<div dir="ltr"><div dir="ltr">Thx for the info. <br></div><div dir="ltr"><div><br></div><div>Since CXX and CUDA are defined together in the project() command, I don't see a way to inject code to use the CXX compiler if no CUDAHOSTCXX or CMAKE_CUDA_HOST_COMPILER is given without replicating the CXX search. I tried moving CUDA out to an enable_language(CUDA) call instead so I could set those variables between project(foo LANGUAGES CXX) and enable_language(CUDA), but I'm seeing other issues with that approach. </div><div><br></div><div>Any other suggestions? If not, we'll just wrap our cmake invocation in a script to help setup the environment properly for now.</div><div><br></div><div>Thx!</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Apr 16, 2019 at 1:10 PM Robert Maynard <<a href="mailto:robert.maynard@kitware.com">robert.maynard@kitware.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">The default implementation is to defer to CUDA for selecting what<br>ever host compiler it would like. To make sure that CMake uses the<br>same CXX and CUDACXX compiler you will need to explicitly state that<br>either through the CUDAHOSTCXX env variable (<br>
<a href="https://cmake.org/cmake/help/v3.12/envvar/CUDAHOSTCXX.html" rel="noreferrer" target="_blank">https://cmake.org/cmake/help/v3.12/envvar/CUDAHOSTCXX.html</a> ) or with<br>CMAKE_CUDA_HOST_COMPILER on the initial configuration of a project.<br>
</blockquote></div></div>