Deep learning target library

R2026b

Target library for deep learning code generation

Description

App Configuration Pane: Deep Learning

The Deep learning library specifies the deep learning library for code generation. You can also specify the deep learning target library in command line by using the coder.DeepLearningConfig function.

Dependencies

For generating generic C/C++ code, or C/C++ code that uses third-party libraries such as Intel^® MKL-DNN and ARM^® Compute libraries, you must use the MATLAB Coder App.

For generating plain CUDA code, or CUDA code that uses NVIDIA^® high-performance deep learning libraries, you must:

• Use the GPU Coder App, or

• Open the MATLAB Coder App, then set Output to CUDA in the Prepare section.

Settings

none

Generate generic C/C++ or plain CUDA code that does not depend on any third-party libraries. Choosing this setting sets Language to C++.

MKL-DNN

Generate code that uses the Intel Math Kernel Library for Deep Neural Networks. Choosing this setting sets Language to C++.

ARM Compute

(To be removed) Generate code that uses the ARM Compute Library. Choosing this setting sets Language to C++.

This setting will be removed in a future release. Use none instead. If you need to deploy to an ARM device that supports Neon instructions, set the InstructionSetExtensions property of the code configuration object to "Neon v7".

CMSIS-NN

(To be removed) Generate code that uses the Common Microcontroller Software Interface Standard - Neural Network library. Choosing this setting sets Language to C.

This setting will be removed in a future release. Use none with the ARM Cortex^®-M code replacement library (CRL) instead.

cuDNN

Generate code that uses the CUDA deep Neural Network library. Choosing this setting sets Language to C++.

TensorRT

Generate code that uses the NVIDIA TensorRT deep learning inference optimizer and run-time library. Choosing this setting sets Language to C++.

Version History

Introduced in R2018a

R2026b: ARM Compute and CMSIS-NN settings will be removed

The ARM Compute and CMSIS-NN deep learning target library settings will be removed in a future release.

Instead, create a deep learning configuration object by using the coder.DeepLearningConfig function with the target library set to "none". If you need to deploy to an ARM device that supports Neon instructions, set the InstructionSetExtensions property of the code configuration object to "Neon v7". For ARM Cortex-M targets, set the CodeReplacementLibrary property to "ARM Cortex-M".

Setting InstructionSetExtensions to "Neon v7" and the CodeReplacementLibrary property require Embedded Coder^®. The "none" target library workflow is more capable in MATLAB^® and Simulink^®, and supports a larger set of layers.