I am trying to build my C++ application that uses caffe, in Debug Mode, VS2013 community, x64. To be able to build version that do not need cuda to run, I added to wrapped each .cu file as indicated below:
#ifndef CPU_ONLY
// .cu file contents
#endif
The project was built and ran fine in CPU_ONLY mode. Undefininig the CPU_ONLY flag, the project builds and runs OK in Release mode, but in Debug, I am getting the following error when trying to compile the *.cu files:
Compiling CUDA source file ..\..\src\caffe\layers\base_data_layer.cu...
>
> >COMMAND
> nvcc fatal : Stray '"' character in command line
COMMAND exited with code 1.
Where COMMAND is the nvcc compiler call command below, newlined for readability.
"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin\nvcc.exe"
-gencode=arch=compute_30,code=\" sm_30,compute_30\"
--use-local-env
--cl-version 2013 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\bin\x86_amd64"
-I"C:\Users\username\Downloads\liblinear-1.8\liblinear-1.8"
-I"C:\Users\username\Downloads\poco-1.6.0\Foundation\include"
-I"C:\Users\username\Downloads\poco-1.6.0\Net\include"
-IC:\opencv_gpu\include -I"C:\Users\username\Downloads\caffe-master\src"
-I"C:\Users\username\Downloads\caffe-master\include"
-IC:\local\boost_1_56_0 -I"C:\Users\username\Downloads\caffe-master\3rdparty\include\openblas"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\lmdb"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\leveldb"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\hdf5"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\google"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\glog"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include\gflags"
-I"C:\Users\username\Downloads\caffe-master\3rdparty\include"
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include"
-I"C:\Users\username\Downloads\cudnn-6.5-win-R1"
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include"
-G --keep-dir x64\Debug -maxrregcount=0 --machine 64 --compile
-cudart static -g -D_SCL_SECURE_NO_WARNINGS -D_CRT_SECURE_NO_WARNINGS -DWIN32 -D_DEBUG -D_CONSOLE -D_LIB -D_UNICODE -DUNICODE
-Xcompiler "/EHsc /W0 /nologo /Od /Zi /RTC1 /MDd "
-o x64\Debug\base_data_layer.cu.obj "C:\Users\username\Downloads\caffe-master\src\caffe\layers\base_data_layer.cu"
The project was able to build successfully in debug mode before adding the CPU_ONLY flags. Any ideas?
Turns out it was a typo.
In project properties->Debug->CUDA C/C++->Device, instead of
compute_30,sm_30
I had
`compute_30, sm_30`
that is, with a space separator.