API#
These pages contain full API reference documentation for SDK components.
For more general information about their usage, please refer to the Runtime, Tools, and Tutorials sections.
Runtime#
Runtime APIs for using the MX3 Accelerator.
Benchmark API for simple tests, or non-realtime processing.
Compiler#
The Neural Compiler compiles neural network models into Dataflow Programs (DFP) to run on the MXA.
Utils#
The Simulator provides a bit-accurate / approximately timed model for MXAs.
A container class for the DFP that provides easy access to metadata like input/output shapes, number of MXAs etc.
A C++ library based on Qt5 to aid development of C++ applications.
Advanced#
The low-level driver API (Advanced use cases only). This API is intended for experienced developers who need fine-grained control over the MXA.