cuda build_error ai_generated true

nvcc 致命错误：在 PATH 中找不到编译器 'cl.exe'

nvcc fatal : Cannot find compiler 'cl.exe' in PATH

ID: cuda/nvcc-include-path-not-found

其他格式: JSON · Markdown 中文 · English
85%修复率
90%置信度
1证据数
2023-04-10首次发现

版本兼容性

版本状态引入弃用备注
CUDA 12.0 active — — —
CUDA 11.8 active — — —
Visual Studio 2022 active — — —
Visual Studio 2019 active — — —

根因分析

在 Windows 上，nvcc 需要在 PATH 中包含 Microsoft Visual C++ 编译器 (cl.exe)，但构建环境缺少必要的 Visual Studio 工具或未配置正确的架构（x64 与 x86）。

English

On Windows, nvcc requires the Microsoft Visual C++ compiler (cl.exe) to be in the PATH, but the build environment is missing the necessary Visual Studio tools or the correct architecture (x64 vs x86) is not configured.

generic

官方文档

https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#compiler

解决方案

  1. 从开始菜单打开“x64 Native Tools Command Prompt for VS 2022”（或 VS 2019），然后在该提示符下运行 nvcc 命令。这会自动设置所有必需的环境变量。
  2. 如果使用 CMake 等构建系统，设置 `CMAKE_GENERATOR` 为 'Visual Studio 17 2022' 并指定 `-A x64` 以确保使用正确的工具链。
  3. 手动将 Visual C++ 编译器目录添加到 PATH：`set PATH=C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.34.31933\bin\Hostx64\x64;%PATH%`（调整版本号）。同时根据 VS 开发者提示符设置 INCLUDE 和 LIB。

无效尝试

常见但无效的做法:

  1. Reinstalling CUDA Toolkit without Visual Studio 95% 失败

    The error is about the missing Visual Studio compiler, not CUDA itself; reinstalling CUDA does not add cl.exe to PATH.

  2. Setting PATH to a random directory containing cl.exe from a different architecture (e.g., x86 instead of x64) 80% 失败

    nvcc requires the x64 native tools; using x86 cl.exe causes link errors or wrong object file format.

  3. Installing only the 'Desktop development with C++' workload but not launching from the correct developer command prompt 75% 失败

    The environment variables (INCLUDE, LIB, PATH) are only set in the Visual Studio developer command prompt; a regular cmd or PowerShell does not inherit them.