Installing Prerequisite Products

To use GPU Coder™ for CUDA^® code generation, you must install and setup the following products. For setup instructions , see Setting Up the Prerequisite Products.

MathWorks Products and Support Packages

• MATLAB^® (required).

• MATLAB Coder™ (required).

• Parallel Computing Toolbox™ (required).

• Simulink^® (required for generating code from Simulink models).

• Deep Learning Toolbox™ (required for deep learning).

• Embedded Coder^® (required for GPU performance analyzer).

• Simulink Coder (required for generating code from Simulink models).

• GPU Coder Interface for Deep Learning support package (required for deep learning).

• MATLAB Coder Support Package for NVIDIA^® Jetson™ and NVIDIA DRIVE^® Platforms (required for deployment to embedded targets such as NVIDIA Jetson and Drive).

• Computer Vision Toolbox™ (recommended).

• Image Processing Toolbox™ (recommended).

For instructions on installing MathWorks^® products, see the MATLAB installation documentation for your platform. If you have installed MATLAB and want to check which other MathWorks products are installed, enter ver in the MATLAB Command Window.

To install the support packages, use the Add-On Explorer in MATLAB.

If MATLAB is installed on a path that contains non 7-bit ASCII characters, such as Japanese characters, GPU Coder does not work because it cannot locate code generation library functions.

Third-Party Hardware

• NVIDIA GPU enabled for CUDA with a compatible graphics driver. For more information, see CUDA GPUs (NVIDIA).

  To see the CUDA compute capability requirements for code generation, consult the following table.

  Target	Compute Capability
  CUDA MEX	See GPU Computing Requirements.
  Source code, static or dynamic library, and executables	3.2 or higher.
  Deep learning applications in 8-bit integer precision	6.1, 7.0 or higher.
  Deep learning applications in half-precision (16-bit floating point)	5.3, 6.0, 6.2 or higher.

• ARM^® Mali graphics processor.

  For the Mali device, GPU Coder supports code generation for only deep learning networks.

Third-Party Software

Required

Software Name	Additional Information	Linux®	Windows®
C/C++ Compiler	N/A	GCC C/C++ compiler For supported versions, see Supported and Compatible Compilers.	Microsoft® Visual Studio® 2022
			Microsoft Visual Studio 2019
			Microsoft Visual Studio 2017
NVIDIA Display Driver	Minimum required driver version CUDA applications can run with limited feature set, on systems with the minimum required driver version.	>= 450.80.02	>= 452.39
	Driver version for full compatibility	>= 520.61.05	>= 520.06

Note

The table lists the driver version requirements for CUDA Toolkit v11.8. For different CUDA Toolkit versions, refer to the CUDA Toolkit release notes (NVIDIA).

Optional

For CUDA MEX, the code generator uses the NVIDIA compiler and libraries installed with MATLAB. Standalone code (static library, dynamically linked library, or executable program) generation has additional software requirements.

Software Name	Information
CUDA Toolkit	GPU Coder has been tested with CUDA Toolkit v9.x-v12.2. To download the CUDA Toolkit, see CUDA Toolkit Archive (NVIDIA).
NVIDIA CUDA deep neural network library (cuDNN) for NVIDIA GPUs	For the host GPU device, GPU Coder has been tested with cuDNN v8.9. To download cuDNN, see cuDNN (NVIDIA).
NVIDIA TensorRT™ high performance inference optimizer and runtime library	For the host GPU device, GPU Coder has been tested with TensorRT v8.6.1. To download TensorRT, see TensorRT (NVIDIA).
ARM Compute Library for Mali GPUs	GPU Coder has been tested with v19.05. For more information, see Compute Library (ARM).
Open Source Computer Vision Library (OpenCV)	Required for deep learning examples. For examples targeting NVIDIA GPUs on the host development computer, use OpenCV v3.1.0. For examples targeting ARM GPUs, use OpenCV v2.4.9 on the ARM target hardware. For more information, see OpenCV.

Tips

 General

• On Windows, a space or special character in the path to the tools, compilers, and libraries can create issues during the build process. You must install third-party software in locations that does not contain spaces or change Windows settings to enable creation of short names for files, folders, and paths. For more information, see Using Windows short names solution in MATLAB Answers.

• You must download the OpenCV source and build the libraries. The examples require separate libs such as, opencv_core.lib and opencv_video.lib. For more information, refer to the OpenCV documentation.

 CUDA Toolkit

• The NVIDIA nvcc compiler relies on tight integration with the host development environment, including the host compiler and runtime libraries. It is recommended that you follow the CUDA Toolkit documentation for detailed information on compiler, libraries, and other platform specific requirements. See, CUDA Toolkit Documentation (NVIDIA).

• It is recommended to select the default installation options that includes nvcc compiler, cuFFT, cuBLAS, cuSOLVER, Thrust libraries, and other tools.

• The nvcc compiler supports multiple versions of GCC and therefore you can generate CUDA code with other versions of GCC. However, there may be compatibility issues when executing the generated code from MATLAB as the C/C++ run-time libraries that are included with the MATLAB installation are compiled for only the supported version of GCC.

• The GPU Performance Analyzer workflow depends on profiling tools from NVIDIA. From CUDA Toolkit v10.1 onwards, NVIDIA restricts access to performance counters to only admin users. To enable GPU performance counters to be used by all users, see the instructions provided in Permission issue with Performance Counters (NVIDIA).

• GPU Coder does not support generating CUDA code by using CUDA Toolkit version 8.

 Deep Learning

• Other versions of these deep learning libraries may have compatibility issues with the features GPU Coder supports this release.

 NVIDIA Embedded Targets

• On the target platforms, use the JetPack or the DriveInstall software for your board to install all the libraries. For more information, see Install and Setup Prerequisites for NVIDIA Boards.

 ARM Mali

• This library must be installed on the ARM target hardware. Do not use a prebuilt library because it might be incompatible with the compiler on the ARM hardware. Instead, build the library from the source code. Build the library on either your host machine or directly on the target hardware. See instructions for building the library on GitHub^®. You can also find information on building the library for CPUs in this post on MATLAB answers.

• When building the Compute Library, enable OpenCL support in the build options. See the ARM Compute Library documentation for instructions. OpenCL library (v1.2 or higher) on the ARM target hardware. See the ARM Compute Library documentation for version requirements. After the build is complete, rename the build folder containing the libraries as lib. Additionally, copy the OpenCL libraries present in the build/opencl-1.2-stubs folder into the lib folder. These steps are required so that the generated makefile can locate the libraries when building the generated code on the target hardware.