按照前文多线程的交错配对方式实现并行规约求和方式,实现CUDA版本的并行规约求和,由于这种方式的规约可以避免线程束的分化,因此不需要进行类似于相邻配对那种方式的优化。
交错与优化相邻模式相比,计算效率提升到1.14倍,性能提高有限,这主要受限于全局内存的加载和存储模式。
并行规约的示意图:
#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <stdio.h>
#include "math.h"
#include "stdlib.h"
//错误检查的宏定义
#define CHECK(call) \
{ \
const cudaError_t status=call; \
if (status!=cudaSuccess) \
{ \
printf("文件:%s,函数:%s,行号:%d\n",__FILE__, \
__FUNCTION__,__LINE__); \
printf("%s", cudaGetErrorString(status)); \
exit(1); \
} \
} \
//核函数
__global__ void Kernel(int *d_data, int *d_local_sum, int N)
{
int tid = threadIdx.x;
int index = blockIdx.x*blockDim.x + threadIdx.x;
int *data = d_data + blockIdx.x*blockDim.x;
if (index >= N) return;
for (int strize = blockDim.x / 2; strize > 0; strize >>= 1)
{
if (tid < strize)
data[tid] += data[tid + strize];
__syncthreads();
}
if (tid == 0)
{
d_local_sum[blockIdx.x] = data[0];
}
}
//主函数
int main()
{
//基本参数设置
cudaSetDevice(0);
const int N = 16777216;
int local_length = 1024;
int total_sum = 0;
dim3 grid(((N + local_length - 1) / local_length), 1);
dim3 block(local_length, 1);
int *h_data = nullptr;
int *h_local_sum = nullptr;
int *d_data = nullptr;
int *d_local_sum = nullptr;
//Host&Deivce内存申请及数组初始化
h_data = (int*)malloc(N * sizeof(int));
h_local_sum = (int*)malloc(int(grid.x) * sizeof(int));
CHECK(cudaMalloc((void**)&d_data, N * sizeof(int)));
CHECK(cudaMalloc((void**)&d_local_sum, int(grid.x) * sizeof(int)));
for (int i = 0; i < N; i++)
h_data[i] = int(10 * sin(0.02*3.14*i));//限制数组元素值,防止最终求和值超过int的范围
//数据拷贝至Device
CHECK(cudaMemcpy(d_data, h_data, N * sizeof(int), cudaMemcpyHostToDevice));
for (int i = 0; i < 200; i++)
//执行核函数
Kernel << <grid, block >> > (d_data, d_local_sum, N);
//数据拷贝至Host
CHECK(cudaMemcpy(h_local_sum, d_local_sum, int(grid.x) * sizeof(int),
cudaMemcpyDeviceToHost));
//同步&重置设备
CHECK(cudaDeviceSynchronize());
CHECK(cudaDeviceReset());
for (int i = 0; i < int(grid.x); i++)
{
total_sum += h_local_sum[i];
}
printf("%d \n", total_sum);
//getchar();
return 0;
}