QIR software architecture - Qualcomm Dragonwing Documentation

The QIR SDK builds on top of the Qualcomm Linux distributions and various function SDKs. The following figure shows the overall software architecture.

Figure: Overall software architecture of QIR SDK

Layers

The QIR SDK builds on top of the open-source Yocto Project and integrates multiple foundational software layers, making it easier to use, customize, and extend. These layers include:

Table: Layers of the QIR SDK

Layer	Description
Qualcomm Linux Distribution	This is a collection of layers from Qualcomm. It provides the foundational Linux environment and multimedia support.
Robotics SDK layers	This is a collection of layers from Qualcomm. It provides the robot-related support.
meta-ros layer	This is a series of OpenEmbedded layers that adds Robot Operating System (ROS) support to embedded Linux distributions built with the Yocto Project. For more details, see meta-ros.

Component function SDKs

The QIR SDK integrates specialized components called function SDKs. The following table lists the details:

Table: Function SDKs used by QIR SDK

Function SDK	Description
Qualcomm® Intelligent Multimedia SDK	The Qualcomm Intelligent Multimedia SDK (IM SDK) provides a unified environment for developing AI and multimedia use cases at the Edge on the Qualcomm® Linux platforms. For more details, see the Qualcomm Intelligent Multimedia SDK (IM SDK) documentation.
Qualcomm® Neural Processing SDK	The Qualcomm Neural Processing SDK enables execution, optimization, and deployment of deep neural networks across Snapdragon CPUs, GPUs, and DSPs, supporting model conversion, quantization, debugging, and application integration through C++/Java. For more details, see the Qualcomm AI Runtime (QAIRT) Overview.
Qualcomm® AI Engine Direct SDK	The Qualcomm AI Engine Direct SDK helps build model files (for example, from PyTorch) to run on various device processors (CPU, GPU, DSP, HTP, or LPAI) across multiple operating systems. For more details, see the Qualcomm AI Runtime (QAIRT) Overview.

The following figure shows a more detailed tech stack. The QIR SDK implements its core functionality modules, based on which it builds the sample applications. Besides, the QIR SDK provides tools to make the development process smooth and faster. <QIR SDK tech stack diagram>

Figure: QIR SDK tech stack

Sample applications

To help you understand the features, the QIR SDK provides sample applications in the following categories:

Sample application categories	Description
Platform samples	Provides ROS samples for common platform types including OCR, Rplidar, Orbbec Camera, and IMU.
Robotics samples	Provides a robotics simulation platform where you can develop your own robotics applications, such as `simulation-sample-AMR-simple-motion`.

Tools

To speed up the deployment of your development environment, the QIR SDK integrates the following tools:

Tool name	Description
Cross-compile toolchain	Provides cross-compile toolchain, which can compile binaries on the host that run on the device.
ROS2 URDF FILE	Provides the URDF file for the robot, which can visualize the robot in ROS2.
Easy-to-use script	Provides the single-command script for configuring the device.

​Layers

​Component function SDKs

​Sample applications

​Tools

Layers

Component function SDKs

Sample applications

Tools