Support building CUDA code with clang
Submitted by Justin Lebar
Assigned to Justin Lebar
Link to original bugzilla bug (#1173)
Description
I'm one of the developers working on CUDA support in clang. I have patches that let me build and test eigen's cuda code compiled with clang, rather than nvcc. With these changes, plus some llvm changes that are still under review, we can compile and run all of eigen's cuda tests successfully. (They are still quite slow; some 100x worse than nvcc.)
Patches in a moment.
Edited by Eigen Bugzilla