Hellooo !
I’m trying to compile cuda 2.0 code with nvcc using gcc (Mingw) on windows xp.
nvcc 's -foreign option (for test only as reference guide says) crash when I tried to compile with it.
As alex_dubinsky advised me here, I compiled my .cu on Linux with dryrun option which gave me the different steps of the linux compilation.
It’s a succession of gcc and cudafe calls ( + nvopencc, ptxas… calls).
First cmd is a gcc call which write a tmpxft_0000130c_00000000-4_cuda_test.cpp1.ii file. It works fine.
Second one is a cudafe call :
cudafe --m32 --diag_error=host_device_limited_call -tused “-IL:\CUDA\include” “-IL:\CUDA\include\cudart” --gen_c_file_name “tmpxft_0000130c_00000000-1_cuda_test.cudafe1.c” --stub_file_name “tmpxft_0000130c_00000000-1_cuda_test.cudafe1.stub.c” --stub_header_file_name “tmpxft_0000130c_00000000-1_cuda_test.cudafe1.stub.h” --gen_device_file_name “tmpxft_0000130c_00000000-1_cuda_test.cudafe1.gpu” --include_file_name tmpxft_0000130c_00000000-3_cuda_test.fatbin.c “tmpxft_0000130c_00000000-4_cuda_test.cpp1.ii”
This cmd crash !
The problem could be a wrong option but I didn’t find any doc about cudafe and its options (btw, “-h”, “–help” lead to a “catastrophic error” ; “?” make cudafe crash :blink: ).
I try to remove --m32, --diag_error=host_device_limited_call, -tused → same problem.
Can anybody help me ? :rolleyes:
is there a doc swh ?
By google-izing, I find this method. I’m gonna try.
Thanks for your help !