New in version 3.17.

This script locates the NVIDIA CUDA toolkit and the associated libraries, but does not require the CUDA language be enabled for a given project. This module does not search for the NVIDIA CUDA Samples.

New in version 3.19: QNX support.

Search Behavior

The CUDA Toolkit search behavior uses the following order:

  1. If the CUDA language has been enabled we will use the directory containing the compiler as the first search location for nvcc.

  2. If the CUDAToolkit_ROOT cmake configuration variable (e.g., -DCUDAToolkit_ROOT=/some/path) or environment variable is defined, it will be searched. If both an environment variable and a configuration variable are specified, the configuration variable takes precedence.

    The directory specified here must be such that the executable nvcc or the appropriate version.txt or version.json file can be found underneath the specified directory.

  3. If the CUDA_PATH environment variable is defined, it will be searched for nvcc.

  4. The user's path is searched for nvcc using find_program(). If this is found, no subsequent search attempts are performed. Users are responsible for ensuring that the first nvcc to show up in the path is the desired path in the event that multiple CUDA Toolkits are installed.

  5. On Unix systems, if the symbolic link /usr/local/cuda exists, this is used. No subsequent search attempts are performed. No default symbolic link location exists for the Windows platform.

  6. The platform specific default install locations are searched. If exactly one candidate is found, this is used. The default CUDA Toolkit install locations searched are:


    Search Pattern



    Other Unix



    C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\vX.Y

    Where X.Y would be a specific version of the CUDA Toolkit, such as /usr/local/cuda-9.0 or C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0


    When multiple CUDA Toolkits are installed in the default location of a system (e.g., both /usr/local/cuda-9.0 and /usr/local/cuda-10.0 exist but the /usr/local/cuda symbolic link does not exist), this package is marked as not found.

    There are too many factors involved in making an automatic decision in the presence of multiple CUDA Toolkits being installed. In this situation, users are encouraged to either (1) set CUDAToolkit_ROOT or (2) ensure that the correct nvcc executable shows up in $PATH for find_program() to find.



The [<version>] argument requests a version with which the package found should be compatible. See find_package version format for more details.



If specified, configuration will error if a suitable CUDA Toolkit is not found.


If specified, the search for a suitable CUDA Toolkit will not produce any messages.


If specified, the CUDA Toolkit is considered found only if the exact VERSION specified is recovered.

Imported targets

An imported target named CUDA::toolkit is provided.

This module defines IMPORTED targets for each of the following libraries that are part of the CUDAToolkit:

CUDA Runtime Library

The CUDA Runtime library (cudart) are what most applications will typically need to link against to make any calls such as cudaMalloc, and cudaFree.

Targets Created:

  • CUDA::cudart

  • CUDA::cudart_static

CUDA Driver Library

The CUDA Driver library (cuda) are used by applications that use calls such as cuMemAlloc, and cuMemFree.

Targets Created:

  • CUDA::cuda_driver


The cuBLAS library.

Targets Created:

  • CUDA::cublas

  • CUDA::cublas_static

  • CUDA::cublasLt starting in CUDA 10.1

  • CUDA::cublasLt_static starting in CUDA 10.1


New in version 3.27.

The NVIDIA Tegra Deep Learning Accelerator cuDLA library.

Targets Created:

  • CUDA::cudla starting in CUDA 11.6


New in version 3.25.

The NVIDIA GPUDirect Storage cuFile library.

Targets Created:

  • CUDA::cuFile starting in CUDA 11.4

  • CUDA::cuFile_static starting in CUDA 11.4

  • CUDA::cuFile_rdma starting in CUDA 11.4

  • CUDA::cuFile_rdma_static starting in CUDA 11.4


The cuFFT library.

Targets Created:

  • CUDA::cufft

  • CUDA::cufftw

  • CUDA::cufft_static

  • CUDA::cufft_static_nocallback starting in CUDA 9.2, requires CMake 3.23+

  • CUDA::cufftw_static


The cuRAND library.

Targets Created:

  • CUDA::curand

  • CUDA::curand_static


The cuSOLVER library.

Targets Created:

  • CUDA::cusolver

  • CUDA::cusolver_static


The cuSPARSE library.

Targets Created:

  • CUDA::cusparse

  • CUDA::cusparse_static


The NVIDIA CUDA Profiling Tools Interface.

Targets Created:

  • CUDA::cupti

  • CUDA::cupti_static

New in version 3.27:

  • CUDA::nvperf_host starting in CUDA 10.2

  • CUDA::nvperf_host_static starting in CUDA 10.2

  • CUDA::nvperf_target starting in CUDA 10.2

  • CUDA::pcsamplingutil starting in CUDA 11.3


The NPP libraries.

Targets Created:

  • nppc:

    • CUDA::nppc

    • CUDA::nppc_static

  • nppial: Arithmetic and logical operation functions in nppi_arithmetic_and_logical_operations.h

    • CUDA::nppial

    • CUDA::nppial_static

  • nppicc: Color conversion and sampling functions in nppi_color_conversion.h

    • CUDA::nppicc

    • CUDA::nppicc_static

  • nppicom: JPEG compression and decompression functions in nppi_compression_functions.h Removed starting in CUDA 11.0, use nvJPEG instead.

    • CUDA::nppicom

    • CUDA::nppicom_static

  • nppidei: Data exchange and initialization functions in nppi_data_exchange_and_initialization.h

    • CUDA::nppidei

    • CUDA::nppidei_static

  • nppif: Filtering and computer vision functions in nppi_filter_functions.h

    • CUDA::nppif

    • CUDA::nppif_static

  • nppig: Geometry transformation functions found in nppi_geometry_transforms.h

    • CUDA::nppig

    • CUDA::nppig_static

  • nppim: Morphological operation functions found in nppi_morphological_operations.h

    • CUDA::nppim

    • CUDA::nppim_static

  • nppist: Statistics and linear transform in nppi_statistics_functions.h and nppi_linear_transforms.h

    • CUDA::nppist

    • CUDA::nppist_static

  • nppisu: Memory support functions in nppi_support_functions.h

    • CUDA::nppisu

    • CUDA::nppisu_static

  • nppitc: Threshold and compare operation functions in nppi_threshold_and_compare_operations.h

    • CUDA::nppitc

    • CUDA::nppitc_static

  • npps:

    • CUDA::npps

    • CUDA::npps_static


The nvBLAS libraries. This is a shared library only.

Targets Created:

  • CUDA::nvblas


The nvGRAPH library. Removed starting in CUDA 11.0

Targets Created:

  • CUDA::nvgraph

  • CUDA::nvgraph_static


The nvJPEG library. Introduced in CUDA 10.

Targets Created:

  • CUDA::nvjpeg

  • CUDA::nvjpeg_static

nvPTX Compiler

New in version 3.25.

The nvPTX (PTX Compilation) library. The PTX Compiler APIs are a set of APIs which can be used to compile a PTX program into GPU assembly code. Introduced in CUDA 11.1 This is a static library only.

Targets Created:

  • CUDA::nvptxcompiler_static starting in CUDA 11.1


The nvRTC (Runtime Compilation) library.

Targets Created:

  • CUDA::nvrtc

New in version 3.26:

  • CUDA::nvrtc_builtins

  • CUDA::nvrtc_static starting in CUDA 11.5

  • CUDA::nvrtc_builtins_static starting in CUDA 11.5


The NVIDIA Management Library. This is a shared library only.

Targets Created:

  • CUDA::nvml


Deprecated since version 3.25: With CUDA 10.0+, use nvtx3.

The NVIDIA Tools Extension. This is a shared library only.

Targets Created:

  • CUDA::nvToolsExt


New in version 3.25.

The header-only NVIDIA Tools Extension Library. Introduced in CUDA 10.0.

Targets created:

  • CUDA::nvtx3


The NVIDIA OpenCL Library. This is a shared library only.

Targets Created:

  • CUDA::OpenCL


The cuLIBOS library is a backend thread abstraction layer library which is static only. The CUDA::cublas_static, CUDA::cusparse_static, CUDA::cufft_static, CUDA::curand_static, and (when implemented) NPP libraries all automatically have this dependency linked.

Target Created:

  • CUDA::culibos

Note: direct usage of this target by consumers should not be necessary.

Result variables


A boolean specifying whether or not the CUDA Toolkit was found.


The exact version of the CUDA Toolkit found (as reported by nvcc --version, version.txt, or version.json).


The major version of the CUDA Toolkit.


The minor version of the CUDA Toolkit.


The patch version of the CUDA Toolkit.


The path to the CUDA Toolkit library directory that contains the CUDA executable nvcc.


List of paths to all the CUDA Toolkit folders containing header files required to compile a project linking against CUDA.


The path to the CUDA Toolkit library directory that contains the CUDA Runtime library cudart.


New in version 3.18.

The path to the CUDA Toolkit directory containing the nvvm directory and either version.txt or version.json.


The path to the CUDA Toolkit directory including the target architecture when cross-compiling. When not cross-compiling this will be equivalent to the parent directory of CUDAToolkit_BIN_DIR.


The path to the NVIDIA CUDA compiler nvcc. Note that this path may not be the same as CMAKE_CUDA_COMPILER. nvcc must be found to determine the CUDA Toolkit version as well as determining other features of the Toolkit. This variable is set for the convenience of modules that depend on this one.