Skip to content

Latest commit

 

History

History
95 lines (75 loc) · 4.05 KB

README.md

File metadata and controls

95 lines (75 loc) · 4.05 KB
Raw

doca_rmax_rx_perf

NVIDIA DOCA is a Software Framework to accelerate application development for the NVIDIA BlueField DPU.

doca_rmax_rx_perf is an RX performance measurement application to be used with the NVIDIA DOCA RMAX library. The code is provided "As Is" and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed.

  • Release date: 27-Oct-2022
  • Update date: 27-Oct-2022
  • Version: 1.0
  • Compatibility:
    • Rivermax: 1.20.10
    • OFED: 5.8-0.2.3.0
    • DOCA: 1.5
    • OS: Ubuntu 20.04
    • Hardware: BlueField 2

Documentation

Prerequisites

  • NVIDIA DOCA SDK 1.5.0 or later. See Installation guide.
  • NVIDIA DOCA RMAX library. Can be installed by SDK Manager (see DOCA SDK installation guide) or manually by installing doca-rmax-libs and libdoca-rmax-libs-dev.
  • CMake 3.6 or later.

Command-line parameters

Usage: doca_doca_rmax_perf [DOCA Flags] [Program Flags]

DOCA Flags:
  -h, --help                        Print a help synopsis
  -v, --version                     Print program version information
  -l, --log-level                   Set the log level for the program <CRITICAL=20, ERROR=30, WARNING=40, INFO=50, DEBUG=60>

Program Flags:
  --list                            List available devices
  -t, --stream-type                 Stream type: generic (default) or RTP-2110
  --scatter-type                    Scattering type: RAW (default), ULP or payload
  --tstamp-format                   Timestamp format: counter (default), nano or synced
  -i, --interface-ip                IP of the local interface to receive data
  -m, --multicast-dst               Multicast address to bind to
  -s, --multicast-src               Source address to read from
  -p, --port                        Destination port to read from
  -r, --header-size                 Header size (default 0)
  -d, --data-size                   Data size (default 1500)
  -k, --packets                     Number of packets to allocate memory for (default 1024)
  -a, --cpu-affinity                Comma separated list of CPU affinity cores for the application main thread
  --sleep                           Amount of microseconds to sleep between requests (default 0)
  --min                             Block until at least this number of packets are received (default 0)
  --max                             Maximum number of packets to return in one completion
  --dump                            Dump packet content

Examples:

  • List available devices: doca_doca_rmax_perf --list
  • Receive a stream: doca_doca_rmax_perf --interface-ip 1.1.64.67 --multicast-dst 1.1.64.67 --multicast-src 1.1.63.5 --port 7000
  • Receive a stream (header-data split mode): doca_doca_rmax_perf --interface-ip 1.1.64.67 --multicast-dst 1.1.64.67 --multicast-src 1.1.63.5 --port 7000 --header-size 20 --data-size 1200

How to build

From the doca_doca_rmax_perf directory run the following commands:

$ cmake -B ./build
$ cmake --build ./build

Enjoy the doca_rmax_rx_perf binary in directory build.

Testing

Was tested in NVIDIA lab:

  • Card: (MT41686 - MBF2M516A-CENOT) BlueField-2 DPU 100GbE Dual-Port QSFP56, Crypto Disabled, 16GB on-board DDR, 1GbE OOB management, Tall Bracket
  • Rivermax: 1.20.9
  • OFED (on BF): MLNX_OFED_LINUX-5.8-0.3.3.0
  • FW: 24.35.0314
  • DOCA: 1.5.0041-1
  • BFB: DOCA_1.5.0_BSP_3.9.3_Ubuntu_20.04-5.20221003

Limitations

  • Supported only on aarch64 platform and Ubuntu 20.04.
  • --scatter-type payload is not yet supported by Rivermax for Generic and ST 2110 streams.