aclnnScatter&aclnnInplaceScatter
Applicable Products
| Product | Supported |
|---|---|
| √ | |
| √ | |
| √ | |
| × | |
| × | |
| × |
Function
Description: Replaces/accumulates/multiplies values from the tensor src into the tensor self element by element according to the specified axis, direction, and corresponding position relationship.
Example:
For a 3D tensor, self is updated according to the following rules:
self[index[i][j][k]][j][k] += src[i][j][k] # If dim == 0 && reduction == 1 self[i][index[i][j][k]][k] *= src[i][j][k] # If dim == 1 && reduction == 2 self[i][j][index[i][j][k]] = src[i][j][k] # If dim == 2 && reduction == 0The following requirements must be met during computation:
self, index, and src must have the same number of dimensions.
For each dimension d, index.size(d) must be less than or equal to src.size(d).
For each dimension d, if d != dim, index.size(d) must be less than or equal to self.size(d).
The value of dim must be in the range [-Number of Dimensions of self, Number of Dimensions of self - 1].
The number of dimensions of self must be less than or equal to 8.
The values in index corresponding to dimension dim must be in the range [0, self.size(dim)-1].
Function Prototype
aclnnScatter and aclnnInplaceScatter implement the same functionality. The differences are as follows. Select the appropriate operator based on your actual scenario.
aclnnScatter: Requires creating a new output tensor object to store the computation result.
aclnnInplaceScatter: Does not require creating a new output tensor object. The computation result is stored directly in the memory of the input tensor.
Each operator is a two-phase API. You must first call the "aclnnScatterGetWorkspaceSize" or "aclnnInplaceScatterGetWorkspaceSize" API to obtain the workspace size required for computation and the executor that encapsulates the operator computation process, and then call the "aclnnScatter" or "aclnnInplaceScatter" API to perform the computation.
aclnnStatus aclnnScatterGetWorkspaceSize(
const aclTensor* self,
int64_t dim,
const aclTensor* index,
const aclTensor* src,
int64_t reduce,
aclTensor* out,
uint64_t* workspaceSize,
aclOpExecutor** executor)aclnnStatus aclnnScatter(
void* workspace,
uint64_t workspaceSize,
aclOpExecutor* executor,
const aclrtStream stream)aclnnStatus aclnnInplaceScatterGetWorkspaceSize(
aclTensor* selfRef,
int64_t dim,
const aclTensor* index,
const aclTensor* src,
int64_t reduce,
uint64_t* workspaceSize,
aclOpExecutor** executor)aclnnStatus aclnnInplaceScatter(
void* workspace,
uint64_t workspaceSize,
aclOpExecutor* executor,
aclrtStream stream)aclnnScatterGetWorkspaceSize
Parameters:
Parameter Input/Output Description Instruction Data Type Data Format Dimension (Shape) Non-Contiguous Tensor self (aclTensor*) Input `self` in the formula. The data type must be the same as src.
The number of dimensions must be the same as index and src.UINT8, INT8, INT16, INT32, INT64, BOOL, FLOAT16, FLOAT32, DOUBLE, COMPLEX64, COMPLEX128, BFLOAT16 ND 0-8 √ dim (int64_t) Input Dimension along which to scatter. The range is [-Number of Dimensions of self, Number of Dimensions of self - 1]. - - - - index (aclTensor*) Input Index tensor. The number of dimensions must be the same as self and src. INT32, INT64 ND 0-8 √ src (aclTensor*) Input `src` in the formula. The data type must be the same as that of self.
The number of dimensions must be the same as that of self and index.UINT8, INT8, INT16, INT32, INT64, BOOL, FLOAT16, FLOAT32, DOUBLE, COMPLEX64, COMPLEX128, BFLOAT16 ND 0-8 √ reduce (int64_t) Input Selects the reduction operation to apply. Available operation options and their corresponding int values are (add, 1), (mul, 2), (none, 0).
The specific meanings of the operations are as follows: 0 indicates replacement, 1 indicates accumulation, and 2 indicates multiplication.- - - - out (aclTensor*) Output Calculation result. The data type must be consistent with self.
The shape must be consistent with self.UINT8, INT8, INT16, INT32, INT64, BOOL, FLOAT16, FLOAT32, DOUBLE, COMPLEX64, COMPLEX128, BFLOAT16 ND 0-8 √ workspaceSize (uint64_t*) Output Returns the workspace size to be allocated on the Device side. - - - - - executor（aclOpExecutor**） Output Returns the op executor, which contains the operator computation flow. - - - - - Return value
aclnnStatus: return code. For details, see aclnn Return Code.
The first-phase API performs input parameter validation and returns an error in the following scenarios:
Return Value Return Code Description ACLNN_ERR_PARAM_NULLPTR 161001 self, index, src, or out is a null pointer. ACLNN_ERR_PARAM_INVALID 161002 The data type of self, index, src, or out is not within the supported range. The data types of self, src, and out are different. The number of dimensions of self, index, and src are inconsistent. The shapes of self and out are inconsistent. The value of dim is not within the range of [-Number of Dimensions of self, Number of Dimensions of self - 1]. The number of dimensions of self exceeds 8. The shapes of self, index, and src do not meet the constraints: for each dimension d, index.size(d) <= src.size(d); and when d != dim, index.size(d) <= self.size(d).
aclnnScatter
Parameters:
Parameter Input/Output Description workspace Input Address of the workspace memory applied for on the Device side. workspaceSize Input Size of the workspace applied for on the Device side, obtained from the first-phase API aclnnScatterGetWorkspaceSize. Executor Input Operator executor, which contains the operator computation flow. Stream Input Specifies the stream for task execution. Return value
aclnnStatus: return code. For details, see aclnn Return Code.
aclnnInplaceScatterGetWorkspaceSize
Parameters:
Parameter Input/Output Description Instruction Data Type Data Format Shape Non-contiguous Tensor selfRef (aclTensor*) Input/Output The target tensor for scatter. The shape must have the same number of dimensions as index and src.
The data type is the same as that of src.
Supports empty tensors.UINT8, INT8, INT16, INT32, INT64, BOOL, FLOAT16, FLOAT32, DOUBLE, COMPLEX64, COMPLEX128, BFLOAT16 ND 0-8 √ dim (int64_t) Input Specifies the dimension along which to perform the scatter operation. The range is [-Number of Dimensions of selfRef, Number of Dimensions of selfRef - 1]. - - - - index (aclTensor*) Input Index tensor. Specifies the positions in the self tensor where the src tensor elements are scattered.
Supports empty tensors.INT32, INT64 ND 0-8 √ src (aclTensor*) Input Source tensor. The data type must be the same as selfRef.
Empty tensor is supported.UINT8, INT8, INT16, INT32, INT64, BOOL, FLOAT16, FLOAT32, DOUBLE, COMPLEX64, COMPLEX128, BFLOAT16 ND 0-8 √ reduce (int64_t) Input Selects the reduction operation to apply. Available operations: 0 for replace, 1 for add, 2 for multiply. - - - - workspaceSize (uint64_t*) Output Returns the workspace size to be allocated on the Device. - - - - - executor（aclOpExecutor**） Output Returns the op executor, which contains the operator computation flow. - - - - - Return value
aclnnStatus: return code. For details, see aclnn Return Code.
The first-phase API performs input parameter validation and returns an error in the following scenarios:
Return Value Error Code Description ACLNN_ERR_PARAM_NULLPTR 161001 selfRef, index, or src is a null pointer. ACLNN_ERR_PARAM_INVALID 161002 The data type of selfRef, index, or src is not within the supported range. The data types of selfRef and src are different. The Number of Dimensions of selfRef, index, and src are inconsistent. The value of dim is not within the range of [-Number of Dimensions of selfRef, Number of Dimensions of selfRef - 1]. The Number of Dimensions of selfRef exceeds 8. The shape does not meet the constraints: for each dimension d, index.size(d) <= src.size(d); and when d != dim, index.size(d) <= selfRef.size(d).
aclnnInplaceScatter
Parameters:
Parameter Input/Output Description workspace Input Workspace memory address applied for on the Device side. workspaceSize Input Workspace size applied for on the Device side, obtained from the first-phase API aclnnInplaceScatterGetWorkspaceSize. executor Input Operator executor, which contains the operator computation process. stream Input Specifies the stream for task execution. Return value
aclnnStatus: return code. For details, see aclnn Return Code.
Constraints
Deterministic computation:
- aclnnScatter&aclnnInplaceScatter default to a deterministic implementation.
Example
The following provides example code for reference only. For details about compilation and running, see Compile and Run Samples.
aclnnScatter Sample Code:
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_scatter.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) {
// Boilerplate for resource initialization.
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);
// Call aclrtMalloc to allocate device memory.
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);
// Call aclrtMemcpy to copy data from the host to the device.
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);
// Compute the strides of a contiguous tensor.
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];
}
// Call aclCreateTensor to create an 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. (Boilerplate) Initialize the device/stream. See the ACL API manual.
// Fill in the deviceId based on your actual device.
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. Construct the input and output. Customize the construction based on the API.
int64_t dim = 1;
int64_t reduce = 1;
std::vector<int64_t> selfShape = {3, 4};
std::vector<int64_t> indexShape = {2, 3};
std::vector<int64_t> srcShape = {2, 3};
std::vector<int64_t> outShape = {3, 4};
void* selfDeviceAddr = nullptr;
void* indexDeviceAddr = nullptr;
void* srcDeviceAddr = nullptr;
void* outDeviceAddr = nullptr;
aclTensor* self = nullptr;
aclTensor* index = nullptr;
aclTensor* src = nullptr;
aclTensor* out = nullptr;
std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
std::vector<int64_t> indexHostData = {0, 0, 2, 1, 0, 2};
std::vector<float> srcHostData = {-1, -2, -3, -4, -5, -6};
std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
// Create the self aclTensor.
ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// Create the index aclTensor.
ret = CreateAclTensor(indexHostData, indexShape, &indexDeviceAddr, aclDataType::ACL_INT64, &index);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// Create the src aclTensor.
ret = CreateAclTensor(srcHostData, srcShape, &srcDeviceAddr, aclDataType::ACL_FLOAT, &src);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// Create the out aclTensor.
ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 3. Call the CANN operator library API. Replace with the actual API name.
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
// Call the first-phase API of aclnnScatter.
ret = aclnnScatterGetWorkspaceSize(self, dim, index, src, reduce, out, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnScatterGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
// Apply for device memory based on the workspaceSize computed by the first-phase API.
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);
}
// Call the second-phase API of aclnnScatter.
ret = aclnnScatter(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnScatter failed. ERROR: %d\n", ret); return ret);
// 4. (Boilerplate) Wait for task execution to finish synchronously.
ret = aclrtSynchronizeStream(stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);
// 5. Obtain the output value and copy the result from device memory to the host side. Modify based on the specific API definition.
auto size = GetShapeSize(outShape);
std::vector<float> resultData(size, 0);
ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,
size * sizeof(resultData[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("result[%ld] is: %f\n", i, resultData[i]);
}
// 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
aclDestroyTensor(self);
aclDestroyTensor(index);
aclDestroyTensor(src);
aclDestroyTensor(out);
// 7. Release device resources. Modify based on the specific API definition.
aclrtFree(selfDeviceAddr);
aclrtFree(indexDeviceAddr);
aclrtFree(srcDeviceAddr);
aclrtFree(outDeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}aclnnInplaceScatter sample code:
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_scatter.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) {
// Boilerplate: resource initialization.
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);
// Call aclrtMalloc to apply for device-side memory.
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);
// Call aclrtMemcpy to copy host-side data to device-side memory.
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);
// Calculate the strides of a contiguous tensor.
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];
}
// Call aclCreateTensor to create an 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. (Boilerplate) Initialize the device/stream. Refer to the ACL API manual.
// Fill in the deviceId based on your actual device.
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. Construct the input and output. Custom construction is required based on the API.
int64_t dim = 1;
int64_t reduce = 1;
std::vector<int64_t> selfRefShape = {3, 4};
std::vector<int64_t> indexShape = {2, 3};
std::vector<int64_t> srcShape = {2, 3};
void* selfRefDeviceAddr = nullptr;
void* indexDeviceAddr = nullptr;
void* srcDeviceAddr = nullptr;
aclTensor* selfRef = nullptr;
aclTensor* index = nullptr;
aclTensor* src = nullptr;
std::vector<float> selfRefHostData = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
std::vector<int64_t> indexHostData = {0, 0, 2, 1, 0, 2};
std::vector<float> srcHostData = {-1, -2, -3, -4, -5, -6};
// Create the selfRef aclTensor.
ret = CreateAclTensor(selfRefHostData, selfRefShape, &selfRefDeviceAddr, aclDataType::ACL_FLOAT, &selfRef);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// Create the index aclTensor.
ret = CreateAclTensor(indexHostData, indexShape, &indexDeviceAddr, aclDataType::ACL_INT64, &index);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// Create the src aclTensor.
ret = CreateAclTensor(srcHostData, srcShape, &srcDeviceAddr, aclDataType::ACL_FLOAT, &src);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 3. Call the CANN operator library API. Replace with the actual API name.
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
// Call the first-phase API of aclnnInplaceScatter.
ret = aclnnInplaceScatterGetWorkspaceSize(selfRef, dim, index, src, reduce, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceScatterGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
// Apply for device memory based on the workspaceSize computed by the first-phase API.
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);
}
// Call the second-phase API of aclnnInplaceScatter.
ret = aclnnInplaceScatter(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceScatter failed. ERROR: %d\n", ret); return ret);
// 4. (Boilerplate) Synchronously wait for task execution to complete.
ret = aclrtSynchronizeStream(stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);
// 5. Obtain the output value and copy the result from the device memory to the host memory. Modify based on the specific API definition.
auto size = GetShapeSize(selfRefShape);
std::vector<float> resultData(size, 0);
ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), selfRefDeviceAddr,
size * sizeof(resultData[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("result[%ld] is: %f\n", i, resultData[i]);
}
// 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
aclDestroyTensor(selfRef);
aclDestroyTensor(index);
aclDestroyTensor(src);
// 7. Release device resources. Modify based on the specific API definition.
aclrtFree(selfRefDeviceAddr);
aclrtFree(indexDeviceAddr);
aclrtFree(srcDeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}