Deep learning library for simulation

Deep learning library to use during simulation

Model Configuration Pane: Simulation Target

Description

The Deep learning library parameter specifies the deep learning library to use during simulation. To enable the Deep learning library parameter, set Language to C++. When the Language is C, the Deep learning library parameter is hidden and the value of this parameter is None.

Dependencies

• When Language is set to C++ and GPU acceleration is disabled:

  • Default value for Target library is MKL-DNN.

  • Option None or MKL-DNN is available for Target library.

• When GPU acceleration on the Simulation Target pane is enabled (requires a GPU Coder™ license):

  • Default value for Target library is cuDNN.

  • Option None, cuDNN or TensorRT is available for Target library.

• If Target library is previously set to None, the value for Target library stays as None when GPU acceleration is enabled or disabled.

Settings

None | MKL-DNN | cuDNN | TensorRT

Default:

• None if GPU acceleration is off and Language is C.

• MKL-DNN if GPU acceleration is off and Language is C++.

• cuDNN if GPU acceleration is on.

None (since R2025a)

Simulates the model that is supported for generic C/C++ and plain CUDA workflows. For more information on networks and layers supported for code generation, see and Networks and Layers Supported for Code Generation (MATLAB Coder).

MKL-DNN

Simulates the model using the Intel^® Math Kernel Library for Deep Neural Networks (Intel MKL-DNN).

cuDNN

Simulate the model using the CUDA^® Deep Neural Network library (cuDNN).

TensorRT

Simulate the model using the NVIDIA^® TensorRT high performance deep learning inference optimizer and run-time library.

Recommended Settings

Application	Setting
Debugging	No impact
Traceability	No impact
Efficiency	No impact
Safety precaution	No impact

Programmatic Use

Parameter: SimDLTargetLibrary

Type: character vector

Value:'None' | 'MKL-DNN' | 'cuDNN' | 'TensorRT'

Default: 'None' for language C | 'MKL-DNN' for language C++

Version History

Introduced in R2020b

R2025a: Version History

Starting from R2025a, for models loaded from earlier releases where Simulink simulation target library is previously set to MKL-DNN and target language is C, the value of Deep learning library is set to None.