影 的专栏

#include #include #include #include #include using namespace std;#define M 3200 //num of person#define N 3200 //num of car#define B_S 32#define tile_x 2#define tile_y 2#define SHOW//P[M]*C[N]=D[M][N

#include #include #include #include #include using namespace std;#define M 3200 //num of person#define N 3200 //num of car#define B_S 32//#define SHOW//P[M]*C[N]=D[M][N]__global__ void distance_gpu(fl

#include #include #include #include #include using namespace std;#define M 3200#define N 3200#define P 3200#define B_S 32//#define SHOW//A[M][P]*B[P][N]=C[M][N]__global__void mulMatri_gpu(float* A, fl

#include #include #include #include #include using namespace std;#define M 3200 //num of person#define N 3200 //num of car#define B_S 32//#define SHOW//P[M]*C[N]=D[M][N]__global__ void distance_gpu(fl

#include #include #include #include using namespace std;#define M 1024#define N 1024#define P 512//void mulMatri(int A[M][P],int B[P][N],int C[M][N],int m,int n,int p);inline void mulMatri(float** A,f

#include #include #include #include #include using namespace std;#define M 5000#define N 50000#define P 1struct Point{ int x; int y;};//void mulMatri(int A[M][P],int B[P][N],int C[M][N],int m,i

1.架构理解1）.包括grid，block，(warp),thread这几个层次，     这就是我们所说的SM；即：grid对应整个GPU，block数量代表SM数量，每个block的thread数量对应ALU数量       2.编写规范       1）核函数前缀__global__，而且必须是void类型。 2）>>,N个block，每个block有M个线程。3）GPU的核函数的执行与CP