CANN/HCCL环状批量收发示例

Point-to-Point Communication - HcclBatchSendRecv (Ring)

【免费下载链接】hccl集合通信库（Huawei Collective Communication Library，简称HCCL）是基于昇腾AI处理器的高性能集合通信库，为计算集群提供高性能、高可靠的通信方案项目地址: https://gitcode.com/cann/hccl

Sample Description

This sample demonstrates how to use theHcclBatchSendRecv()API to implement point-to-point communication in a ring topology. It covers the following functions:

• CallaclrtGetDeviceCount()to detect devices and query the number of available devices.

• CallHcclGetRootInfo()and userank 0as the root rank to generate the rootinfo identifier.

  The rootinfo identifier contains the device IP address and device ID. This information must be broadcast to all ranks in the cluster to initialize the communicator.

• In each thread, callHcclCommInitRootInfo()to initialize the communicator based on the rootinfo identifier.

• Call theHcclBatchSendRecv()API to send data to the next node while receiving data from the previous node, and display the result.

Directory Structure

├── main.cc # Sample source file ├── Makefile # Compilation and build configuration file └── batch_send_recv_ring # Compiled executable file

Environment Preparation

Environment Requirements

This sample supports the following products in a single-server N-card configuration (N >= 2):

• Ascend 950PR / Ascend 950DT
• Atlas A3 Training Series Products / Atlas A3 Inference Series Products
• Atlas A2 Training Series Products
• Atlas Training Series Products

Setting Environment Variables

# Set CANN environment variables. The following uses the root user default installation path as an example. source /usr/local/Ascend/cann/set_env.sh

Compiling and Running the Sample

Run the following commands in the sample code directory:

make make test

Note: You can set theHCCL_OP_EXPANSION_MODEenvironment variable to configure the task orchestration expansion location of communication algorithms. For the supported ranges for different product models, see the usage instructions for this environment variable in the Environment Variable List.

# Set the orchestration expansion location of communication algorithms to the AI CPU on the Device side. The Device side automatically selects the corresponding scheduler based on the hardware model. export HCCL_OP_EXPANSION_MODE=AI_CPU

Sample Output

ThesendBufcontent on each node is initialized to the Device ID. Data is sent to the next node and received from the previous node. Therefore, each node receives the Device ID of the previous node.

Found 8 NPU device(s) available rankId: 0, output: [ 7 7 7 7 7 7 7 7 ] rankId: 1, output: [ 0 0 0 0 0 0 0 0 ] rankId: 2, output: [ 1 1 1 1 1 1 1 1 ] rankId: 3, output: [ 2 2 2 2 2 2 2 2 ] rankId: 4, output: [ 3 3 3 3 3 3 3 3 ] rankId: 5, output: [ 4 4 4 4 4 4 4 4 ] rankId: 6, output: [ 5 5 5 5 5 5 5 5 ] rankId: 7, output: [ 6 6 6 6 6 6 6 6 ]

【免费下载链接】hccl集合通信库（Huawei Collective Communication Library，简称HCCL）是基于昇腾AI处理器的高性能集合通信库，为计算集群提供高性能、高可靠的通信方案项目地址: https://gitcode.com/cann/hccl