Open vucinicv opened 8 years ago
This could ultimately be due to the fact that we're forced to use a relatively old version of Xcode in order to be able to build with nvcc. A newer CUDA toolkit might cure this problem.
One possible WAR for the thread-local issue might be to use the __thread
attribute under clang instead.
According to http://stackoverflow.com/a/25393790, it's supported. But I can't find it in the clang documentation. One possible solution is to just check if it builds.
Tagging @chuckseberino as he's run into this recently as well.
Using __thread works for me on OSX 10.10.5, XCode 7.2.1
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1 Apple LLVM version 7.0.2 (clang-700.1.81) Target: x86_64-apple-darwin14.5.0
thread_local storage is unsupported on some versions(Mavericks at least) of mac.