C++openmp并行计算实例

C++利用内部库<omp.h>进行并行加速的实例，但实际效果确更慢，找不出原因。希望各路大神帮忙。

基本功能是计算$\pi$:
$$\pi =4{\int }_0^1\frac{1}{1+x^2}dx$$

//g++ avx_omp.cpp -fopenmp -O2 
#include <iostream>
#include <ctime>
#include <omp.h>
using namespace std; 
//正常的逐个累加运算
double compute_pi_naive(size_t dt){
	double pi = 0.0;
	double delta = 1.0/dt;
	for (size_t i =0;i<dt;i++){
		double x = (double)i/dt;
		pi += delta /(1+x*x);
//		cout<<omp_get_thread_num()<<endl;
	}
	return pi*4.0;
}
//多线程并行
double compute_pi_omp(size_t dt){
	double pi = 0.0;
	double delta = 1.0/dt;
	omp_set_num_threads(4);//设置四个核
	#pragma omp parallel
        {
	#pragma omp for  reduction(+:pi)
	for (size_t i =0;i<dt;i++){
		double x = (double)i/dt;
		pi += delta /(1+x*x);
//		cout<<omp_get_thread_num()<<endl;
	}
    }
	return pi*4.0;
return pi*4.0;
}
int main(){
	clock_t start,end;
	size_t dt = 134217728;
	double result1,result2;
	//普通函数计时
	start = clock();
	result1 = compute_pi_naive(dt);
	end  = clock();
	cout<<"naive:\n"<< result1 <<endl<<end- start <<endl;
    //omp 
	start = clock();
	result2 = compute_pi_omp(dt);
	end  = clock();
	cout<<"omp:\n" <<result2 <<endl<<end- start <<endl;

	return 0;
}

利用了四个线程，结果竟然更慢，真是无语了。