Qualcomm Hexagon SDK
The Qualcomm® Hexagon™ SDK provides a set of tools, software, and documentation that assists developers in running applications on Qualcomm Technologies-powered devices and leveraging the Hexagon DSPs. The Hexagon SDK is a software development kit that enables embedded developers to access embedded computing resources on the Hexagon DSP. Using the Hexagon SDK, cutting-edge developers with expertise working in native programming environments can tap into world-class, hardware-enabled multimedia features to deliver premium, interactive user experiences in rapid development time.
The SDK is designed to help ensure multimedia processing efficiency, which means increased fluidity, low latency, and superior app performance. It can also help overcome development hurdles such as lack of battery and insufficient processing power. The Hexagon SDK is also designed to enable high computational burdens on the CPU to be offloaded into a heterogeneous computing environment with the use of shared remote code objects.
Features developed on the Hexagon DSP have been deployed by a broad OEM ecosystem that extends to device and consumer product manufacturers for products including tablets, HD TVs, and automotive products. Each Qualcomm Technologies chipset includes multiple Hexagon DSPs such as the compute DSP (cDSP), audio DSP (aDSP), and sensor DSP (also know as sensor low power island (SLPI )). Each of these DSPs implements a specific instruction set architecture (ISA) version. The ISAs supported by the Hexagon SDK are:
The cDSP, which is intended for compute-intensive tasks such as image processing, computer vision, and camera streaming, also includes an instruction set extension for fixed-point vector operations called Hexagon vector eXtensions (HVX) that is not present on other DSPs. Compared to the host CPU, the DSP typically runs at a lower clock speed but provides more parallelism opportunities at the instruction level. This often makes the DSP a better alternative in terms of throughput and/or power consumption. As a result, it is preferable to offload as many large compute-intensive tasks as possible onto the DSP to reduce power consumption of the device and free up cycles on the CPU for additional features.
QRB5165 has an additional hardware block called the Hexagon tensor accelerator (HTA). The HTA is a dedicated, scalable, power-efficient, programmable hardware accelerator for fixed-point deep convolutional neural networks (DCNN) models. The HTA is a part of the Hexagon DSP that can offload neural network inference tasks to the HTA or the cDSP. The HTA offers high-level programming access through the Qualcomm Neural Processing SDK and the Hexagon-HTA-NN (also known as DirectHTA) API in the Hexagon SDK as well as the Google Android NN-API in Android NDK.
Refer to the Qualcomm Developer Network for full documentation to enable access to computing resources on the Hexagon DSP. The following diagram shows some important points of the Hexagon SDK.