[PATCH] D55269: [CUDA][OpenMP] Fix nvidia-cuda-toolkit detection on Debian/Ubuntu

Tue Dec 4 12:28:44 PST 2018

jdenny added a comment.

@tra, @Hahnfeld: Thanks for your replies.

In D55269#1318901 <https://reviews.llvm.org/D55269#1318901>, @tra wrote:

> I'm not sure that's something that needs to be fixed in clang.
>
> IIUIC, Debian has added a shim that pretends to be a monolithic CUDA install:
>  https://bugs.launchpad.net/ubuntu/+source/clang/+bug/1706326
>  That change seems to be in Ubuntu bionic (18.04) https://packages.ubuntu.com/en/bionic/nvidia-cuda-toolkit

apt confirms that's what I have: nvidia-cuda-toolkit 9.1.85-3ubuntu1

> With that fix in place --cuda-path=/usr/lib/cuda should work.

Seems to.  To be clear, I'm trying to address the use case where cmake/clang finds the cuda installation automatically.

> --cuda-path=/usr was never supposed to work -- /usr is *not* the root of the CUDA SDK.

/usr/lib/cuda/bin/nvcc doesn't exist, so that's probably why FindCUDA.cmake finds /usr/bin/nvcc (also installed by nvidia-cuda-toolkit).  Is it fair then to say that /usr/lib/cuda isn't the root either?

> I guess that just adding the check for isUbuntu() should make clang work on Ubuntu 18.04+.

It fixes the first issue I reported.  It does not fix the second.

It seems that nvidia-cuda-toolkit still isn't installing a complete CUDA install in one location.  Even if it installed it all to /usr/lib/cuda, FindCUDA.cmake would probably still see /usr/bin/nvcc and assume /usr is the CUDA install root.

What's the path forward?

Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55269/new/

https://reviews.llvm.org/D55269