Skip to content
No description, website, or topics provided.
C Cuda Makefile
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Makefile
README.md
rdma_client.c
rdma_common.c
rdma_common.h
rdma_gpu.cu
rdma_server.c

README.md

Overview

This program uses GPUDirect RDMA to communicate between server and client GPU. Source code is created using the following code.

Requirements

Operating Environment

The environment used for the operation check is as follows. There is no guarantee that it will work in other environments.

  • OS
    • Server : Linux 4.10.0
    • Client : Linux 4.4.64
  • GPU : NVIDIA Quadro M4000
  • NIC : Mellanox ConnectX-4 VPI
  • Driver
    • NVIDIA GPU driver 375.66
    • CUDA 8.0.61
    • Mellanox OFED 4.1
    • nvidia-peer-memory 1.0.7

Procedure

First, run the following command on the two hosts.

$ make

After command execution, rdma_server and rdma_client are created. Second, run the following command on the server.

$ ./rdma_server [-a <server_addr>] [-p <server_port>]

Third, run the following command on the client.

$ ./rdma_client [-a <server_addr>] [-p <server_port>]
You can’t perform that action at this time.