0

I am trying to do some coding on GPU, and I have visual studio and intel composer installed, when I start a CUDA Project, wrote the code then compile, everything will be fine if I ask MS's compiler to compile the host code, however if I ask intel's compiler to do that, it will always fail, is there anyway to let NVCC work with intel's C++ compiler (v 13.0)?

user2003564
  • 213
  • 2
  • 9
  • 1
    [Currently, Intel compilers are not supported by nvcc.](http://stackoverflow.com/questions/9433722/how-to-use-intel-c-compiler-with-cuda-nvcc) – sgarizvi Feb 11 '13 at 12:21
  • @sgar91 That's too bad, I am trying to write a library which include both cuda subroutines and Intel's optimized subroutines, it seems that the two cannot be packed together then. – user2003564 Feb 11 '13 at 12:24
  • Try this: Compile each file separately. cu files with `nvcc` and other files with `icl`. Then link all the object files which are created as the output of compilation. – sgarizvi Feb 11 '13 at 12:27
  • Sure, but I cannot use the optimized routines freely on the host part this way like I thought, anyway thanks. – user2003564 Feb 11 '13 at 12:34

0 Answers0