关于cuda编程的一些问题

[sand_and_fish]

小菜菜有几个关于cuda 的问题想请教一下：
1、请问大家在vs环境下，都是如何调试.cu文件的呢？
2、nvcc 具体的该怎么用呢？是直接在dos里面敲入命令就可以么？
3、memcheck的用法和存在的意义是什么呢？
4、最后呢，还有一个程序希望大家能帮我想想问题出在哪里呢？话不多说，下面先贴代码

cudaError_t addWithCuda(float *r, float *a, float *b, float *c, size_t size);

__global__ void addKernel(float *r, float *a, float *b, float *c)
{
	int i = threadIdx.x;
	int tem = a[i] + b[i];
	tem=tem+c[i];
	r[i]=tem;

}

int main()
{
	const int arraySize = 1000;
	float *a ;
	float *b ;
	float *c ;
	float * r,*r1;
	a=(float *)malloc(sizeof(float)*arraySize);
	b=(float *)malloc(sizeof(float)*arraySize);
	c=(float *)malloc(sizeof(float)*arraySize);
	r=(float *)malloc(sizeof(int)*arraySize);
	r1=(float *)malloc(sizeof(int)*arraySize);
	for (int i=0; i<arraySize; i++)
	{
		a[i]=(float)i;
		b[i]=1.0;
		c[i]=1.;
		r[i]=0.;
		r1[i]=a[i]+b[i]+c[i];
	}

	// Add vectors in parallel.
	cudaError_t cudaStatus = addWithCuda(r, a, b,c, arraySize);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "addWithCuda failed!");
		return 1;
	}
	for (int i=0;i<arraySize;i++)
	{
		printf("%f+%f+%f=%f  r1[i]:%f\n",a[i],b[i],c[i],r[i],r1[i]);
	}
	// cudaThreadExit must be called before exiting in order for profiling and
	// tracing tools such as N sight and Visual Profiler to show complete traces.
	cudaStatus = cudaThreadExit();
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaThreadExit failed!");
		return 1;
	}
	free(a);
	free(b);
	free(c);
	free(r);
	free(r1);
	return 0;
}

// Helper function for using CUDA to add vectors in parallel.
cudaError_t addWithCuda(float *r, float *a, float *b, float *c, size_t size)
{
	float *dev_a = 0;
	float *dev_b = 0;
	float *dev_c = 0;
	float *dev_r = 0;
	int threadsNum=256;
	int blockNum=(size+threadsNum-1)/threadsNum;
	cudaError_t cudaStatus;

	// Choose which GPU to run on, change this on a multi-GPU system.
	cudaStatus = cudaSetDevice(0);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaSetDevice failed!  Do you have a CUDA-capable GPU installed?");
		goto Error;
	}

	// Allocate GPU buffers for three vectors (two input, one output)    .

	cudaStatus = cudaMalloc((void**)&dev_r, size * sizeof(float));
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMalloc failed!");
		goto Error;
	}
	cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(float));
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMalloc failed!");
		goto Error;
	}

	cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(float));
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMalloc failed!");
		goto Error;
	}

	cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(float));
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMalloc failed!");
		goto Error;
	}

	// Copy input vectors from host memory to GPU buffers.
	cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(float), cudaMemcpyHostToDevice);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMemcpy failed!");
		goto Error;
	}

	cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(float), cudaMemcpyHostToDevice);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMemcpy failed!");
		goto Error;
	}
	cudaStatus = cudaMemcpy(dev_c, c, size * sizeof(float), cudaMemcpyHostToDevice);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMemcpy failed!");
		goto Error;
	}

	// Launch a kernel on the GPU with one thread for each element.
	addKernel<<<blockNum, threadsNum>>>(dev_r, dev_a, dev_b,dev_c);
	
	// cudaThreadSynchronize waits for the kernel to finish, and returns
	// any errors encountered during the launch.
	cudaStatus = cudaThreadSynchronize();
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaThreadSynchronize returned error code %d after launching addKernel!\n", cudaStatus);
		goto Error;
	}

	// Copy output vector from GPU buffer to host memory.
	cudaStatus = cudaMemcpy(r, dev_r, size * sizeof(float), cudaMemcpyDeviceToHost);
	if (cudaStatus != cudaSuccess) {
		fprintf(stderr, "cudaMemcpy failed!");
		goto Error;
	}

Error:
	cudaFree(dev_c);
	cudaFree(dev_a);
	cudaFree(dev_b);
	cudaFree(dev_r);

	return cudaStatus;
}

这段代码的问题在于：输出的结果不正确，或者说，数据个数较少的时候是正确的，但是数据数据多一点就会发生错误，这是为什么呢？
下面是程序的运行结果：

最后一列的r1的值是cpu跑出来的正确的结果

[sand_and_fish]

哪位大神来回复一下吧

[sand_and_fish]

引用 2 楼 cc19851 的回复:

1.直接就可以调试 2.nvcc就是编译器，使用和gcc，icc类似 3.没用过 4.int i = threadIdx.x; ---〉int i = blockIdx.x*blockDim.x+threadIdx.x; if(i<size)

我还是不是特别明白，第二个问题哎，没有用过gcc，都是直接vs自带的编译器的呀 编程能力不太好，问的问题太白了呀，惭愧惭愧呢

[cc19851]

1.直接就可以调试 2.nvcc就是编译器，使用和gcc，icc类似 3.没用过 4.int i = threadIdx.x; ---〉int i = blockIdx.x*blockDim.x+threadIdx.x; if(i<size)

[cc19851]

nvcc 会调用 cl，就是vs使用的编译器。 装完toolkit和sdk后，vs就会和cuda集成，也就是加入了cuda的规则文件，这时cu文件会默认调用nvcc进行编译，所以vs可以直接编译cuda程序，但是有的依赖项需要你自己设置，如头文件目录，附加库目录，也就是命令行下的 -I,-L。 如果windows下开发的话，基本不需要在命令行下make

[yjoe61]

cuda-memcheck <memcheck参数> <你的程序> <你的程序的参数> 可以查你程序里有没有非法访问之类的。