Skip to content

Advanced setup for AVA Platform#

Overview#

This instruction explains how to install advanced software for AVA platform.

  1. Desktop environment(XFCE)
  2. NVIDIA driver
  3. NVIDIA Container Toolkit

Checkout the repository for AVA platform#

  1. m5p3nc3r / meta-ewaol-machine

    git clone git@github.com:m5p3nc3r/meta-ewaol-machine.git -b kirkstone-dev

  2. Copy meta-ewaol-ext directory to the project root.

    cp -rf meta-ewaol-machine/meta-ewaol-ext ~/meta-adlink-ampere

  3. Download a missing patch in meta-ewaol-ext layer.

    wget -P ~/meta-adlink-ampere/meta-ewaol-ext/recipes-ewaol/recipes-container/nvidia-container-toolkit/files https://raw.githubusercontent.com/autowarefoundation/open-ad-kit-docs/main/docs/version-2.0/start-guide/installation/patch/0002-TEMPORARY-force-aarch64-rpm.patch

  4. Modify ava.yml.

     header:
   version: 1
   includes:
     - repo: meta-ewaol
-      file: meta-ewaol-config/kas/baremetal.yml
+      file: meta-ewaol-config/kas/baremetal-sdk.yml

 repos:
   meta-ewaol:
     path: meta-ewaol

   meta-adlink-ampere:

+  meta-openembedded:
+    path: layers/meta-openembedded
+    layers:
+      meta-oe:
+      meta-gnome:
+      meta-multimedia:
+      meta-xfce:
+
+  meta-ewaol-ext:
+    path: meta-ewaol-ext


 machine: ava

 bblayers_conf_header:
   base: |
     POKY_BBLAYERS_CONF_VERSION = "2"
     BBPATH = "${TOPDIR}"
     BBFILES ?= ""

+local_conf_header:
+  meta-at: |
+    XSERVER:append = " xserver-xorg-extension-glx xserver-xorg-module-libwfb xserver-xorg-module-exa"
+    IMAGE_INSTALL:append = " packagegroup-core-x11 packagegroup-xfce-extended acpid xf86-video-modesetting mesa-demos nvidia-container-toolkit"
+    DISTRO_FEATURES:append = " opengl x11 glx"
+    PACKAGECONFIG:append:pn-xserver-xorg = " xinerama"
+    IMAGE_FEATURES:append =" x11 x11-base"
+    INSANE_SKIP:${PN}:append = " already-stripped"
+    FILES:${PN}:append =" /usr/share/containers/oci/hooks.d/oci-nvidia-hook.json"

 target:
-  - ewaol-baremetal-image
+  - ewaol-baremetal-sdk-image

  5. Build via kas.

    kas build ava.yml

Flash yocto image#

For example;

sudo bmaptool copy --bmap build/tmp_baremetal/deploy/images/ava/ewaol-baremetal-sdk-image-ava.wic.bmap build/tmp_baremetal/deploy/images/ava/ewaol-baremetal-sdk-image-ava.wic.gz /dev/sdb

Extend rootfs partition#

Follow the instructions Extend rootfs partition.

Boot your machine#

You can see XFCE desktop. XFCE

NVIDIA driver installation#

  1. Create the files required for compiling external modules.

    cd /usr/src/kernel
make modules_prepare

  2. Check NVIDIA graphics card.

    lspci | grep -i nvidia

    💬 You can find a NVIDIA graphics card such as GeForce RTX 3070 Ti.

    000d:01:00.0 VGA compatible controller: NVIDIA Corporation GA104 [GeForce RTX 3070 Ti] (rev a1)
000d:01:00.1 Audio device: NVIDIA Corporation GA104 High Definition Audio Controller (rev a1)

  3. Open the URL in browser.

    Official Drivers NVIDIA

  4. Select your graphics card, then click SEARCH.

    Search NVIDIA driver

  5. Click DOWNLOAD.

    Download NVIDIA driver

  6. Copy link to the file.

    NVIDIA URL

  7. Download NVIDIA driver on AVA platform.

    For example;

    wget https://jp.download.nvidia.com/XFree86/aarch64/515.65.01/NVIDIA-Linux-aarch64-515.65.01.run
chmod +x NVIDIA-Linux-aarch64-515.65.01.run

  8. Stop display manager to install NVIDIA driver.

    systemctl stop xserver-nodm

  9. Install NVIDIA driver.

    For example;

    ./NVIDIA-Linux-aarch64-515.65.01.run

    Building kernel module starts.

    Building kernel module

  10. Press ⬅ to select Yes and then press ↩.

    X configuration

  11. Press ↩.

    Installation completed

Change display output from motherboard to GPU#

You are using VGA, so switch to GPU.

  1. Turn off AVA platform

  2. Connect HDMI or Display Port from your graphics card to your monitor.

  3. Change display mode to HDMI or Display Port on your monitor if needed.

  4. Turn on AVA platform, then the desktop window will be shown via GPU.

Confirm nvidia-docker works (Optional)#

You can confirm nvidia-docker works by the following command.

docker run --gpus all --rm nvidia/cuda-arm64:11.4.0-base nvidia-smi

You can see the outputs like below.

root@ava:~# docker run --gpus all --rm nvidia/cuda-arm64:11.4.0-base nvidia-smi
Unable to find image 'nvidia/cuda-arm64:11.4.0-base' locally
11.4.0-base: Pulling from nvidia/cuda-arm64
55c604a74c4b: Pull complete
657fae4b9575: Pull complete
b2cf3c1bfea9: Pull complete
71492f856142: Pull complete
c74b3fce51ac: Pull complete
Digest: sha256:625c8265d0f88d4250d48958113f1184f96db794fbe5d6d5cdd782f9916ec718
Status: Downloaded newer image for nvidia/cuda-arm64:11.4.0-base
Thu Aug 25 23:17:40 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 515.65.01    Driver Version: 515.65.01    CUDA Version: 11.7     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 0000000D:01:00.0  On |                  N/A |
|  0%   35C    P8    18W / 290W |    234MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+