QtCreator + CUDA-5.5

[danioto]

Hello everyone!

I'm facing now the problem with integrating CUDA-5.5 with QtCreator on Ubuntu 12.04. Steps I followed:
1. I installed CUDA-5.5 with samples, then builded them and run correctly.
2. I added to .bashrc file:
export PATH=$PATH:/usr/local/cuda-5.5/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-5.5/lib64
3. I made a program, which looks like follows:
cuda_main.cu:

#include <stdio.h>
#include <cuda_runtime.h>
 
__global__ void cube(float * d_out, float * d_in){
   int index=threadIdx.x;
   float f=d_in[index];
   d_out[index]=f*f*f;
}
 
extern "C"
int cuda_main()
{
   const int ARRAY_SIZE = 96;
   const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float);
 
   // generate the input array on the host
   float h_in[ARRAY_SIZE];
   for (int i = 0; i < ARRAY_SIZE; i++) {
       h_in[i] = float(i);
   }
   float h_out[ARRAY_SIZE];
 
   // declare GPU memory pointers
   float * d_in;
   float * d_out;
 
   // allocate GPU memory
   cudaMalloc((void**) &d_in, ARRAY_BYTES);
   cudaMalloc((void**) &d_out, ARRAY_BYTES);
 
   // transfer the array to the GPU
   cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice);
 
   // launch the kernel
   cube<<<1, ARRAY_SIZE>>>(d_out, d_in);
 
   // copy back the result array to the CPU
   cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost);
 
   // print out the resulting array
   for (int i =0; i < ARRAY_SIZE; i++) {
       printf("%f", h_out[i]);
       printf(((i % 4) != 3) ? "\t" : "\n");
   }
 
   cudaFree(d_in);
   cudaFree(d_out);
 
   return 0;
}
 
int main(int argc, char *argv[])
{
    cuda_main();
    return 0;
}

#include <stdio.h>
#include <cuda_runtime.h>

__global__ void cube(float * d_out, float * d_in){
   int index=threadIdx.x;
   float f=d_in[index];
   d_out[index]=f*f*f;
}

extern "C"
int cuda_main()
{
   const int ARRAY_SIZE = 96;
   const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float);

   // generate the input array on the host
   float h_in[ARRAY_SIZE];
   for (int i = 0; i < ARRAY_SIZE; i++) {
       h_in[i] = float(i);
   }
   float h_out[ARRAY_SIZE];

   // declare GPU memory pointers
   float * d_in;
   float * d_out;

   // allocate GPU memory
   cudaMalloc((void**) &d_in, ARRAY_BYTES);
   cudaMalloc((void**) &d_out, ARRAY_BYTES);

   // transfer the array to the GPU
   cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice);

   // launch the kernel
   cube<<<1, ARRAY_SIZE>>>(d_out, d_in);

   // copy back the result array to the CPU
   cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost);

   // print out the resulting array
   for (int i =0; i < ARRAY_SIZE; i++) {
       printf("%f", h_out[i]);
       printf(((i % 4) != 3) ? "\t" : "\n");
   }

   cudaFree(d_in);
   cudaFree(d_out);

   return 0;
}

int main(int argc, char *argv[])
{
    cuda_main();
    return 0;
}

To copy to clipboard, switch view to plain text mode 

*.pro file:

CONFIG += console
CONFIG -= qt
TEMPLATE = app
SOURCES += cuda_main.cu
SOURCES -= cuda_main.cu
CUDA_SOURCES += cuda_main.cu
 
# project build directories
DESTDIR = $$system(pwd)
OBJECTS_DIR = $$DESTDIR/Obj
 
# C++ flags
QMAKE_CXXFLAGS_RELEASE = -O3
 
# Path to cuda toolkit install
CUDA_DIR = /usr/local/cuda-5.5
 
# Path to header and libs files
INCLUDEPATH += $$CUDA_DIR/include
QMAKE_LIBDIR += $$CUDA_DIR/lib64        # Note I'm using a 64 bits Operating system
 
# libs used in your code
LIBS += -lcudart \
        -lcuda
 
# GPU architecture
CUDA_ARCH = sm_12                       # I've a old device. Adjust with your compute capability
 
# Here are some NVCC flags I've always used by default.
NVCCFLAGS = --compiler-options -fno-strict-aliasing -use_fast_math --ptxas-options=-v
 
# Prepare the extra compiler configuration (taken from the nvidia forum - i'm not an expert in this part)
CUDA_INC = $$join(INCLUDEPATH,' -I','-I',' ')
cuda.commands = $$CUDA_DIR/bin/nvcc -m64 -O3 -arch=$$CUDA_ARCH -c $$NVCCFLAGS \
                $$CUDA_INC $$LIBS ${QMAKE_FILE_NAME} -o ${QMAKE_FILE_OUT} 2>&1 \
                | sed -r \"s/\\(([0-9]+)\\)/:\\1/g\" 1>&2
 
cuda.dependency_type = TYPE_C
cuda.depend_command = $$CUDA_DIR/bin/nvcc -O3 -M $$CUDA_INC $$NVCCFLAGS ${QMAKE_FILE_NAME}
cuda.input = CUDA_SOURCES
cuda.output = ${OBJECTS_DIR}${QMAKE_FILE_BASE}_cuda.o
 
# Tell Qt that we want add more stuff to the Makefile
QMAKE_EXTRA_COMPILERS += cuda

CONFIG += console
CONFIG -= qt
TEMPLATE = app
SOURCES += cuda_main.cu
SOURCES -= cuda_main.cu
CUDA_SOURCES += cuda_main.cu

# project build directories
DESTDIR = $$system(pwd)
OBJECTS_DIR = $$DESTDIR/Obj

# C++ flags
QMAKE_CXXFLAGS_RELEASE = -O3

# Path to cuda toolkit install
CUDA_DIR = /usr/local/cuda-5.5

# Path to header and libs files
INCLUDEPATH += $$CUDA_DIR/include
QMAKE_LIBDIR += $$CUDA_DIR/lib64        # Note I'm using a 64 bits Operating system

# libs used in your code
LIBS += -lcudart \
        -lcuda

# GPU architecture
CUDA_ARCH = sm_12                       # I've a old device. Adjust with your compute capability

# Here are some NVCC flags I've always used by default.
NVCCFLAGS = --compiler-options -fno-strict-aliasing -use_fast_math --ptxas-options=-v

# Prepare the extra compiler configuration (taken from the nvidia forum - i'm not an expert in this part)
CUDA_INC = $$join(INCLUDEPATH,' -I','-I',' ')
cuda.commands = $$CUDA_DIR/bin/nvcc -m64 -O3 -arch=$$CUDA_ARCH -c $$NVCCFLAGS \
                $$CUDA_INC $$LIBS ${QMAKE_FILE_NAME} -o ${QMAKE_FILE_OUT} 2>&1 \
                | sed -r \"s/\\(([0-9]+)\\)/:\\1/g\" 1>&2

cuda.dependency_type = TYPE_C
cuda.depend_command = $$CUDA_DIR/bin/nvcc -O3 -M $$CUDA_INC $$NVCCFLAGS ${QMAKE_FILE_NAME}
cuda.input = CUDA_SOURCES
cuda.output = ${OBJECTS_DIR}${QMAKE_FILE_BASE}_cuda.o

# Tell Qt that we want add more stuff to the Makefile
QMAKE_EXTRA_COMPILERS += cuda

To copy to clipboard, switch view to plain text mode 

4. Application builded correctly, but when I tried to run it from terminal i got message:
./CUDATry: error while loading shared libraries: libcudart.so.5.5: cannot open shared object file: No such file or directory
That's what ldd ./CUDATry return:

linux-vdso.so.1 =>  (0x00007fffb9dff000)
        libcudart.so.5.5 => /usr/local/cuda-5.5/lib64/libcudart.so.5.5 (0x00007fbfd3eca000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fbfd3ae3000)
        libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fbfd38de000)
        libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fbfd35de000)
        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fbfd33c1000)
        librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fbfd31b8000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fbfd4119000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fbfd2ebc000)
        libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fbfd2ca6000)

linux-vdso.so.1 =>  (0x00007fffb9dff000)
        libcudart.so.5.5 => /usr/local/cuda-5.5/lib64/libcudart.so.5.5 (0x00007fbfd3eca000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fbfd3ae3000)
        libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fbfd38de000)
        libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fbfd35de000)
        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fbfd33c1000)
        librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fbfd31b8000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fbfd4119000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fbfd2ebc000)
        libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fbfd2ca6000)

To copy to clipboard, switch view to plain text mode 

So it looks like libcudart.so.5.5 was included properly... I dont know what to do, what do you think about it?

[syamcr]

A wild guess:
Try this:
export LD_LIBRARY_PATH=/usr/local/cuda-5.5/lib64
./CUDATry

If that works, you may have to add the library path by creating/editing: /etc/ld.so.conf.d/cuda.conf
And then run ldconfig once to update.

[danioto]

I forgot to show my .bashrc file:

######################JAVA############################
export JAVA_HOME=/usr/lib/jvm/java-6-openjdk/
######################################################
 
 
##########Zmienne Srodowiskowe CUDA###################
export PATH=$PATH:/usr/local/cuda-5.5/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-5.5/lib64:/usr/local/cuda-5.5/lib
######################################################
 
 
##########Zmienne Srodowiskowe ROS####################
source /opt/ros/groovy/setup.bash
source ~/Programowanie/ROS/Groovy/catkin_workspace/devel/setup.bash
export ROS_PACKAGE_PATH=/opt/ros/groovy/share:/opt/ros/groovy/stacks:~/Programowanie/ROS/Groovy/catkin_workspace/src
export ROS_MASTER_URI=http://monster:11311
######################################################

######################JAVA############################
export JAVA_HOME=/usr/lib/jvm/java-6-openjdk/
######################################################


##########Zmienne Srodowiskowe CUDA###################
export PATH=$PATH:/usr/local/cuda-5.5/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-5.5/lib64:/usr/local/cuda-5.5/lib
######################################################


##########Zmienne Srodowiskowe ROS####################
source /opt/ros/groovy/setup.bash
source ~/Programowanie/ROS/Groovy/catkin_workspace/devel/setup.bash
export ROS_PACKAGE_PATH=/opt/ros/groovy/share:/opt/ros/groovy/stacks:~/Programowanie/ROS/Groovy/catkin_workspace/src
export ROS_MASTER_URI=http://monster:11311
######################################################

To copy to clipboard, switch view to plain text mode 

It should be enought, right?

[danioto]

Don't anybody have a single clue?

[danioto]

Ok, problem fixed. It turns out, LD_LIBRARY_PATH in isn't the best idea. After removing it from .bashrc and adding new line in file /etc/ld.so.conf:

/usr/local/cuda-5.5/lib64

/usr/local/cuda-5.5/lib64

To copy to clipboard, switch view to plain text mode 

and run:

sudo ldconfig

sudo ldconfig

To copy to clipboard, switch view to plain text mode 

everything works fine. I hope it will help someone.