aclnnSplitTensor
支持的产品型号
- Atlas 训练系列产品。
- Atlas A2训练系列产品/Atlas 800I A2推理产品。
- Atlas 推理系列产品。
接口原型
每个算子分为两段式接口,必须先调用“aclnnSplitTensorGetWorkspaceSize”接口获取计算所需workspace大小以及包含了算子计算流程的执行器,再调用“aclnnSplitTensor”接口执行计算。
aclnnStatus aclnnSplitTensorGetWorkspaceSize(const aclTensor* self, uint64_t splitSections, int64_t dim, aclTensorList* out, uint64_t* workspaceSize, aclOpExecutor** executor)
aclnnStatus aclnnSplitTensor(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
功能描述
算子功能:将输入self沿dim轴按照splitSections大小均匀切分。若dim轴无法被整除,则非最后一块的大小等于splitSections,最后一块小于splitSections。
aclnnSplitTensorGetWorkspaceSize
参数说明:
- self(aclTensor*,计算输入):表示被split的输入tensor,数据类型支持FLOAT、FLOAT16、DOUBLE、BFLOAT16(仅Atlas A2训练系列产品/Atlas 800I A2推理产品支持)、INT32、INT64、INT16、INT8、UINT8、BOOL、COMPLEX128和COMPLEX64。支持非连续的Tensor,数据格式支持ND。
- splitSections(uint64_t,计算输入):数据类型支持UINT64,表示沿dim轴均匀切分后的块大小, 最后一块可以小于splitSections。
- dim(int64_t,计算输入):数据类型支持INT64,表示输入tensor被split的维度。
- out(aclTensorList*,计算输出):表示被split后的输出tensor的列表,数据类型支持FLOAT、FLOAT16、DOUBLE、BFLOAT16(仅Atlas A2训练系列产品/Atlas 800I A2推理产品支持)、INT32、INT64、INT16、INT8、UINT8、BOOL、COMPLEX128和COMPLEX64。支持非连续的Tensor,数据格式支持ND,每个输出的dtype需要保持一致。
- 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。
返回值:
aclnnStatus:返回状态码,具体参见aclnn返回码。
161001 (ACLNN_ERR_PARAM_NULLPTR): 1. 传入的self、out是空指针。
161002 (ACLNN_ERR_PARAM_INVALID): 1. self和out的数据类型不在支持的范围之内。
2. self的长度不在支持的范围之内。
3. out中的tensor长度不在支持的范围之内。
4. dim的取值越界。
5. 被split的维度shape为0且splitSections不为0。
6. 被split的维度shape不为0且splitSections为0。
aclnnSplitTensor
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aclnnSplitTensorGetWorkspaceSize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
无。
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 <chrono>
#include <algorithm>
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_split_tensor.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,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 = 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;
}
void CheckResult(const std::vector<std::vector<int64_t>> &shapeList, const std::vector<void *> addrList) {
for (size_t i = 0; i < shapeList.size(); i++) {
auto size = GetShapeSize(shapeList[i]);
std::vector<float> resultData(size, 0);
auto ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), addrList[i],
size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return);
for (int64_t j = 0; j < size; j++) {
LOG_PRINT("result[%ld] is: %f\n", j, resultData[j]);
}
}
}
int main() {
// 1.(固定写法)device/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;
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> selfShape = {4, 2};
std::vector<int64_t> shape1 = {2, 2};
std::vector<int64_t> shape2 = {2, 2};
uint64_t splitSections = 2;
int64_t dim = 0;
void* selfDeviceAddr = nullptr;
void* shape1DeviceAddr = nullptr;
void* shape2DeviceAddr = nullptr;
aclTensor* self = nullptr;
aclTensor* shape1Addr = nullptr;
aclTensor* shape2Addr = nullptr;
std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
std::vector<float> shape1HostData = {0, 1, 4, 5};
std::vector<float> shape2HostData = {2, 3, 6, 7};
// 创建self aclTensor
ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
CHECK_RET(ret == ACL_SUCCESS, return ret);
ret = CreateAclTensor(shape1HostData, shape1, &shape1DeviceAddr, aclDataType::ACL_FLOAT, &shape1Addr);
CHECK_RET(ret == ACL_SUCCESS, return ret);
ret = CreateAclTensor(shape2HostData, shape2, &shape2DeviceAddr, aclDataType::ACL_FLOAT, &shape2Addr);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建out aclTensorList
std::vector<aclTensor*> tmp = {shape1Addr, shape2Addr};
aclTensorList* out = aclCreateTensorList(tmp.data(), tmp.size());
CHECK_RET(out != nullptr, return ret);
// 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aclnnSplitTensor第一段接口
ret = aclnnSplitTensorGetWorkspaceSize(self, splitSections, dim, out, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSplitTensorGetWorkspaceSize 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) {
auto 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aclnnSplitTensor第二段接口
ret = aclnnSplitTensor(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSplitTensor failed. ERROR: %d\n", ret); return ret);
ret = aclrtSynchronizeStream(stream);
CheckResult({shape1, shape2}, {shape1DeviceAddr, shape2DeviceAddr});
// 6. 释放aclTensor和aclScalar,需要根据具体API的接口定义修改
aclDestroyTensor(self);
aclDestroyTensorList(out);
// 7. 释放device 资源
aclrtFree(selfDeviceAddr);
aclrtFree(shape1DeviceAddr);
aclrtFree(shape2DeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}