aclnnMaxPool

Applicable Products

Applicable Products	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:

Performs a max pooling operation on an input tensor with dim=3 or 4 dimensions.

Formula:
- When ceilMode=False, the formula for deriving the H and W dimensions of the output tensor shape:
  $[H_{out}, W_{out}]=[\lfloor{\frac{H_{in}+ padding\_size_{Htop} + padding\_size_{Hbottom} - {dilation\_size \times(k_h - 1) - 1}}{s_h}}\rfloor + 1,\lfloor{\frac{W_{in}+ padding\_size_{Wleft} + padding\_size_{Wright} - {dilation\_size \times(k_w - 1) - 1}}{s_w}}\rfloor + 1]$
- When ceilMode=True, the derivation formula for the H and W dimensions in the output tensor shape is:
  $[H_{out}, W_{out}]=[\lceil{\frac{H_{in}+ padding\_size_{Htop} + padding\_size_{Hbottom} - {dilation\_size \times(k_h - 1) - 1}}{s_h}}\rceil + 1,\lceil{\frac{W_{in}+ padding\_size_{Wleft} + padding\_size_{Wright} - {dilation\_size \times(k_w - 1) - 1}}{s_w}}\rceil + 1]$
  - When the top-left corner of the sliding window is on the bottom or right padding region or outside the boundary (no valid value can be obtained), the result of that sliding window is discarded, and the corresponding spatial axis shape is reduced by 1 based on the above derivation formula: $\begin{cases} H_{out}=H_{out} - 1& \text{if } (H_{out}-1)*s_h>=H_{in}+padding\_size_{Htop} \\ W_{out}=W_{out} - 1& \text{if } (W_{out}-1)*s_w>=W_{in}+padding\_size_{Wleft} \\ \end{cases}\\$

Function Prototype

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

aclnnStatus aclnnMaxPoolGetWorkspaceSize(
  const aclTensor   *self,
  const aclIntArray *kernelShape,
  const aclIntArray *strides,
  const int64_t      autoPad,
  const aclIntArray *pads,
  const aclIntArray *dilations,
  const int64_t      ceilMode,
  aclTensor         *out,
  uint64_t          *workspaceSize,
  aclOpExecutor     **executor)

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

aclnnMaxPoolGetWorkspaceSize

Parameters:

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
self	Input	Tensor to be computed.	-	FLOAT16, FLOAT, BFLOAT16, INT32, INT64, UINT8, INT16, INT8, UINT16.	ND	3-4	√
kernelShape	Input	Specifies the window size for max pooling.	Has a length of 1 or 2, and all array elements must be greater than 0.	INT64	-	-	-
strides	Input	The stride of the sliding window.	The array length is 0, 1, or 2, and all array elements must be greater than 0. When the array length is 0, strides defaults to 1.	INT64	-	-	-
autoPad	Input	Specifies the padding mode.	0 represents "NOTSET", and only the value 0 is supported.	INT64	-	-	-
pads	Input	Specifies the padding at the start and end positions along the spatial axes, corresponding to padding_size in the formula.	The length is 0, 1, 2, or 4.	INT64	-	-	-
dilations	Input	Dilation value along the kernel spatial axis, corresponding to dilation_size in the formula.	Only supports input scenarios with a value of 1, with a length of 0, 1, 2, or 4.	INT64	-	-	-
ceilMode	Input	Rounding mode for computing the output shape.	When set to 0, represents False, rounding down; when set to a non-zero value, represents True, rounding up.	INT64	-	-	-
out	Output	Output tensor.	Has the same data type as self. The shape is derived from the formula above. The data format and dimensions are consistent with the input self.	FLOAT16, FLOAT, BFLOAT16, INT32, INT64, UINT8, INT16, INT8, UINT16.	ND	3-4	√
workspaceSize	Output	Returns the workspace size to be requested on the Device side.	-	-	-	-	-
executor	Output	Returns the op executor, which contains the operator computation flow.	-	-	-	-	-

Atlas training products: The data type of the self and out parameters does not support BFLOAT16, INT32, INT64, UINT8, INT16, INT8, or UINT16.
Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products: The data types BFLOAT16, INT32, INT64, UINT8, INT16, INT8, and UINT16 are not supported for the self and out parameters.

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 Code	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	The passed self or out is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data type or data format of self is not within the supported range.
		The dimension of self is not 3D or 4D.
		The output shape computed based on max pooling semantics is inconsistent with the specified shape.
		The length of kernelShape is not equal to 1 or 2.
		kernelShape contains values less than or equal to 0.
		The length of strides is not equal to 0, 1, or 2.
		A value in strides is less than or equal to 0.
		The sum of pad amounts in a single spatial axis direction must be less than or equal to kernelShape in the corresponding direction.
		The length of dilations is not 0, 1, 2, or 4.
		A value in dilations is not equal to 1.

aclnnMaxPool

Parameters:

Parameter	Input/Output	Description
workspace	Input	Specifies the workspace memory address applied for on the Device side.
workspaceSize	Input	Specifies the workspace size applied for on the Device side, which is obtained by the first-phase API aclnnMaxPoolGetWorkspaceSize.
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.

Constraints

Deterministic computation:
aclnnMaxPool 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 <cstdio>
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_max_pool.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-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);
  // Copy host-side data to device-side memory using aclrtMemcpy.
  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 the aclCreateTensor API to create an aclTensor.
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_NCHW,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. (Boilerplate) Initialize the device/stream. See the ACL API manual.
  // Set deviceId based on the 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 interface.
  std::vector<int64_t> selfShape = {2, 2, 3, 3};
  std::vector<int64_t> outShape = {2, 2, 2, 2};
  std::vector<int64_t> kernel_size = {2, 2};
  std::vector<int64_t> strides_size = {2, 2};
  std::int64_t autoPads = 0;
  std::vector<int64_t> padding_size = {0, 0, 0, 0};
  std::vector<int64_t> dilation_size = {1, 1};
  std::int64_t ceilMode = 1;

  void* selfDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* out = nullptr;
  aclIntArray* kernel_shape = aclCreateIntArray(kernel_size.data(), 2);
  aclIntArray* strides = aclCreateIntArray(strides_size.data(), 2);
  aclIntArray* padding = aclCreateIntArray(padding_size.data(), 4);
  aclIntArray* dilations = aclCreateIntArray(dilation_size.data(), 2);
  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, 25, 26, 27,
                                     28, 29, 30, 31, 32, 33, 34, 35, 36};
  std::vector<float> outHostData = {0, 0, 0, 0, 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 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. Modify it to the specific API name.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnMaxPool.
  ret = aclnnMaxPoolGetWorkspaceSize(self, kernel_shape, strides, autoPads, padding, dilations, ceilMode, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnMaxPoolGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate 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 aclnnMaxPool second-phase API.
  ret = aclnnMaxPool(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnMaxPool failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate) Synchronize and 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(outShape);
  std::vector<float> outData(size, 0);
  ret = aclrtMemcpy(outData.data(), outData.size() * sizeof(outData[0]), outDeviceAddr,
                    size * sizeof(outData[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, outData[i]);
  }

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

  // 7. Release device resources.
  aclrtFree(selfDeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}