This layer provides recipes for integrating the DeepX M1 hardware accelerator, its runtime (DXRT), and associated applications into Yocto Project-based Linux distributions.
The meta-deepx-m1 layer is essential for building embedded Linux images that target the DeepX M1 platform. It encapsulates all necessary software components, including low-level kernel modules, core inference libraries, and demonstration applications.
Key responsibilities include:
- Providing M1-specific kernel drivers (
dx-driver). - Integrating the DeepX Runtime (DXRT) libraries (
dx-rt). - Enabling ONNX Runtime (ORT) execution through DXRT.
- Packaging DeepX Applications (e.g., streaming demos).
This layer depends on several core Yocto layers. Ensure these layers are included in your conf/bblayers.conf file before attempting to use meta-deepx-m1.
| Layer Name | URL/Source | Notes |
|---|---|---|
poky/meta |
(Included by default) | Core metadata layer |
meta-openembedded |
(e.g., meta-oe, meta-python) |
Required for common utilities and Python support |
meta-yocto-bsp |
(Target specific BSP) | The specific Board Support Package (BSP) layer for the M1 target hardware |
Clone this repository into your Yocto build environment's source directory:
cd <your-yocto-build-dir>
git clone -b scarthgap https://github.com/DEEPX-AI/meta-deepx-m1.gitFor more detailed setup instructions and examples, please refer to the following documentation: