Cuda compilation time issue nvcc 10x slower than cl to parse boost/shared_ptr.hpp


Am I the only developer losing too much time waiting for nvcc?

For example nvcc takes 11s on my computer (quad-core, 4GB of RAM, windows 7)

to compile the following piece of code, whereas cl (visual studio 2010)

takes only 1s:

# include <boost/shared_ptr.hpp>

Is there any tips to speed up c++ compilation with nvcc on windows?

I checked nvcc’s docs, it doesn’t say anything about precompiled

header support.