Compile and link multiple files in CUDA

Question

I have encountered a problem in Visual Studio while using the CUDA toolkit. I want to write my code in multiple .cu files. However, It always gives me a building failure. I did a lot of search on it but I cannot find a solution. Here are what I tried:

Make sure the Item type is CUDA C/C++ not C++
Set general relocatable code rcd+=true(Which is the solution for most posts)
Change the host runtime library to multi-threaded debug(/Mtd)

But none of them works, I got the same error.

My endpoint is to create a matrix class that can be used on both host and device. To make the problem simple, I only keep the construct and print function in the class(But has the same error). Here is the header of the matrix:

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <iostream>
#ifndef __alldev__
#define __alldev__ __host__ __device__
#endif
template <class T>
class Matrix {
    T* m = NULL;
    T* dev_m = NULL;
public:
    int rowNum = 0;
    int colNum = 0;
    Matrix(T* data, int rowNum, int colNum, int location = 0);
    void print();
    __alldev__ int ind(int i, int j) {
        return (i - 1) + (j - 1)*colNum;
    }
};

Here is the .cu file of the matrix:

#include "Matrix.cuh"
template <class T>
Matrix<T>::Matrix(T* data, int rowNum, int colNum, int location = 0) {
    this->rowNum = rowNum;
    this->colNum = colNum;
    m = new T[rowNum*colNum];
    for (int i = 1; i <= rowNum; i++) {
        for (int j = 1; j <= colNum; j++) {
            m[ind(i, j)] = data[ind(i, j)];
        }
    }
}

template <class T>
void Matrix<T>::print() {
    for (int i = 1; i <= rowNum; i++) {
        for (int j = 1; j <= colNum; j++) {
            std::cout << m[ind(i, j)] << "  ";
        }
        std::cout << std::endl;
    }
}

The main file:

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <stdio.h>
#include "Matrix.cuh"
int main()
{
    int N = 10;
    float * data = new float[N];
    for (int i = 0; i < N; i++) {
        data[i] = i;
    }
    Matrix<float> test(data,N,1);
    test.print();
}

The above code works fine when they are in the same file, but does not work when in different files. The most strange thing is that if I comment out the kernel function call, the error message will be changed. I totally have no clue of it.

Here are the normal level output from VS:

1>------ Build started: Project: test1, Configuration: Debug x64 ------
1>Build started 8/31/2018 7:27:07 PM.
1>Target InitializeBuildStatus:
1>  Creating "x64\Debug\test1.tlog\unsuccessfulbuild" because "AlwaysCreate" was specified.
1>Target AddCudaCompileDeps:
1>  Skipping target "AddCudaCompileDeps" because all output files are up-to-date with respect to the input files.
1>Target AddCudaCompileDeps:
1>  Skipping target "AddCudaCompileDeps" because all output files are up-to-date with respect to the input files.
1>Target AddCudaCompilePropsDeps:
1>  Skipping target "AddCudaCompilePropsDeps" because all output files are up-to-date with respect to the input files.
1>Target AddCudaCompilePropsDeps:
1>  Skipping target "AddCudaCompilePropsDeps" because all output files are up-to-date with respect to the input files.
1>Target AddCudaCompilePropsDeps:
1>  Skipping target "AddCudaCompilePropsDeps" because all output files are up-to-date with respect to the input files.
1>Target CudaBuild:
1>  Target CudaBuildCore:
1>    Compiling CUDA source file kernel.cu...
1>    Target CudaBuildCore:
1>      Skipping target "CudaBuildCore" because all output files are up-to-date with respect to the input files.
1>    Target CudaBuildCore:
1>      Skipping target "CudaBuildCore" because all output files are up-to-date with respect to the input files.
1>
1>    C:\Users\wangj\documents\visual studio 2017\Projects\test1\test1>"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\bin\nvcc.exe" -gencode=arch=compute_30,code=\"sm_30,compute_30\" --use-local-env --cl-version 2017 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.10.25017\bin\HostX86\x64" -x cu -rdc=true -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\include"  -G   --keep-dir x64\Debug -maxrregcount=0  --machine 64 --compile -cudart static  -g   -DWIN32 -DWIN64 -D_DEBUG -D_CONSOLE -D_MBCS -Xcompiler "/EHsc /W3 /nologo /Od /FS /Zi /RTC1 /MDd " -o x64\Debug\kernel.cu.obj "C:\Users\wangj\documents\visual studio 2017\Projects\test1\test1\kernel.cu"
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(838): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(1772): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(2628): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(3477): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(4417): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(5319): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(6229): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7104): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7914): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h(776): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h(1636): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\device_double_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\sm_20_intrinsics.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\sm_20_intrinsics.h(925): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(838): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(1772): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(2628): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(3477): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(4417): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(5319): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(6229): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7104): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7914): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h(776): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt/device_functions.h(1636): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\device_double_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\sm_20_intrinsics.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\sm_20_intrinsics.h(925): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    kernel.cu
1>    C:/Users/wangj/documents/visual studio 2017/Projects/test1/test1/kernel.cu(16): warning C4244: '=': conversion from 'int' to 'float', possible loss of data
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h : warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(838): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(1772): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(2628): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(3477): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(4417): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(5319): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(6229): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7104): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>    c:\program files\nvidia gpu computing toolkit\cuda\v9.1\include\crt\math_functions.h(7914): warning C4819: The file contains a character that cannot be represented in the current code page (936). Save the file in Unicode format to prevent data loss
1>  Done building target "CudaBuildCore" in project "test1.vcxproj".
1>
1>  Done building project "test1.vcxproj".
1>Target CudaLink:
1>  C:\Users\wangj\documents\visual studio 2017\Projects\test1\test1>"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\bin\nvcc.exe" -dlink -o x64\Debug\test1.device-link.obj -Xcompiler "/EHsc /W3 /nologo /Od /Zi /RTC1 /MDd " -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\lib\x64" cudart.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib  -gencode=arch=compute_30,code=sm_30 -G --machine 64 x64\Debug\kernel.cu.obj x64\Debug\Matrix.cu.obj x64\Debug\Matrix.cuh.obj
1>  cudart.lib
1>  kernel32.lib
1>  user32.lib
1>  gdi32.lib
1>  winspool.lib
1>  comdlg32.lib
1>  advapi32.lib
1>  shell32.lib
1>  ole32.lib
1>  oleaut32.lib
1>  uuid.lib
1>  odbc32.lib
1>  odbccp32.lib
1>  kernel.cu.obj
1>  Matrix.cu.obj
1>  Matrix.cuh.obj
1>Target Link:
1>  kernel.cu.obj : error LNK2019: unresolved external symbol "public: __cdecl Matrix<float>::Matrix<float>(float *,int,int,int)" (??0?$Matrix@M@@QEAA@PEAMHHH@Z) referenced in function main
1>  kernel.cu.obj : error LNK2019: unresolved external symbol "public: void __cdecl Matrix<float>::print(void)" (?print@?$Matrix@M@@QEAAXXZ) referenced in function main
1>  C:\Users\wangj\documents\visual studio 2017\Projects\test1\x64\Debug\test1.exe : fatal error LNK1120: 2 unresolved externals
1>Done building target "Link" in project "test1.vcxproj" -- FAILED.
1>
1>Done building project "test1.vcxproj" -- FAILED.
1>
1>Build FAILED.
1>
1>Time Elapsed 00:00:06.70
========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========

I will really appreciate it if any one can me any suggestion.

I don't see how this could work under any circumstance. Where are `Matrix::set()`, `Matrix::get` , `Matrix::HostToDevice()` and `Matrix::deviceToHost()` implemented? They are not even in your class definition. — Robert Crovella, Aug 31 '18 at 23:19
@Robert Crovella Thanks for your comment, I forgot to remove them in the main function when I'm trying to make a simple example. They were defined in my project. I will update my question. — Jeff, Aug 31 '18 at 23:24
chopping down and posting an example that you have not actually tested is oftentimes not very useful. The above is one example why. If you want help, I suggest you actually test what you post, and post the output from that test, not some other test, in your question. you might want to read about [mcve] — Robert Crovella, Aug 31 '18 at 23:28
@Robert Crovella Thanks for the information, I did write the sample code like the link, but I copy the main function from my project to this post by mistake (I opened two VS when I'm writing the post), so that's why there are some missing functions. That's my fault. The sample code behaves exactly as I described. I have updated the output (the Previous one is also from my project) — Jeff, Aug 31 '18 at 23:35
note that if you have the `Matrix` template implementation in a separate file by itself, there is a good chance you are going to run into a template instantiation issue at link time. This isn't unique or specific to CUDA. See [here](https://stackoverflow.com/questions/2351148/explicit-instantiation-when-is-it-used). With the output I see now, this appears to be the case. You should explicitly instantiate your `Matrix` class for type `float`, in `Matrix.cu` At the bottom of your `Matrix.cu` file, put this: `template class Matrix;` — Robert Crovella, Aug 31 '18 at 23:39
@Robert Crovella It works! I really really appreciate your help! I cannot solve it without your tips. This problem is in my blind spot. Would you mind to write a short answer? I would select you as the correct answer. Thanks! — Jeff, Aug 31 '18 at 23:58

Compile and link multiple files in CUDA

0 Answers0