When will CUDA support libc++ on Mavericks?

I am aware of the fact that the latest CUDA (6.0) does not yet support libc++ on Mac OS X Mavericks. However, since libc++ has become the default when installing new libraries (through, say, MacPorts), I am wondering if CUDA will be able to support libc++ in the near future (maybe in 6.5?) Thank you.

I agree.

Qt5 compile with libc++. So right now it’s more or less not possible to compile with Qt 5 and CUDA easily…

Seems the issue is not resolved with Cuda 6.5 unfortunately.

Nvidia developers, we have to maintain a Qt4 backward compatibility just for that, and it’s really slow it down. And we’re certainly not the only one, first google search leads to this page : https://github.com/BVLC/caffe/issues/405

So please consider this as a high priority !

I am using CUDA 6.0 on OS X 10.9 together with Qt 5 (using libstdc++ for the CUDA part and libc++ for the Qt part). It is really unfortunate that this means I can not pass any objects containing STL stuff between CUDA code and the rest of the application.

Actually scrap that. I haven’t worked on the project in a while. It turns out I am using libstdc++ for the whole project, but a mix should still be possible.

See also:


The point remains. We need libc++ support. Even using libstdc++ for everything has issues (see first link above).