[PATCH] D54608: [CUDA] updated CompileCudaWithLLVM.rst

[Justin Lebar via Phabricator via llvm-commits]

jlebar added inline comments.


================
Comment at: llvm/docs/CompileCudaWithLLVM.rst:27
+7.0 through 9.2. If you need support for CUDA 10, you will need to use clang
+built from r342924 or newer.
 
----------------
tra wrote:
> jlebar wrote:
> > Is there a way to query clang (or its source code) to see what versions it supports, so that people don't have to ping us manually?  Like, if we write the documentation saying "current release of clang", are we going to update it for every new release?
> There's no convenient way at the moment. The easiest way for the user right now is to check  [[https://github.com/llvm-mirror/clang/blob/master/lib/Headers/__clang_cuda_runtime_wrapper.h#L65 | __clang_cuda_runtime_wrapper.h]] 
> 
> We could enhance CUDA installation detector to issue a warning if it finds and unsupported CUDA version and list supported versions there.
> 
> 
> 
> 
Sure, I'm wondering if e.g. we shouldn't point users at __clang_cuda_runtime_wrapper.h?

> We could enhance CUDA installation detector to issue a warning if it finds and unsupported CUDA version and list supported versions there.

Something like that might be nice.


Repository:
  rL LLVM

https://reviews.llvm.org/D54608