# Auto Platform Overview

The Qualcomm® AI Engine Direct SDK supports a family of automotive SA-series SoCs
(SA8295 through SA8797). These platforms share the same QNN API and tooling as other
Qualcomm® AI Engine Direct targets, but have automotive-specific target environments,
deployment workflows, and hardware configurations.

Use this page to find the documentation that applies to your chipset, target environment,
and what you are trying to do.

| Environment | Description |
| --- | --- |
| **LA GVM** | Android guest VM (`aarch64-android` toolchain). |
| **LV GVM** | Linux OE guest VM (`aarch64-linux-oe-gcc9.3` toolchain). |
| **QC Linux PVM** | QC Linux primary VM on SA8797 (`aarch64-oe-linux-gcc11.2` toolchain). |

## Run Your First Model

These tutorials walk you through getting a model running on your target device end-to-end — from model conversion through on-device execution.

**To start, choose the right tutorial for you based on your target device:**

| SoC | Target environment | Tutorial |
| --- | --- | --- |
| SA8295, SA8540, SA8620, SA8650, SA8775 | LA GVM (Android) or LV GVM (Linux OE) | [HTP Backend on SA-series (LA GVM and LV GVM)](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_tutorial_2.html) |
| SA8295, SA8540, SA8620, SA8650, SA8775 | LV GVM (Linux OE) | [GPU Backend on SA-series (LV GVM)](https://docs.qualcomm.com/doc/80-63442-10/topic/gpu_auto_tutorial_2.html) |
| SA8797 | QC Linux PVM | [HTP and LPAI on SA8797 QC Linux PVM](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_qclinux.html) |
| x86 host | QEMU emulation (no hardware required) | [HTP Emulation on x86 using QEMU](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_qemu.html) |

Note

The HTP tutorials above use the serialized context binary workflow
(`qnn-context-binary-generator` → `qnn-net-run --retrieve_context`), which is
the recommended approach for automotive platforms. Before running these tutorials,
complete model conversion using
[Tutorial: Converting and executing a CNN model](https://docs.qualcomm.com/doc/80-63442-10/topic/tutorial_convert_execute_cnn_model.html).

## Custom Op Packages

If your model uses operations not natively supported by Qualcomm® AI Engine Direct,
you will need to build and deploy a custom op package alongside your model.

| SoC | Target environment | Tutorial |
| --- | --- | --- |
| SA8295, SA8540, SA8620, SA8650, SA8775 | LA GVM (Android) or LV GVM (Linux OE) | [HTP with Custom Ops on SA-series (LA GVM and LV GVM)](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_tutorial_3.html) |

Note

Custom op package support for SA8797 QC Linux PVM is not yet documented.
Refer to the HTP Backend page for op package API reference.

## Build an App

Once your model runs correctly with `qnn-net-run`, the next step is integrating it
into a C++ application using the QNN API directly.

- [Sample App Tutorial](https://docs.qualcomm.com/doc/80-63442-10/topic/sample_app.html) — walks through the QNN API workflow
(loading libraries, creating context, preparing graphs, executing inference) using
the `qnn-sample-app` example included in the SDK. This tutorial is not
Auto-specific and applies to all platforms including SA-series.

## Optimize

After your application is working, use these pages to tune for performance on
your specific hardware.

- [Selecting NSP0 or NSP1 for HTP execution](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_single_nsp.html) —
control which CDSP/NSP executes your model using `device_id` in the backend
config. Applies to all SA-series SoCs with HTP.
- [HTP and HTP MCP optimization (O=3, P-points, performance estimates)](https://docs.qualcomm.com/doc/80-63442-10/topic/htp_auto_optimization.html) —
enable graph optimization levels, experiment with P-point compiler configurations,
and generate performance estimates using `qnn-context-binary-generator`.
Applies to all SA-series SoCs with HTP.

## QNX Targets

QNX-specific documentation is **not** included in this section of the SDK.
It is distributed as a separate Add-On and is only available to customers with
the appropriate entitlements.

If you have a QNX entitlement, your SDK zip includes the QNX Add-On documentation at:

Add-Ons/QNX/
    Copy to clipboard

Last Published: Jun 04, 2026

[Previous Topic
Saver Tutorial: Save execution sequence with Saver and replay on a backend](https://docs.qualcomm.com/bundle/publicresource/80-63442-10/topics/saver_tutorial.md) [Next Topic
Tutorial: Running QNN HTP on SA-series](https://docs.qualcomm.com/bundle/publicresource/80-63442-10/topics/htp_auto_tutorial_2.md)