CUDA使用多维数组

https://devtalk.nvidia.com/default/topic/821510/is-it-possible-to-process-multidimensional-arrays-inside-the-kernel-/

#include <stdio.h>
#include <stdlib.h>
#include <time.h>

#include <cuda.h>
#include <cuda_runtime.h>

#define FILE_MAX 3
#define MOL_MAX 3
#define ATOM_MAX 3
#define XMAX 5
#define YMAX 5
#define ZMAX 5

typedef struct full {
	int array[FILE_MAX][MOL_MAX][ATOM_MAX][ZMAX][YMAX][XMAX];
} full_t;

typedef struct coord {
	int data[ZMAX][YMAX][XMAX];
} coord_t;

__global__ void sample(coord_t * a)
{
	int x = threadIdx.x + blockDim.x * blockIdx.x;
	int y = threadIdx.y + blockDim.y * blockIdx.y;
	int z = threadIdx.z + blockDim.z * blockIdx.z;

	a->data[z][y][x] = z * 100 + y * 10 + x;
	a->data[z][y][x]++;

}

int main(void)
{
	int i, j, k;
	int x, y, z;
	dim3 BlockPerGrid(1, 1, 1);
	dim3 ThreadPerBlock(XMAX, YMAX, ZMAX);
	size_t size;

	full_t *top;
	size = sizeof(full);

	cudaMallocManaged(&top, size);

	for (i = 0; i < FILE_MAX; i++) {
		for (j = 0; j < MOL_MAX; j++) {
			for (k = 0; k < ATOM_MAX; k++) {
				sample << < BlockPerGrid, ThreadPerBlock >> > ((coord_t *) &(top->array[i][j][k][0][0][0]));
				cudaDeviceSynchronize();
			}
		}
	}

	for (i = 0; i < FILE_MAX; i++) {
		for (j = 0; j < MOL_MAX; j++) {
			for (k = 0; k < ATOM_MAX; k++) {
				for (z = 0; z < ZMAX; z++) {
					for (y = 0; y < YMAX; y++) {
						for (x = 0; x < XMAX; x++) {
							printf(" %d", top->array[i][j][k][z][y][x]);
						}
						printf("\n");
					}
					printf("\n");
				}
				printf("\n");
			}
			printf("\n");
		}
		printf("\n");
		printf("\n");
	}

	cudaFree(top);

}