请教各位一下，在linux下如何将.cpp和.cu文件一起编译啊

菜鸟-xl 2018-04-24 03:24:20

各位好，是这样的，我写了五个文件，分别是.cpp和.cu的，但是不知道如何一起编译，上网找了好长时间也没有找到解决方法，因为我在Linux下运行，没有图形界面，不能在VS上编译，实在没辙了，只好请教一下各位



#include <stdio.h>

#include <iostream>

#include <iomanip>

#include <math.h>

#include "GPUTSPSolver.h"

using namespace std;

CGPUTSPSolver solverGPU;

int main(int argc, char **argv)

{

	solverGPU.CleanCudaMemory();

	solverGPU.PrepareCudaMemory();

	solverGPU.GeneInitCudaMemory();

	solverGPU.CleanCudaMemory();

	return 0;

}

这个是main.cpp



#ifndef _GPU_TSP_SOLVER

#define _GPU_TSP_SOLVER

#endif

#ifdef WIN32 

#include "cuda_runtime.h"

#include "device_launch_parameters.h"

#endif

#include "GPUTSPSolverKernel.cuh"

class CGPUSolver {

	int THREADSPERBLOCK;

	int nMaxPoplation;

	int *d_gene;

	int *d_gene_randi;

	int *h_gene;

	int *h_gene_randi;

	curandState *d_cudaState;

public:

	void GenInitCudaMemory(void);

	void CleanCudaMemory(void);

	void PrepareCudaMemory(void);

	void GeneInitCudaMemory(void);

	CGPUSolver();

	~CGPUSolver();

}

这个是GPUTSPSolver.h



#include "GPUTSPSolver.h"

#include <stdio.h>

#include <stdlib.h>

CGPUTSPSolver::CGPUTSPSolver(){

	THREADSPERBLOCK = 256;

	d_gene = NULL;

	d_cudaState = NULL;

	d_gene_randi = NULL;

	h_gene = NULL;

	h_gene_randi = NULL;

}

CGPUTSPSolver::~CGPUTSPSolver() {

}

void CGPUTSPSolver::CleanCudaMemory(void){

	printf("cleaning cuda memories\n");

	if (d_gene)	cudaFree(d_gene);

	if (d_gene_randi)	cudaFree(d_gene_randi);

	if (d_cudaState)	cudaFree(d_cudaState);

	if (h_gene) free(h_gene);

	if (h_gene_randi) free(h_gene_randi);

}

void CGPUTSPSolver::PrepareCudaMemory(void){

	printf("preparing cuda memories...\n");

	cudaMalloc((void **)&d_gene , 2048 * 100 * sizeof(int));

	cudaMalloc((void **)&d_gene_randi , 2048 * 100 * sizeof(int));

	cudaMalloc((void **)&d_cudaState, sizeof(curandState));

	h_gene       = (int *)  malloc(2048 * 100 * sizeof(int));

	h_gene_randi = (int *)  malloc(2048 * 100 * sizeof(int));

	printf("cuda device memory successfully allocated (%d cities, %d genes)\n", 100, 2018);

}

void CGPUTSPSolver::GeneInitCudaMemory(void) {

	int blocksPerGrid = (2048 + THREADSPERBLOCK - 1) / THREADSPERBLOCK;

	if (!d_gene) printf("no gene memory\n");

	printf("gene initialization with %d threads per block x (%d blocks)\n", THREADSPERBLOCK, blocksPerGrid);

	d_geneInit(blocksPerGrid, THREADSPERBLOCK, d_cudaState, 2048, 100, d_gene, d_gene_randi);

	printf("initSolver GPU\n");

	cudaMemcpy(h_gene_randi, d_gene_randi, sizeof(int)*2048*100, cudaMemcpyDeviceToHost);

	for(int i = 0;i<204800;i++){

			printf("h_gene_randi[%d] = %d\n", i, h_gene_randi[i]);

		};

	printf("randi!");

}

这个是GPUTSPSolver.cpp



#include "GPUTSPSolverKernel.cuh"



#include "cuda_runtime.h"

#include "device_launch_parameters.h"

#include <curand.h>

#include <curand_kernel.h>





#include <stdio.h>

#include <math.h>



void d_geneInit(int blocks, int threads, curandState *pState, unsigned int nPopulation, unsigned int numCities, int *gene, int *d_gene_randi) {



	printf("gene initialization at device started....\n");

	d_geneInitKernel << < blocks, threads >> >(pState, nPopulation, numCities, gene, d_gene_randi);

	printf("gene initialization at device done\n");

}



__global__ void d_geneInitKernel(curandState_t *pstate, unsigned int nPopulation, unsigned int numCities, int *gene, int *d_gene_randi) {

	

	// tId: gene idx ( tId-th gene with numCities elements)

	int tId = threadIdx.x + blockIdx.x * blockDim.x;

	if (tId >= nPopulation) return;



	// gene initialization - cards straight

	for (int i = 0; i < numCities; i++) {

		gene[tId*numCities+i] = i;

	}

	

	// gene shuffle

	for (int i = 1; i < numCities; i++) {

		int rIdx = randi(pstate, threadIdx.x+blockIdx.x+blockDim.x,  (unsigned int) 1, (unsigned int) numCities - 1);

		d_gene_randi[tId*numCities + i] = rIdx;

		int t = gene[tId*numCities + i];

		gene[tId * numCities + i] = gene[tId * numCities + rIdx];

		gene[tId * numCities + rIdx] = t;

	}

}



__device__ unsigned int randi(curandState *pState, int add, int min, int max) {

	curandState localState = *pState;

	unsigned int rndval = min + (curand(&localState) * add + add)%(max-min+1);

	*pState = localState;

	return rndval;

}

这个是GPUTSPSollverKernel.cu



#ifndef _GPUTSPKERNEL__H_2015_YMKANG__

#define _GPUTSPKERNEL__H_2015_YMKANG__

#endif



#include "cuda_runtime.h"

#include "device_launch_parameters.h"

#include <curand.h>

#include <curand_kernel.h>



void d_geneInit(int blocks, int threads, curandState *pState, unsigned int nPopulation, unsigned int numCities, int *gene,int *d_gene_randi);

__global__ void d_geneInitKernel(curandState_t *state, unsigned int nPopulation, unsigned int numCities, int *gene, int *d_gene_randi);

__device__ unsigned int randi(curandState *pState, int add, int min, int max);

这个是GPUTSPSollverKernel.cuh
求教各位了，怎么把它们一起编译啊？

...全文