I have a C++
static library project that compiles without any errors, nor it shows any errors in the Error list
window. So all seems to be working OK. Except that when I try to open the source (CTRL
+ left click on a library or library specific item), then the file of an incorrect version is opened. Why?
To elaborate. I have CUDA 9.1
, CUDA 9.2
and CUDA 10.2
installed on my PC
. My project has following CUDA Linker
config:
Additional Libary Directories: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2\lib\x64;%(AdditionalLibraryDirectories)
Additional Dependencies: cudart.lib;%(AdditionalDependencies)
In the CUDA linker
's Command line
I have:
"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2\bin\nvcc.exe" -dlink -o x64\Debug\Core3.device-link.obj -Xcompiler "/EHsc /nologo /Zi " -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2\lib\x64" cudart.lib
In the CUDA C/C++
's Command line
I have:
So, it looks OK. But in the code, when I click on any of the includes - they point to 9.2
source. Example:
#include <cuda.h>
#include "cuda_runtime.h"
#include <device_launch_parameters.h>
the path is: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\ v9.2 \include
Furthermore I have checked the environment variables. All good. Also checked which props
files are used in .vcxproj
file. Properties CUDA 10.2.props
are used everywhere.
So my question is - why does the IntelliSence
still thinks that I am using an older version of CUDA
? And more importantly - how do I make him use a proper, 10.2
version?
EDIT 1: more information requested in comments. Include directories - defaults used.