遍历数组和线程时发生访问冲突

Question

我在Windows中玩Threading，并组成了这个示例，它应该添加128Mb阵列的所有位置。 我创建了x个线程来计算总和，因此我将数组除以x个和平度，然后让每个线程计算这些和平度之一。 在我尝试创建超过64个线程之前，一切都很好。 例如，如果我创建65个线程，则会在我的add函数中遇到访问冲突。 我的猜测是，它是一个超出范围的数组，我只是不明白为什么在64个线程之后出现此错误。

#include <windows.h>
#include <tchar.h>
#include <strsafe.h>

#define MAX_ARRAY_SIZE 128 * 1024 * 1024
#define MAX_THREADS_BUFFER 512

DWORD dwSampleData[MAX_ARRAY_SIZE];
DWORD dwTotal;

DWORD WINAPI DoWork(LPVOID lpParam);
void ErrorHandler(LPTSTR lpszFunction);
VOID InitializedwSampleData();
VOID CreateThreadsAndDoWork(DWORD MaxThreads, DWORD dwPrintIntermediateResults);

typedef struct _THREAD_ARGS{
    DWORD * pdwSampleData;
    DWORD dwOffset;
    DWORD dwSize;
    DWORD dwPrintIntermediateResults;
}THREAD_ARGS, *PTHREAD_ARGS;

DWORD _tmain()
{
    InitializedwSampleData();
    CreateThreadsAndDoWork(1, FALSE);
    CreateThreadsAndDoWork(2, FALSE);
    CreateThreadsAndDoWork(4, FALSE);
    CreateThreadsAndDoWork(8, FALSE);
    CreateThreadsAndDoWork(16, FALSE);
    CreateThreadsAndDoWork(32, FALSE);
    CreateThreadsAndDoWork(64, FALSE);
    CreateThreadsAndDoWork(128, FALSE); // <----------- More than 64 threads

    printf("Press any key to finish");
    getchar();
    return 0;
}

VOID InitializedwSampleData(){
    DWORD i;

    for (i = 0; i < MAX_ARRAY_SIZE; i++){
        dwSampleData[i] = 1;
    }
}

VOID CreateThreadsAndDoWork(DWORD MaxThreads, DWORD dwPrintIntermediateResults){
    PTHREAD_ARGS pDataArray[MAX_THREADS_BUFFER];
    DWORD   dwThreadIdArray[MAX_THREADS_BUFFER];
    HANDLE  hThreadArray[MAX_THREADS_BUFFER];
    DWORD BeginTickCount;

    // Reset dwTotal;
    dwTotal = 0;

    // Get Initial Tick Count
    BeginTickCount = GetTickCount();

    // Create MAX_THREADS worker threads.
    for (DWORD i = 0; i < MaxThreads; i++)
    {
        // Allocate memory for thread data.
        pDataArray[i] = (PTHREAD_ARGS)HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
            sizeof(THREAD_ARGS));

        if (pDataArray[i] == NULL)
        {
            // If the array allocation fails, the system is out of memory
            // so there is no point in trying to print an error message.
            // Just terminate execution.
            ExitProcess(2);
        }

        // Generate data for each thread to work with.
        pDataArray[i]->dwOffset = i * (MAX_ARRAY_SIZE / MaxThreads);
        pDataArray[i]->dwSize = MAX_ARRAY_SIZE / MaxThreads;
        pDataArray[i]->pdwSampleData = dwSampleData;
        pDataArray[i]->dwPrintIntermediateResults = dwPrintIntermediateResults;

        // Create the thread to begin execution on its own.
        hThreadArray[i] = CreateThread(
            NULL,                   // default security attributes
            0,                      // use default stack size  
            DoWork,                 // thread function name
            pDataArray[i],          // argument to thread function 
            0,                      // use default creation flags 
            &dwThreadIdArray[i]);   // returns the thread identifier 


        // Check the return value for success.
        // If CreateThread fails, terminate execution. 
        // This will automatically clean up threads and memory. 

        if (hThreadArray[i] == NULL)
        {
            ErrorHandler(TEXT("CreateThread"));
            ExitProcess(3);
        }
    } // End of main thread creation loop.

    // Wait until all threads have terminated.
    WaitForMultipleObjects(MaxThreads, hThreadArray, TRUE, INFINITE);

    // Close all thread handles and free memory allocations.
    for (DWORD i = 0; i < MaxThreads; i++)
    {
        CloseHandle(hThreadArray[i]);
        if (pDataArray[i] != NULL)
        {
            HeapFree(GetProcessHeap(), 0, pDataArray[i]);
            pDataArray[i] = NULL;    // Ensure address is not reused.
        }
    }
    // Print Results
    _tprintf(TEXT("Computation task with %d thread(s): Added to %d in %d mills\n"), MaxThreads, dwTotal, GetTickCount() - BeginTickCount);
}

DWORD WINAPI DoWork(LPVOID lpParam)
{
    DWORD i;
    DWORD sum = 0;

    for (i = ((PTHREAD_ARGS)lpParam)->dwOffset; i < ((PTHREAD_ARGS)lpParam)->dwSize + ((PTHREAD_ARGS)lpParam)->dwOffset; i++){
        sum += ((PTHREAD_ARGS)lpParam)->pdwSampleData[i]; // <------------ ACCESS VIOLATION ERROR
    }

    dwTotal += sum;
    if (((PTHREAD_ARGS)lpParam)->dwPrintIntermediateResults){
        _tprintf(TEXT("\nSUM = %d\n"), sum);
    }

    return 0;
}

void ErrorHandler(LPTSTR lpszFunction)
{
    // Retrieve the system error message for the last-error code.
    LPVOID lpMsgBuf;
    LPVOID lpDisplayBuf;
    DWORD dw = GetLastError();

    FormatMessage(
        FORMAT_MESSAGE_ALLOCATE_BUFFER |
        FORMAT_MESSAGE_FROM_SYSTEM |
        FORMAT_MESSAGE_IGNORE_INSERTS,
        NULL,
        dw,
        MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
        (LPTSTR)&lpMsgBuf,
        0, NULL);

    // Display the error message.
    lpDisplayBuf = (LPVOID)LocalAlloc(LMEM_ZEROINIT,
        (lstrlen((LPCTSTR)lpMsgBuf) + lstrlen((LPCTSTR)lpszFunction) + 40) * sizeof(TCHAR));
    StringCchPrintf((LPTSTR)lpDisplayBuf,
        LocalSize(lpDisplayBuf) / sizeof(TCHAR),
        TEXT("%s failed with error %d: %s"),
        lpszFunction, dw, lpMsgBuf);
    MessageBox(NULL, (LPCTSTR)lpDisplayBuf, TEXT("Error"), MB_OK);

    // Free error-handling buffer allocations.
    LocalFree(lpMsgBuf);
    LocalFree(lpDisplayBuf);
}

Answer 1

WaitForMultipleObjects对可以等待的线程数有限制。 该限制是MAXIMUM_WAIT_OBJECTS ，其值为64。

这意味着当您调用带有128个句柄的WaitForMultipleObjects时，它将立即返回一个错误，您将忽略该错误。 完全清楚地说，在线程完成之前， WaitForMultipleObjects将返回。 然后，您将释放堆内存，结果是，试图访问现在已释放的内存时，仍在运行的线程将失败。

要学习的主要课程始终是检查Win32 API函数的返回值。 如果检查了WaitForMultipleObjects返回的值，就会发现问题。

若要解决此问题，您将需要在循环中重复调用WaitForMultipleObjects 。 等待第一个MAXIMUM_WAIT_OBJECTS线程。 然后是下一个WaitForMultipleObjects线程。 依此类推，直到没有其他可等待的为止。