OpenMP 並行提速

<strong><span style="font-size:24px;">//測試VS2010下開啓OpenPM並行提速</span></strong>

<span style="white-space:pre">	</span>LARGE_INTEGER tmp,time1,time2;
	long freq,costTime;
	QueryPerformanceFrequency(&tmp);
	freq = tmp.QuadPart;
	int k = 0;
	CString str;
	int kk[100];
	QueryPerformanceCounter(&time1);
	for (int i = 0;i<100;++i)
	{
		kk[k] = i;
		k++;
		Sleep(10);
	}
	QueryPerformanceCounter(&time2);
	costTime = (long)((time2.QuadPart - time1.QuadPart) * 1000 / freq);

如上代碼運行時間costTime = 999。

開啓OpenMP並行提速：

1、項目---屬性---配置屬性---C/C++---語言---OpenMP支持---是（如下圖）

2、循環前添加

<span style="color:#ff0000;">#pragma omp parallel for</span>

如下圖

	LARGE_INTEGER tmp,time1,time2;
	long freq,costTime;
	QueryPerformanceFrequency(&tmp);
	freq = tmp.QuadPart;
	int k = 0;
	CString str;
	int kk[100];
	QueryPerformanceCounter(&time1);

<pre name="code" class="cpp"><span style="white-space:pre">	</span><span style="color:#ff0000;">#pragma omp parallel for</span>

for (int i = 0;i<100;++i){kk[k] = i;k++;Sleep(10);}QueryPerformanceCounter(&time2);costTime = (long)((time2.QuadPart - time1.QuadPart) * 1000 / freq);

如上代碼運行時間costTime = 135左右。

我的機器是8核的，costTime 約等於 ceil（100/8）×10 = 130

測試可顯著提速，可看下運行完後kk數組中數據，可看出二者間的區別。

不開啓OpenMP 是按順序0-99；開啓後數據混亂，且每次運行都不一樣。