aclnnStdMeanCorrection
接口原型
每个算子有两段接口，必须先调用“aclnnXxxGetWorkspaceSize”接口获取入参并根据计算流程计算所需workspace大小，再调用“aclnnXxx”接口执行计算。两段式接口如下：
- 第一段接口：aclnnStatus aclnnStdMeanCorrectionGetWorkspaceSize(const aclTensor *self, const aclIntArray *dim, int64_t correction, bool keepdim, aclTensor *stdOut, aclTensor *meanOut, uint64_t *workspaceSize, aclOpExecutor **executor)
- 第二段接口：aclnnStatus aclnnStdMeanCorrection(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)
功能描述
aclnnStdMeanCorrectionGetWorkspaceSize
- 接口定义：
aclnnStatus aclnnStdMeanCorrectionGetWorkspaceSize(const aclTensor *self, const aclIntArray *dim, int64_t correction, bool keepdim, aclTensor *stdOut, aclTensor *meanOut, uint64_t *workspaceSize, aclOpExecutor **executor)
- 参数说明：
- self(aclTensor*, 计算输入)：Device侧的aclTensor，数据类型支持FLOAT、FLOAT16。支持非连续的Tensor，数据格式支持ND。
- dim(aclIntArray*, 计算输入)：Host侧的aclIntArray，指定计算的维度，数据类型支持INT32、INT64，取值范围为[-self.dim(), self.dim()-1]，且其中元素值不能相同。
- correction(int64_t, 计算输入)：Host侧的整型，公式中的δN（修正值），数据类型支持INT64。
- keepdim(bool, 计算输入)：Host侧的BOOL类型，是否在输出张量中保留输入张量的维度。若为false，不保留该维度。
- stdOut(aclTensor*, 计算输出)：Device侧的aclTensor，标准差的计算结果，数据类型支持FLOAT、FLOAT16。支持非连续的Tensor，数据格式支持ND。
- meanOut(aclTensor*, 计算输出)：Device侧的aclTensor，均值的计算结果，数据类型支持FLOAT、FLOAT16。支持非连续的Tensor，数据格式支持ND。
- workspaceSize：返回用户需要在Device侧申请的workspace大小。
- executor：返回op执行器，包含了算子计算流程。
- 返回值：
返回aclnnStatus状态码，具体参见aclnn返回码。
第一段接口完成入参校验，出现以下场景时报错：
- 返回161001（ACLNN_ERR_PARAM_NULLPTR）：传入的self、stdOut、meanOut是空指针。
- 返回161002（ACLNN_ERR_PARAM_INVALID）：
- self、stdOut、meanOut数据类型不在支持的范围之内。
- dim数组的维度超出self的维度范围。
- dim数组中元素重复。
aclnnStdMeanCorrection
- 接口定义：
aclnnStatus aclnnStdMeanCorrection(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)
- 参数说明：
- workspace：在Device侧申请的workspace内存起址。
- workspaceSize：在Device侧申请的workspace大小，由第一段接口aclnnStdMeanCorrectionGetWorkspaceSize获取。
- executor：op执行器，包含了算子计算流程。
- stream：指定执行任务的AscendCL stream流。
- 返回值：
返回aclnnStatus状态码，具体参见aclnn返回码。
调用示例
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_std_mean_correction.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，AscendCL初始化
auto ret = aclInit(nullptr);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
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);
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AscendCL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 = {2, 3, 4};
std::vector<int64_t> stdOutShape = {2, 4};
std::vector<int64_t> meanOutShape = {2, 4};
void* selfDeviceAddr = nullptr;
void* stdOutDeviceAddr = nullptr;
void* meanOutDeviceAddr = nullptr;
aclTensor* self = nullptr;
aclTensor* stdOut = nullptr;
aclTensor* meanOut = nullptr;
std::vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
19, 20, 21, 22, 23, 24};
std::vector<float> stdOutHostData = {1, 2, 3, 4, 5, 6, 7, 8.0};
std::vector<float> meanOutHostData = {1, 2, 3, 4, 5, 6, 7, 8.0};
std::vector<int64_t> dimData = {1};
int64_t correction = 1;
bool keepdim = false;
// 创建self aclTensor
ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建stdOut aclTensor
ret = CreateAclTensor(stdOutHostData, stdOutShape, &stdOutDeviceAddr, aclDataType::ACL_FLOAT, &stdOut);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建meanOut aclTensor
ret = CreateAclTensor(meanOutHostData, meanOutShape, &meanOutDeviceAddr, aclDataType::ACL_FLOAT, &meanOut);
CHECK_RET(ret == ACL_SUCCESS, return ret);
const aclIntArray *dim = aclCreateIntArray(dimData.data(), dimData.size());
CHECK_RET(dim != nullptr, return ACL_ERROR_INTERNAL_ERROR);
// 3. 调用CANN算子库API，需要修改为具体的算子接口
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
// 调用aclnnStdMeanCorrection第一段接口
ret = aclnnStdMeanCorrectionGetWorkspaceSize(self, dim, correction, keepdim, stdOut, meanOut, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnStdMeanCorrectionGetWorkspaceSize 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aclnnStdMeanCorrection第二段接口
ret = aclnnStdMeanCorrection(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnStdMeanCorrection 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(stdOutShape);
std::vector<float> stdResultData(size, 0);
ret = aclrtMemcpy(stdResultData.data(), stdResultData.size() * sizeof(stdResultData[0]), stdOutDeviceAddr, 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("stdResultData[%ld] is: %f\n", i, stdResultData[i]);
}
std::vector<float> meanResultData(size, 0);
ret = aclrtMemcpy(meanResultData.data(), meanResultData.size() * sizeof(meanResultData[0]), meanOutDeviceAddr, 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("meanResultData[%ld] is: %f\n", i, meanResultData[i]);
}
// 6. 释放aclTensor和aclScalar，需要根据具体API的接口定义修改
aclDestroyTensor(self);
aclDestroyTensor(stdOut);
aclDestroyTensor(meanOut);
// 7. 释放device资源，需要根据具体API的接口定义修改
aclrtFree(selfDeviceAddr);
aclrtFree(stdOutDeviceAddr);
aclrtFree(meanOutDeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtDestroyContext(context);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}
父主题： NN类算子接口