CUDA9.1 Visual Studio Integration 安裝失敗問題，如何解決？

最近更新微軟2018.4的更新，也更新過397.31的NVIDIA顯卡驅動，VS的cuda程序可編譯但是找不到CUDA，並報錯。重裝cuda後發現一直卡在Visual Studio Integration Fail。如何解決呢

---

1. 用 zip 打開，解壓到一個文件夾中
2. 打開setup.exe，執行 自定義安裝
3. 取消勾選 VS integration
4. 安裝完成後，找到 CUDAVisualStudioIntegrationextrasvisual_studio_integrationMSBuildExtensions 複製裡面的所有文件
5. 打開自己的 VS2017 的安裝目錄，比如 C:Program Files (x86)Microsoft Visual Studio2017CommunityCommon7IDEVCVCTargetsBuildCustomizations，把第四步中選中的文件複製進來即可

---

這個問題在Nvidia的devtalk論壇里討論過。

https://devtalk.nvidia.com/default/topic/1033111/cuda-setup-and-installation/cuda-9-1-cannot-install-due-to-failed-visual-studio-integration/?

devtalk.nvidia.com

https://devtalk.nvidia.com/default/topic/1027299/cuda-setup-and-installation/cuda-9-failed-to-support-the-latest-visual-studio-2017-version-15-5/post/5225807/#5225807?

devtalk.nvidia.com

當時管理員給出的解決方案有兩種：

1. 卸載當前的VS 2017。下載安裝15.4.5版本的VS 2017，再安裝CUDA 9。Installing an Earlier Release of Visual Studio 2017
2. 不卸載當前的VS 2017，安裝140工具集，之後所有CUDA項目都是用140工具集，而不是141工具集。

Using the latest VS2017 Community (15.6.6) and CUDA 9.1 on Windows 10 x64 with a GTX 1050 Ti. Upon following the instructions in this thread I was able to succesfully build all the samples.

Steps (nothing new, just confirming what was already indicated above):

- Install/Add "VC++ 2015.3 v140 toolset for desktop (x86,x64)" using the VS Installer;

- Install CUDA 9.1 Toolkit;

- Reboot;

- Open "Samples_vs2017.sln" in VS2017;

- For each project do update settings to Platform Toolset = "Visual Studio 2015 (v140)"; you cand select multiple projects;

- Rebuild all.

https://devtalk.nvidia.com/default/topic/1027299/cuda-setup-and-installation/cuda-9-failed-to-support-the-latest-visual-studio-2017-version-15-5/post/5251105/#5251105?

devtalk.nvidia.com

---

可以改環境變數試試

其他東西不用重灌 重點是CUDA要用自訂安裝

這裡有細節 :

http://markjong001.pixnet.net/blog?

markjong001.pixnet.net

---

推薦閱讀：