aclnnForeachCopy
支持的产品型号
- Atlas A2训练系列产品/Atlas 800I A2推理产品。
接口原型
每个算子分为两段式接口,必须先调用“aclnnForeachCopyGetWorkspaceSize”接口获取入参并根据计算流程计算所需workspace大小,再调用“aclnnForeachCopy”接口执行计算。
aclnnStatus aclnnForeachCopyGetWorkspaceSize(const aclTensorList *x, aclTensorList *out, uint64_t *workspaceSize, aclOpExecutor **executor)
aclnnStatus aclnnForeachCopy(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)
功能描述
算子功能:用于实现两个张量列表内容的复制,要求输入和输出两个张量列表形状相同。
计算公式:
aclnnForeachCopyGetWorkspaceSize
参数说明:
- src(aclTensorList*,计算输入):公式中的
src
,Device侧的aclTensorList,数据类型支持FLOAT、FLOAT16、BFLOAT16、INT32。数据格式支持ND,shape维度不高于8维。 - dst(aclTensorList*,计算输出):公式中的
dst
,Device侧的aclTensorList,数据类型支持FLOAT、FLOAT16、BFLOAT16、INT32。数据格式支持ND,shape维度不高于8维。数据格式和shape跟入参x
一致。 - workspaceSize(uint64_t*,出参):返回需要在Device侧申请的workspace大小。
- executor(aclOpExecutor**,出参):返回op执行器,包含了算子计算流程。
- src(aclTensorList*,计算输入):公式中的
返回值:
aclnnStatus:返回状态码,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): 1. 传入的src或dst是空指针。 返回161002(ACLNN_ERR_PARAM_INVALID): 1. src或dst的数据类型不在支持的范围之内。 2. src、dst的shape不一致。 3. self维度大于8。
aclnnForeachCopy
参数说明:
- workspace(void*,入参):在Device侧申请的workspace内存地址。
- workspaceSize(uint64_t,入参):在Device侧申请的workspace大小,由第一段接口aclnnForeachCopyGetWorkspaceSize获取。
- executor(aclOpExecutor*,入参):op执行器,包含了算子计算流程。
- stream(aclrtStream,入参):指定执行任务的AscendCL Stream流。
返回值:
aclnnStatus:返回状态码,具体参见aclnn返回码。
约束与限制
- dtype仅支持FLOAT32、FLOAT16、BFLOAT16和INT32。
调用示例
示例代码如下,仅供参考,具体编译和执行过程请参考编译与运行样例。
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_foreach_copy.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 shapeSize = 1;
for (auto i : shape) {
shapeSize *= i;
}
return shapeSize;
}
int Init(int32_t deviceId, aclrtStream *stream)
{
// 固定写法,acl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 = 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侧数据复制到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/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;
aclrtStream stream;
auto ret = Init(deviceId, &stream);
CHECK_RET(ret == ACL_SUCCESS, 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> selfShape1 = {2, 3};
std::vector<int64_t> selfShape2 = {1, 3};
std::vector<int64_t> outShape1 = {2, 3};
std::vector<int64_t> outShape2 = {1, 3};
void *input1DeviceAddr = nullptr;
void *input2DeviceAddr = nullptr;
void *out1DeviceAddr = nullptr;
void *out2DeviceAddr = nullptr;
aclTensor *input1 = nullptr;
aclTensor *input2 = nullptr;
aclTensor *out1 = nullptr;
aclTensor *out2 = nullptr;
std::vector<float> input1HostData = {1, 2, 3, 4, 5, 6};
std::vector<float> input2HostData = {7, 8, 9};
std::vector<float> out1HostData(6, 0);
std::vector<float> out2HostData(3, 0);
// 创建input1 aclTensor
ret = CreateAclTensor(input1HostData, selfShape1, &input1DeviceAddr, aclDataType::ACL_FLOAT, &input1);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建input2 aclTensor
ret = CreateAclTensor(input2HostData, selfShape2, &input2DeviceAddr, aclDataType::ACL_FLOAT, &input2);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建out1 aclTensor
ret = CreateAclTensor(out1HostData, outShape1, &out1DeviceAddr, aclDataType::ACL_FLOAT, &out1);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建out2 aclTensor
ret = CreateAclTensor(out2HostData, outShape2, &out2DeviceAddr, aclDataType::ACL_FLOAT, &out2);
CHECK_RET(ret == ACL_SUCCESS, return ret);
std::vector<aclTensor *> tempInput{input1, input2};
aclTensorList *tensorListInput = aclCreateTensorList(tempInput.data(), tempInput.size());
std::vector<aclTensor *> tempOutput{out1, out2};
aclTensorList *tensorListOutput = aclCreateTensorList(tempOutput.data(), tempOutput.size());
// 3. 调用CANN算子库API,需要修改为具体的Api名称
uint64_t workspaceSize = 0;
aclOpExecutor *executor;
// 调用aclnnForeachCopy第一段接口
ret = aclnnForeachCopyGetWorkspaceSize(tensorListInput, tensorListOutput, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnForeachCopyGetWorkspaceSize 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aclnnForeachCopy第二段接口
ret = aclnnForeachCopy(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnForeachCopy 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侧内存上的结果复制至host侧,需要根据具体API的接口定义修改
auto size = GetShapeSize(outShape1);
std::vector<float> out1Data(size, 0);
ret = aclrtMemcpy(out1Data.data(),
out1Data.size() * sizeof(out1Data[0]),
out1DeviceAddr,
size * sizeof(out1Data[0]),
ACL_MEMCPY_DEVICE_TO_HOST);
for (int64_t i = 0; i < size; i++) {
LOG_PRINT("out1 result[%ld] is: %f\n", i, out1Data[i]);
}
size = GetShapeSize(outShape2);
std::vector<float> out2Data(size, 0);
ret = aclrtMemcpy(out2Data.data(),
out2Data.size() * sizeof(out2Data[0]),
out2DeviceAddr,
size * sizeof(out2Data[0]),
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("out2 result[%ld] is: %f\n", i, out2Data[i]);
}
// 6. 释放aclTensor和aclScalar,需要根据具体API的接口定义修改
aclDestroyTensorList(tensorListInput);
aclDestroyTensorList(tensorListOutput);
// 7.释放device资源,需要根据具体API的接口定义修改
aclrtFree(input1DeviceAddr);
aclrtFree(input2DeviceAddr);
aclrtFree(out1DeviceAddr);
aclrtFree(out2DeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}