本博文主要讲解下基于cuda的矩阵相乘,cuda特别擅长的就是矩阵乘法,而且也比较容易实现。通过矩阵乘法的实现,可以比较容易理解cuda的核心思想。网上也有很多基于cuda实现的矩阵乘法,但是感觉都不完成,要不就是有错,本文给出的代码都是经过验证可行的,希望能够帮助到大家。
矩阵乘法实现方式一:矩阵乘法的逐点实现方式,具体如下图所示
对应实现代码:
-
#include <stdio.h>
-
#include <stdlib.h>
-
#include <cuda_runtime.h>
-
-
-
__global__
void MatMul(
int *M,
int *N,
int *P,
int width)
-
{
-
int x = threadIdx.x;
-
int y = threadIdx.y;
-
-
float Pervalue =
0;
-
-
float elem1 =
0.0,elem2 =
0.0,value =
0.0;
-
for(
int i =
0;i < width;i++)
-
{
-
elem1 = M[y * width + i];
//取M矩阵的一行
-
elem2 = N[i * width + x];
//取N矩阵的一列
-
-
value += elem1 * elem2;
//求和
-
}
-
-
P[y * width + x] = value;
-
}
-
-
int main()
-
{
-
const
int ND =
30;
-
int a[ND][ND],b[ND][ND],c[ND][ND];
-
int *M,*N,*P;
-
-
int width = ND;
-
int NUM =
900;
-
dim3 blockSize(ND,ND);
-
-
cudaEvent_t start,stop;
-
float elapsedTime =
0;
-
cudaEventCreate(&start);
-
cudaEventCreate(&stop);
-
-
//设备端内存分配
-
cudaMalloc((
void**)&M,ND * ND *
sizeof(
int));
-
cudaMalloc((
void**)&N,ND * ND *
sizeof(
int));
-
cudaMalloc((
void**)&P,ND * ND *
sizeof(
int));
-
-
//初始化
-
for(
int i =
0;i < ND;i++)
-
{
-
for(
int j =
0;j < ND;j++)
-
{
-
a[i][j] =
2;
-
b[i][j] =
3;
-
}
-
}
-
-
int Size = ND * ND;
-
//数据拷贝,主机到设备
-
cudaMemcpy(M,a,Size *
sizeof(
int),cudaMemcpyHostToDevice);
-
cudaMemcpy(N,b,Size *
sizeof(
int),cudaMemcpyHostToDevice);
-
-
cudaEventRecord(start,
0);
-
MatMul<<<
1,blockSize>>>(M,N,P,width);
//调用核函数
-
cudaThreadSynchronize();
-
cudaEventRecord(stop,
0);
-
cudaEventSynchronize(stop);
-
cudaEventElapsedTime(&elapsedTime,start,stop);
-
-
cudaMemcpy(c,P,Size *
sizeof(
int),cudaMemcpyDeviceToHost);
-
-
printf(
"c0 = %d \n",c[
0][
0]);
-
-
//释放设备内存
-
cudaFree(M);
-
cudaFree(N);
-
cudaFree(P);
-
-
return
0;
-
}
运行结果:
矩阵相乘实现方式二:矩阵乘法分块实现,具体如下图所示
具体代码实现:
-
#include <stdio.h>
-
#include <stdlib.h>
-
#include <cuda_runtime.h>
-
-
-
#define TILE_WIDTH 10
-
-
//核函数的具体实现
-
__
global__ void matmul(int *M,int *N,int *P,int width)
-
{
-
__shared__
float Mds[TILE_WIDTH][TILE_WIDTH];
-
__shared__
float Nds[TILE_WIDTH][TILE_WIDTH];
-
-
int bx = blockIdx.x;
-
int
by = blockIdx.y;
-
int tx = threadIdx.x;
-
int ty = threadIdx.y;
-
-
int Col = bx * TILE_WIDTH + tx;
-
int Row =
by * TILE_WIDTH + ty;
-
-
int Pervalue =
0;
-
-
for(
int i =
0;i < width / TILE_WIDTH;i++)
//有多少个TILE_WIDTH,每个循环计算一个块的大小
-
{
-
Mds[ty][tx] = M[Row * width + (i * TILE_WIDTH + tx)];
-
Nds[ty][tx] = N[Col + (i * TILE_WIDTH + ty) * width];
-
__syncthreads();
-
-
-
for(
int k =
0;k < TILE_WIDTH;k++)
//TILE_WIDTH相乘
-
Pervalue += Mds[ty][k] * Nds[k][tx];
-
__syncthreads();
-
}
-
-
P[Row * width + Col] = Pervalue;
-
}
-
-
-
int main()
-
{
-
const
int Nd =
30;
-
int Size = Nd * Nd;
-
int *M,*N,*P;
-
int width = Nd /
3;
-
-
int a[Nd][Nd];
-
int b[Nd][Nd];
-
int c[Nd][Nd];
-
-
//线程块以及线程的划分
-
dim3 gridSize(Nd / width,Nd / width);
-
dim3 blockSize(width,width);
-
-
cudaEvent_t start,stop;
-
float elapsedTime;
-
cudaEventCreate(&start);
-
cudaEventCreate(&stop);
-
-
//设备内存分配
-
cudaMalloc((
void**)&M,Size *
sizeof(
int));
-
cudaMalloc((
void**)&N,Size *
sizeof(
int));
-
cudaMalloc((
void**)&P,Size *
sizeof(
int));
-
-
//初始化
-
for(
int i =
0;i < Nd;i++)
-
{
-
for(
int j =
0;j < Nd;j++)
-
{
-
a[i][j] =
2;
-
b[i][j] =
3;
-
}
-
}
-
-
//数据拷贝,主机到设备
-
cudaMemcpy(M,a,Size *
sizeof(
int),cudaMemcpyHostToDevice);
-
cudaMemcpy(N,b,Size *
sizeof(
int),cudaMemcpyHostToDevice);
-
-
cudaEventRecord(start,
0);
-
matmul<<<gridSize,blockSize>>>(M,N,P,Nd);
//调用核函数
-
cudaThreadSynchronize();
-
cudaEventRecord(stop,
0);
-
cudaEventSynchronize(stop);
-
cudaEventElapsedTime(&elapsedTime,start,stop);
-
-
-
cudaMemcpy(c,P,Size *
sizeof(
int),cudaMemcpyDeviceToHost);
-
printf(
"c0 = %d\n",c[
0][
0]);
-
-
-
cudaFree(M);
-
cudaFree(N);
-
cudaFree(P);
-
-
return
0;
-
}
运行结果:
本文也参考了网上的一些资料,主要是做了一定的修改以及程序的完备,图片就直接网上copy的,水平有限,有不当之处,请指教,谢谢!