEND_PATH，即CANN包安装目录
if(NOT "$ENV{ASCEND_CUSTOM_PATH}" STREQUAL "")
    set(ASCEND_PATH $ENV{ASCEND_CUSTOM_PATH})
else()
    set(ASCEND_PATH "/usr/local/Ascend/")
endif()
# 设置INCLUDE_BASE_DIR，即头文件所在目录
set(INCLUDE_BASE_DIR "${ASCEND_PATH}/include")

# (固定写法)Header path
include_directories(
    ${INCLUDE_BASE_DIR}
    ${INCLUDE_BASE_DIR}/aclnn
)

# 根据实际情况，设置可执行文件名（如opapi_test），并指定待运行算子文件*.cpp所在目录
add_executable(opapi_test
               test_add.cpp) 

# (固定写法)linked libraries
target_link_libraries(opapi_test PRIVATE
                      ${ASCEND_PATH}/lib64/libascendcl.so
                      ${ASCEND_PATH}/lib64/libnnopbase.so
                      ${ASCEND_PATH}/lib64/libopapi.so)

# 可执行文件在CMakeLists文件所在目录的bin目录下
install(TARGETS opapi_test DESTINATION ${CMAKE_RUNTIME_OUTPUT_DIRECTORY})

编译与运行

关于编译和运行的详细操作，可参见应用调试章节中“编译及运行应用”内容。

根据前文示例代码和CMakeLists文件，按需定义算子的调用代码（*.cpp）和编译脚本（CMakeLists.txt）。
编译并运行。
1. 在任意目录下以运行用户（默认HwHiAiUser）执行如下命令设置环境变量。其中${INSTALL_DIR}指CANN软件安装后文件存储路径，请根据实际情况替换该路径。
```
source ${INSTALL_DIR}/bin/setenv.bash
```
2. 进入CMakeLists.txt所在目录，执行如下命令，新建build目录存放生成的编译文件。
```
mkdir -p build 
```
3. 进入CMakeLists.txt所在目录，执行cmake命令编译代码，执行make命令生成可执行文件。
```
cmake ./ -DCMAKE_CXX_COMPILER=g++ -DCMAKE_SKIP_RPATH=TRUE
make
```
  编译成功后，会在当前目录的bin目录下生成opapi_test可执行文件。
4. 进入bin目录，运行可执行文件opapi_test。
```
./opapi_test
```
  以Add算子的运行结果为例，结果如下：

调用NN类算子接口示例代码

基本原理

示例代码

CMakeLists文件

编译与运行