cudaMemcpy2D()和cudaMallocPitch()的使用

#include "head.h"

void gpuInit()
{
	cudaSetDevice(0);
	size_t width = 5;
	size_t height = 5;
	float *h_data, *d_data;
	size_t pitch;

	h_data = (float*)malloc(sizeof(float)*width * height);
	for(int i = 0; i < width * height; ++i)
	{
		h_data[i] = (float)i;
	}
	std::cout << "Allow memory" << std::endl;
	cudaMallocPitch((void**)&d_data,&pitch,sizeof(float)*width, height);
	std::cout << "copy to device" << std::endl;
	cudaMemcpy2D(d_data,pitch,h_data,sizeof(float)*width,sizeof(float)*width,height,cudaMemcpyHostToDevice);

	gpu<<< 1, 1>>>(d_data,height,width,pitch);
	cudaDeviceSynchronize();
	std::cout << "copy to Host" << std::endl;
	cudaMemcpy2D(h_data,sizeof(float)*width,d_data,pitch,sizeof(float)*width,height,cudaMemcpyDeviceToHost);

	for(int i = 0; i < width * height; ++i)
	{
		std::cout << h_data[i] << " ";
		if((i + 1) % width == 0)
		{
			std::cout << std::endl;
		}
	}
	free(h_data);
	cudaFree(d_data);
}
__global__ void gpu(float *devPtr,int height,int width,int pitch)
{
	int row, col;
	float *rowHead;
	for(row = 0;row < height; ++row)
	{
		rowHead = (float*)((char*)devPtr + row * pitch);
		for(col = 0; col < width;++col)
		{
			printf("\t%f",rowHead[col]);
			rowHead[col]++;
		}
		printf("\n");
	}
}