aclnnTriangularSolve

Applicable Products

Product	Supported
Ascend 950PR/Ascend 950DT	√
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	×
Atlas inference products	×
Atlas training products	√

Function

Description: Solves a system of equations with a square upper or lower triangular invertible matrix A and multiple right-hand sides b.
Formula:
$AX = b$
Where $A$ is an upper triangular square matrix (or a lower triangular square matrix when upper is false), with no zero elements on its main diagonal. $b$ and $A$ are two-dimensional matrices or batches of two-dimensional matrices. When the input is a batch, the returned output $X$ is also the corresponding batch. When the main diagonal of $A$ contains zeros, or elements very close to zero, and unitriangular is false, the output may contain $NaN$ .

Function Prototype

Each operator is divided into a two-phase API. The aclnnTriangularSolveGetWorkspaceSize API must be called first to obtain the workspace size required for computation and the executor that encapsulates the operator computation flow. Then, the aclnnTriangularSolve API is called to perform the computation.

aclnnStatus aclnnTriangularSolveGetWorkspaceSize(
  const aclTensor* self,
  const aclTensor* A,
  bool             upper,
  bool             transpose,
  bool             unitriangular,
  aclTensor*       xOut,
  aclTensor*       mOut,
  uint64_t*        workspaceSize,
  aclOpExecutor**  executor)

aclnnStatus aclnnTriangularSolve(
  void*            workspace,
  uint64_t         workspaceSize,
  aclOpExecutor*   executor,
  aclrtStream      stream)

aclnnTriangularSolveGetWorkspaceSize

Parameters

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
self (aclTensor*)	Input	b in the formula, the right-hand side of the equation.	The data type is consistent with A. self[-2] = A[-2]. Except for the last two dimensions, the remaining dimensions of A and self satisfy the broadcast relationship.	FLOAT, DOUBLE, COMPLEX64, COMPLEX128	ND	2-8	√
A (aclTensor*)	Input	A in the formula, the coefficient matrix.	The data type is consistent with self. The last two axes are equal. Except for the last two dimensions, the remaining dimensions of A and self satisfy the broadcast relationship.	FLOAT, DOUBLE, COMPLEX64, COMPLEX128	ND	2-8	√
upper (bool)	Input	Controls the computation property of A in the formula, specifying whether it participates in computation as an upper or lower triangular matrix.	Defaults to true, meaning A is an upper triangular matrix. When upper is false, A is a lower triangular matrix.	BOOL	-	-	-
transpose (bool)	Input	Computation attribute that controls whether A or A^T is used in the control formula.	Defaults to false. When transpose is true, A^TX=b is computed.	BOOL	-	-	-
unitriangular (bool)	Input	A computation attribute that controls whether A in the formula is treated as a unit triangular matrix.	Defaults to false. When unitriangular is true, the main diagonal elements of A are treated as 1 instead of being referenced from A. When unitriangular is true, the input self and A, and the output xOut and mOut support only the FLOAT data type.	BOOL	-	-	-
xOut (aclTensor*)	Output	X in the formula, the solution of the equation.	The data type is consistent with self. The shape must satisfy the AX=b constraint with A and b after broadcast. After A and self satisfy the broadcast relation, the last axis dim=self[-1].	FLOAT, DOUBLE, COMPLEX64, COMPLEX128	ND	-	√
mOut (aclTensor*)	Output	Copies the upper triangle (or lower triangle) of A after broadcast.	The data type is consistent with self. After A and self satisfy the broadcast relationship, the last axis of the resulting dimension is dim = A[-1].	FLOAT, DOUBLE, COMPLEX64, COMPLEX128	ND	-	√
workspaceSize (uint64_t*)	Output	Returns the workspace size to be allocated on the Device side.	-	-	-	-	-
executor (aclOpExecutor**)	Output	Returns the operator executor, which includes the operator computation flow.	-	-	-	-	-

Return Value

aclnnStatus: return code. For details, see aclnn Return Codes.

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	A null pointer exists in the input self, A, xOut, or mOut.
ACLNN_ERR_PARAM_INVALID	161002	The data type or data format of self, A, xOut, or mOut is not within the supported range.
ACLNN_ERR_PARAM_INVALID	161002	The shapes of self, A, xOut, and mOut do not satisfy the constraints.

aclnnTriangularSolve

Parameters

Parameter	Input/Output	Description
workspace	Input	Address of the workspace memory applied for on the device side.
workspaceSize	Input	Workspace size obtained by the first-phase API `aclnnTriangularSolveGetWorkspaceSize`.
executor	Input	Operator executor, which includes the operator computation flow.
stream	Input	Specifies the stream for task execution.

Return Value

aclnnStatus: return code. For details, see aclnn Return Codes.

Constraints

Deterministic computation description: aclnnTriangularSolve defaults to a deterministic implementation.

Example

The following provides example code for reference only. For details about compilation and running, see Compile and Run Samples.

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_triangular_solve.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 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 host data to the device 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);

  // 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 the aclCreateTensor API 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.
  // Set 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.
  std::vector<int64_t> selfShape = {3, 1};
  std::vector<int64_t> otherShape = {3, 3};
  std::vector<int64_t> xOutShape = {3, 1};
  std::vector<int64_t> mOutShape = {3, 3};
  void* selfDeviceAddr = nullptr;
  void* otherDeviceAddr = nullptr;
  void* xOutDeviceAddr = nullptr;
  void* mOutDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* other = nullptr;
  aclTensor* xOut = nullptr;
  aclTensor* mOut = nullptr;
  bool upper = true;
  bool transpose = false;
  bool unitriangular = false;
  std::vector<float> selfHostData = {1, 2, 3};
  std::vector<float> otherHostData = {1, 2, 3, 0, 4, 5, 0, 0, 6};
  std::vector<float> xOutHostData = {-0.2500, -0.1250, 0.5000};
  std::vector<float> mOutHostData = {1, 2, 3, 0, 4, 5, 0, 0, 6};
  // Create the self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the other aclTensor.
  ret = CreateAclTensor(otherHostData, otherShape, &otherDeviceAddr, aclDataType::ACL_FLOAT, &other);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the xOut aclTensor.
  ret = CreateAclTensor(xOutHostData, xOutShape, &xOutDeviceAddr, aclDataType::ACL_FLOAT, &xOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the mOut aclTensor.
  ret = CreateAclTensor(mOutHostData, mOutShape, &mOutDeviceAddr, aclDataType::ACL_FLOAT, &mOut);
  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 aclnnTriangularSolve.
  ret = aclnnTriangularSolveGetWorkspaceSize(self, other, upper, transpose, unitriangular, xOut, mOut, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTriangularSolveGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Apply for device memory based on the workspaceSize calculated 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 aclnnTriangularSolve.
  ret = aclnnTriangularSolve(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTriangularSolve failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate) Synchronously wait for the 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 xSize = GetShapeSize(xOutShape);
  std::vector<float> resultData(xSize, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), xOutDeviceAddr,
                    xSize * 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 < xSize; i++) {
    LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
  }

  auto mSize = GetShapeSize(mOutShape);
  std::vector<float> mResultData(mSize, 0);
  ret = aclrtMemcpy(mResultData.data(), mResultData.size() * sizeof(mResultData[0]), mOutDeviceAddr,
                    mSize * sizeof(mResultData[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 < mSize; i++) {
    LOG_PRINT("result[%ld] is: %f\n", i, mResultData[i]);
  }

  // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(other);
  aclDestroyTensor(xOut);
  aclDestroyTensor(mOut);

  // 7. Release device resources. Modify based on the specific API definition.
  aclrtFree(selfDeviceAddr);
  aclrtFree(otherDeviceAddr);
  aclrtFree(xOutDeviceAddr);
  aclrtFree(mOutDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}