Minimal cuda program(memory transfer only)

CMakeLists-cudamemcpy.txt

cmake_minimum_required(VERSION 2.8)

project(cudamemcpy)

include(FindCUDA)
include_directories(${CUDA_INCLUDE_DIRS})

set( SRCS cudamemcpy.cpp )

add_executable( cudamemcpy ${SRCS} )

target_link_libraries( cudamemcpy ${CUDA_LIBRARIES} )

cudamemcpy.cpp

#include <cuda_runtime.h>
#include <vector>
#include <iostream>
#include <algorithm>

bool checkCudaStatus( cudaError_t status ) {
    if( status != cudaSuccess ) {
        std::cerr << cudaGetErrorString( status ) << std::endl;
        return false;
    } else return true;
}
// perform cpu->gpu->cpu copy
int main( int argc, char** argv ) {
    const size_t ARRAY_SIZE = 0x10000;
    const double INIT_VALUE_IN  = 1.0;
    const double INIT_VALUE_OUT = 0.0;
    std::vector< double > host_array_in( ARRAY_SIZE, INIT_VALUE_IN );
    std::vector< double > host_array_out( ARRAY_SIZE, INIT_VALUE_OUT );
    double* dev_array = 0;
    cudaError_t status = cudaMalloc( &dev_array, ARRAY_SIZE * sizeof( double ) );
    if( !checkCudaStatus( status ) ) return -1;
    status = cudaMemcpy( dev_array, &host_array_in[ 0 ], ARRAY_SIZE * sizeof( double ), cudaMemcpyHostToDevice );
    if( !checkCudaStatus( status ) ) return -1;
    status = cudaMemcpy( &host_array_out[ 0 ], dev_array, ARRAY_SIZE * sizeof( double ), cudaMemcpyDeviceToHost );
    if( !checkCudaStatus( status ) ) return -1;
    std::cout << std::boolalpha
              << std::equal( host_array_in.begin(), host_array_in.end(), host_array_out.begin() )
              << std::endl;
    status = cudaFree( dev_array );
    if( !checkCudaStatus( status ) ) return -1;
    return 0;
}