Sample Overview

Sample Obtaining

Visit the Ascend samples repository on Gitee and download the sample package that matches your CANN version in use. For the version mapping, see "Release Notes" in the README file. Find the "acl_operator_add" sample in the python/level1_single_api/1_acl/4_blas/acl_operator_add directory.

Function Description

This sample verifies the functionality of the custom operator by converting the custom operator file into a single-operator offline model file and loading the file using the ACL for execution.

This sample implements the matrix-matrix addition operation: C = A + B. In this operation, A, B, and C are 8 x 16 matrices and are of the int32 type. The result is an 8 x 16 matrix.

Main APIs

Table 1 shows the main APIs.

**Table 1** **Main APIs**
Function	ACL Module	ACL Interface Function	Description
Resource initialization	Initialization	acl.init	Initializes the ACL configuration.
	Device management	acl.rt.set_device	Specifies the device for computation.
	Context management	acl.rt.create_context	Creates a context.
	Stream management	acl.rt.create_stream	Creates a stream.
	Operator loading and execution	acl.op.set_model_dir	Loads the directory of a model file.
Data postprocessing	Operator loading and execution	acl.op.create_attr	Creates data of the aclopAttr type.
	--	acl.create_tensor_desc	Creates data of the aclTensorDesc type.
	--	acl.get_tensor_desc_size	Obtains the size of a tensor description.
	--	acl.create_data_buffer	Creates data of the aclDataBuffer type.
Data exchange	Memory management	acl.rt.memcpy	Sends data from the host to the device or from the device to the host.
	Memory management	acl.rt.malloc	Allocates device memory.
	Memory management	acl.rt.malloc_host	Allocates memory on the host.
Single-operator inference	Operator loading and execution	acl.op.execute	Loads and executes an operator. It is an asynchronous interface.
Common module	--	acl.util.ptr_to_numpy	Obtains the numpy.ndarray object based on the pointer address.
Common module	--	acl.util.numpy_to_ptr	Obtains the pointer address of the memory data of the numpy.ndarray object.
Allocation destruction	Memory management	acl.rt.free	Frees device memory.
	Memory management	acl.rt.free_host	Frees the memory on the host.
	Stream management	acl.rt.destroy_stream	Destroys a stream.
	Context management	acl.rt.destroy_context	Destroys a context.
	Device management	acl.rt.reset_device	Resets the current device and reclaims the resources on the device.
	Deinitialization	acl.finalize	Deinitializes ACL.

Single-Operator Matrix Addition Process

Figure 1 shows the single-operator matrix addition process.

Figure 1 Single-operator matrix addition process

Directory Structure

The following is an example of the directory structure after the model file has been converted. The op_models folder is generated after the conversion.

acl_operator_add
├──src
│ ├── acl_execute_add.py // Running file
│ └── constant.py // Constant definition
└── test_data
  ├── config
  │ ├── acl.json // Configuration file for system initialization
  │ └── add_op.json // Description information of the matrix-matrix addition operator
  └── op_models
     └── 0_Add_3_2_8_16_3_2_8_16_3_2_8_16.om //Model file of the matrix addition operator

Parent topic: Matrix-Matrix Addition