做了個小程序，用來驗證Interlocked的效果：

上圖分別是注釋掉“標記1”和“標記2”后的截圖，和取消注釋后的截圖，下面是代碼：

    #include <windows.h>
#include <stdio.h>
#include <float.h>
#include <process.h>
#include <conio.h>

#include <iostream>
#include <tchar.h>

using namespace std;

LONG g = 0, flag = 0, T = 1, F = 0;

void threadfun1(void *pNull);
void threadfun2(void *pNull);

int main()
{
	TCHAR	szBuf[10240] = {0};// = TEXT("Thread Interlocked Experiment BEGIN!!!\n");
	HANDLE	harr[2];

	harr[0] = (HANDLE)_beginthread(threadfun1, 0, szBuf);
	harr[1] = (HANDLE)_beginthread(threadfun2, 0, szBuf);

	// 如果不等兩個線程結束，則極有可能主函數結束了，而那兩個線程沒結束！;
	// 也就是說，屏幕上將無任何輸出！;
	WaitForMultipleObjects(2, harr, TRUE, INFINITE);

	_tprintf(szBuf);
	printf("%d\n", g);
	getch();
	return 0;
}

void threadfun1(void *pNull)
{
	// 標記1
	while (InterlockedExchange(&flag, T) == T)
		;
	g++;
	int i, len, j;
	TCHAR *pStr = (TCHAR*)pNull;
	for (i = 0; i < 10; ++i)
	{
		for (j = 0; j < 10; ++j)
		{
			len = lstrlen(pStr);
			pStr[len++] = '1';
			pStr[len] = 0;
		}
		len = lstrlen(pStr);
		pStr[len++] = '\n';
		pStr[len] = 0;
		Sleep(10);	// 這樣使實驗的效果更明顯;
	}
	InterlockedExchange(&flag, F);
}

void threadfun2(void *pNull)
{
	// 標記2
	while(InterlockedExchange(&flag, T) == T)
		;
	g++;
	int i, len, j;
	TCHAR *pStr = (TCHAR*)pNull;
	for (i = 0; i < 10; ++i)
	{
		for (j = 0; j < 10; ++j)
		{
			len = lstrlen(pStr);
			pStr[len++] = '2';
			pStr[len] = 0;
		}
		len = lstrlen(pStr);
		pStr[len++] = '\n';
		pStr[len] = 0;
		Sleep(10);
	}
	InterlockedExchange(&flag, F);
}

  

Chapter 8 Thread Synchronization in User Mode