#  Cmake Cuda

**Repository Path**: denglinAI/cmake-cuda

## Basic Information

- **Project Name**:  Cmake Cuda
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-11-27
- **Last Updated**: 2025-11-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

**Now these repo is implemented base on cmake 3.25.3. Additionally, cmake 3.30 has some compatibility bug, please use cmake between cmake 3.25 and cmake 3.29.**

## how to use find_package(CUDA)
1. set dlgpu architecture, you can also add other compiler flags here
    ``` cmake
    set(CUDA_NVCC_FLAGS "--cuda-gpu-arch=dlgput64")
    # other compiler flags
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -DXXX -DYYY")
    ```

2. add current repo to `CMAKE_MODULE_PATH`. Additionally, you can use `-DCMAKE_MODULE_PATH="<current repo dir>" when typing cmake command
    ``` cmake
    list(APPEND CMAKE_MODULE_PATH "<current repo dir>")
    ```

3. import CUDA module
    ``` cmake
    find_package(CUDA REQUIRED)
    ```

4. if you want to use separate compilation, set CUDA_SEPARABLE_COMPILATION to ON
    ``` cmake
    set(CUDA_SEPARABLE_COMPILATION ON)
    ```

5. use cuda_add_executable, cuda_add_library to compile cu files

These are examples under `./example/test_find_package` and `./example/test_cuda*` directory, check it to see details.

## how to use enable_language(CUDA)
1. you must set dlgpu architecture first, otherwise there is an error in cmake configure phase
    ``` cmake
    set(CMAKE_CUDA_ARCHITECTURES dlgput64)
    ```

2. add other compiler flags
    ``` cmake
    set(CMAKE_CUDA_FLAGS "-DXXX -DYYY")
    ```

3. add current repo to `CMAKE_MODULE_PATH`. Additionally, you can use `-DCMAKE_MODULE_PATH="<current repo dir>" when typing cmake command
    ``` cmake
    list(APPEND CMAKE_MODULE_PATH "<current repo dir>")
    ```

4. use CUDA language
    ``` cmake
    enable_language(CUDA)
    ```

5. use add_executable, add_library to compile cu files
These are examples under `./example/test_enable_language*` directory, check it to see details.

## how to execute examples
- if you want to execute all examples, just `cd examples` and run `test.sh`
- if you want to try single example, you should `cd` corresponding directory, and run commnad blow:
``` cmake
cmake -B build -DCMAKE_MODULE_PATH="<current repo dir>"
```