Re: Question about Interlocked functions - C / C++ / MFC Discussion Boards

If you want to use Interlocked*() to do locking then read about SpinLocks.

A very simple version might look something like (note: doesn't suppport timeout, non-reentrant, doesn't enforce thread ownership, ...):

class  SpinLock
{
   volatile  long  lock;
 
public:
    SpinLock( void ) : lock(0)  {;}
 
    void  Aquire ( void ) {
        while( InterlockedCompareExchange(&lock, 1, 0) ) {
            SwitchToThread();
        }
    }
 
    void  Release ( void ) {
        lock = 0;
    }
};

...cmk

Save the whales - collect the whole set

void Release ( void ) { lock = 0; }

why not use:

InterlockedExchange(&lock, 0);

I think that it's better. isn't?

HOW WHAT wrote:
void Release ( void ) { lock = 0; }
why not use:
InterlockedExchange(&lock, 0);

http://msdn2.microsoft.com/en-us/library/ms684122.aspx[^]
Simple reads and writes to properly-aligned 32-bit variables are atomic operations.

So, lock = 0; is atomic.
InterlockedExchange will generate a full memory barrier

http://msdn2.microsoft.com/en-us/library/12a04hfd(VS.80).aspx[^]
A write to a volatile object (volatile write) has Release semantics ... This allows volatile objects to be used for memory locks and releases in multithreaded applications.

For an explaination of full memory barrier and release semantics read:
http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi/linux/bks/SGI_Developer/books/LX_DevDriver_PG/sgi_html/apa.html[^]

...cmk

Save the whales - collect the whole set

Sorry, my bad englist...

Umm, your meaning is if the code is a instruction, unneed use Interlocked functions? it unable happen some question?

If "lock = 0" modification "if (0 == lock)", need Interlocked functions?

and i want let nether code safe in multithreading, how to do?

if (0 == lock)
{
lock = 1;
}
else
{
lock = 0;
}

Thanks very much for your answer.

Interlocked* is needed if you want to do 2 or more things with the value.

In your example:

if( 0 == lock )  lock = 1;
else             lock = 0;

You are doing (at least) 2 things:
1. if( 0 == lock )
2. lock = ...;

Imagine 2 threads, Thread1 and Thread2.
Start with lock = 0.
Thread1 checks 1. above: if( 0 == lock ) and returns true
Thread2 then checks 1. above: if( 0 == lock ) and returns true
Thread1 then does 2. above: lock = 1
Thread2 then does 2. above: lock = 1
Problem: both threads think they are locked.

You must use Interlocked* to make it like 1 instruction (i.e. atomic):

InterlockedCompareExchange(&lock, 1, 0)

...cmk

Save the whales - collect the whole set

volatile bool Sentinel = true;<br />
<br />
unsigned ThreadFunc1( void* pArguments ) {<br />
   while (Sentinel){;}<br />
      <br />
......<br />
   return 0;<br />
} <br />
<br />
unsigned  ThreadFunc2( void* pArguments ) {<br />
   Sentinel = false; // exit critical section<br />
   return 0;<br />
}

--------------------------------------------
Ago, i think if a thread readning a address, and other thread write this address, that will be conflict. But why above not happen error and crash.

you say "You are doing (at least) 2 things", but InterlockedExchange only 1 operation, how to use it.

underside is my some code, can u help me see is it safe?

1.

volatile LONG g_Value = -1;<br />
<br />
LONG Read()<br />
{<br />
......<br />
    return InterlockedExchange(&g_Value, g_Value);<br />
}<br />
<br />
VOID Write(LONG NewValue)<br />
{<br />
......<br />
    return InterlockedExchange(&g_Value, NewValue);<br />
}<br />
<br />
VOID Thread1()<br />
{<br />
    if (Read >= 0)<br />
    {<br />
        Write(-1);<br />
        <br />
        for (INT i = 0; i < 10; i++)<br />
        {<br />
            Write(i);<br />
            if (RunThread3() == error)<br />
            {<br />
                Write(-1);<br />
            }<br />
        }<br />
    }<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
    // code same with Thread1 function<br />
}

-----------------------------------------------------

2.

volatile LONG g_nIndex = -1;<br />
<br />
VOID Thread1()<br />
{<br />
    InterlockedExchange(&g_nIndex, 1);<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
    do<br />
    {<br />
    }<br />
    while (InterlockedExchange(&g_nIndex, g_nIndex) < -1);<br />
}<br />
<br />
VOID Thread3()<br />
{<br />
    INT nIndex = GetIndex();<br />
    InterlockedExchange(&g_nIndex, nIndex);<br />
}

-----------------------------------------------------

3.

volatile LONG g_nData1 = 0L;<br />
<br />
VOID Thread1()<br />
{<br />
    if (InterlockedExchange(&g_nData1, g_nData1) == 0)<br />
    {<br />
        InterlockedExchangePointer(&g_nData1, 1);<br />
    }<br />
<br />
.................<br />
<br />
    if (InterlockedExchangePointer(&g_nData1, g_nData1) == 1)<br />
    {<br />
        InterlockedExchangePointer(&g_nData1, 0L);<br />
    }<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
    // code same with Thread1<br />
}

-------------------------------------------------------

4.

volatile LONG g_bIsLock = 0L;<br />
<br />
VOID Thread1()<br />
{<br />
    if (InterlockedExchange(&g_bIsLock, g_bIsLock) == 1)<br />
    {<br />
        InterlockedExchange(&g_bIsLock, 0L);<br />
    }<br />
    else<br />
    {<br />
        InterlockedExchange(&g_bIsLock, 1);<br />
    }<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
    InterlockedExchange(&g_bIsLock, g_bIsLock);<br />
}<br />
<br />
VOID Thread3()<br />
{<br />
    if (InterlockedExchange(&g_bIsLock, g_bIsLock))<br />
    {<br />
        <br />
    }<br />
}

Thanks very much for your help.

Your code does not make sense.

Do not use:
if( InterlockedExchange() ) {
InterlockedExchange();
}
This is bad, you are doing 2 things (test and set) in 2 instructions.

Instead use:
while( InterlockedCompareExchange(&lock, 1, 0) ) {
// lock was == 0, now == 1, 'this' thread set lock = 1
}
This is better, you are doing 2 things in 1 instruction.

...cmk

Save the whales - collect the whole set

Yes, my code not make sense, it only write example for my question.

Is nether code right now? Which Interlocked functions unneed?

1.

volatile LONG g_Value = -1;<br />
<br />
LONG Read()<br />
{<br />
	// return g_Value;<br />
	return InterlockedExchange(&g_Value, g_Value);<br />
}<br />
<br />
VOID Write(LONG NewValue)<br />
{<br />
	// g_Value = NewValue;<br />
	InterlockedExchange(&g_Value, NewValue);<br />
}<br />
<br />
VOID Thread1()<br />
{<br />
	if (Read() >= 0)<br />
	{<br />
		Write(-1);<br />
<br />
		for (INT i = 0; i < 10; i++)<br />
		{<br />
			......<br />
			Write(i);<br />
<br />
			if (RunThread2() == error)<br />
			{<br />
				......<br />
				Write(-1);<br />
			}<br />
		}<br />
	}<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
	// code same with Thread1 function<br />
}

-----------------------------------------------------

2.

volatile LONG g_nIndex = -1;<br />
<br />
VOID Thread1()<br />
{<br />
	// g_nIndex = 1;<br />
	InterlockedExchange(&g_nIndex, 1);<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
	do<br />
	{<br />
		......<br />
	}<br />
	// while (g_nIndex < -1);<br />
	while (InterlockedExchange(&g_nIndex, g_nIndex) < -1);<br />
}<br />
<br />
VOID Thread3()<br />
{<br />
	INT nIndex = GetIndex();<br />
<br />
	// g_nIndex = nIndex;<br />
	InterlockedExchange(&g_nIndex, nIndex);<br />
}

-----------------------------------------------------

3.

volatile LONG g_nData1 = 0L;<br />
<br />
VOID Thread1()<br />
{<br />
	// if (0L == g_nData1) g_nData = 1;<br />
	InterlockedCompareExchange(g_nData1, 1, 0L);<br />
<br />
	.................<br />
<br />
	// if (1 == g_nData1) g_nData = 0L;<br />
	InterlockedCompareExchange(g_nData1, 0, 1);<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
	// code same with Thread1<br />
}

-------------------------------------------------------

4.

volatile LONG g_bIsLock = 0L;<br />
<br />
VOID Thread1()<br />
{<br />
	// if (1 == g_bIsLock) g_bIsLock = 0;<br />
	InterlockedCompareExchange(g_nData1, 0L, 1);<br />
}<br />
<br />
VOID Thread2()<br />
{<br />
	// if (0 == g_bIsLock)<br />
	if (InterlockedExchange(&g_bIsLock, g_bIsLock) == 0L)<br />
	{<br />
		......<br />
	}<br />
}<br />
<br />
VOID Thread3()<br />
{<br />
	// if (1 == g_bIsLock)<br />
	if (InterlockedExchange(&g_bIsLock, g_bIsLock) == 1)<br />
	{<br />
		......<br />
	}<br />
}

---------------------------------------------

And can u tell me why nether MSDN example unuseInterlocked functions, but it can work right and have not resource conflict( a thread read value, the other one write value)

volatile bool Sentinel = true;<br />
<br />
unsigned ThreadFunc1( void* pArguments ) {<br />
	while (Sentinel){;}<br />
<br />
	......<br />
	return 0;<br />
} <br />
<br />
unsigned ThreadFunc2( void* pArguments ) {<br />
	Sentinel = false; // exit critical section<br />
	return 0;<br />
}

http://msdn2.microsoft.com/en-us/library/ms684122.aspx[^][^]

Unclear why those code run not happen error, if
if "volatile bool Sentinel = true;" -> "bool Sentinel = true;" , have some different?

HOW WHAT wrote:
Unclear why those code run not happen error, if
if "volatile bool Sentinel = true;" -> "bool Sentinel = true;" , have some different?

http://msdn2.microsoft.com/en-us/library/12a04hfd(VS.80).aspx[^]

volatile is important.
It tells the compiler that the value could change any time.
e.g.

void  Thread1( void )
{
    while( Sentinal ) {
        ...
    }
}
void  Thread2( void )
{
    ...
    Sentinal = false;
    (wait for Thread1 to finish)
    ...
}

If Sentinal is NOT volatile then Thread1 may cache the value of Sentinal when the Thread1 starts and never reload it becuase it is not used anywhere else in Thread1 (it optimizes the load out). In the example above this could cause Thread2 to wait forever becuase Thread1 never exits.

If Sentinal IS volatile then Thread1 reloads the value in each iteration of the while loop.
Therefore when Thread2 changes Sentinal Thread1 will see the change.

...cmk

Save the whales - collect the whole set

May i understand that it's real-time and not optimize code of all about this value?

I do not understand what you are saying.

...cmk

Save the whales - collect the whole set

is there a libary or something where i can get numbers which have like 50 digits without truncation, for a program for find prime numbers etc???

also why i'm here

***************************************************
***************************************************

i'm trying to use the .clear() with vectors but they never work

Code:

#include <iostream>
#include <vector>

using namespace std;
int main()
{
vectornumbers;

int createindex = 100;

while(createindex)
{
numbers.push_back(createindex);
createindex--;
}
cout << numbers[20] << endl;
numbers.clear();
cout << numbers[20] << endl;

system("PAUSE");
return 0;
}

what am i doing wrong?

thanks i'n advanced

g3RC4n wrote:
what am i doing wrong?

Not using <pre> tags for one thing. The other thing is that your code snippet (i.e., adding to a vector) has nothing to do with your question (i.e., large numbers, primes).

"A good athlete is the result of a good and worthy opponent." - David Crow

"To have a respect for ourselves guides our morals; to have deference for others governs our manners." - Laurence Sterne

DavidCrow wrote:
Not using <pre> tags for one thing.

Laugh | :laugh:

can you just help me out? Frown | :(

google "large integer library c++"

Peter
"Until the invention of the computer, the machine gun was the device that enabled humans to make the most mistakes in the smallest amount of time."

thank you, found it

I have this VBA code: (uses MS HTML library MSHTML.tlb)

Public ie As InternetExplorer

Sub RacingPost_Login()

Set ie = CreateObject("InternetExplorer.Application")
ie.Visible = True
'Go to this Web Page!
ie.navigate "https://reg.racingpost.co.uk/cde/login_iframe.sd"

' (waits here for good connection)

With ie.document.forms(0)
.in_un.Value = "USERNAME"
.in_pw.Value = "PASSWORD"
.submit
End With

I would like to do this in C++. I've tried for 4+ days.
Any help is greatly appreciated.
Thanks,
Roy

baloneyman wrote:
I would like to do this in C++. I've tried for 4+ days.

May I see what you've done in 4+ days?

Hi, Most of the attempts were erased. I was just using this stub to
plug things in.

I can get logged just by tagging the parameters to the url and doing
a navigate. But that way ends up at a splash screen which the VBA code
bypasses.

This one I think was from MSDN:

CString strHeaders =
   _T("Content-Type: application/x-www-form-urlencoded");

CString strFormData = _T("in_un=xxx&in_pw=xxx&process=IN&PARGS=http%253A%252F%252Fwww.racingpost.co.uk%252Fnews%252Fhome.sd");

CInternetSession session;
CHttpConnection* pConnection =
   session.GetHttpConnection(_T("https://reg.racingpost.co.uk/cde/login_iframe_rp.sd"));

CHttpFile* pFile =
   pConnection->OpenRequest(CHttpConnection::HTTP_VERB_POST, _T("rp_sign"));

BOOL result = pFile->SendRequest(strHeaders,
   (LPVOID)(LPCTSTR)strFormData, strFormData.GetLength());

The form is:

  <FORM name=rpsign onsubmit="rp_sign(document.rpsign);return false" 
method=post>
  <TBODY>
  <TR vAlign=top height=25>
    <TD width=114 colSpan=2><INPUT class=B10 maxLength=30 size=14 
      value=xxx name=in_un></TD>
    <TD width=136 colSpan=2><INPUT class=B10 
      style="BACKGROUND-IMAGE: url(../images/furniture/top_password.gif); BACKGROUND-REPEAT: no-repeat" 
      onfocus=removeBG(this); type=password maxLength=30 size=14 name=in_pw 
      xonFocus="this.backgroundimage:url('../images/furniture/top_password.gif');"></TD></TR>
  <TR>
    <TD width=52><INPUT class=A8 style="WIDTH: 52px" onclick="rp_sign(document.rpsign);return false" type=submit value="LOG IN"></TD>
    <TD width=62><INPUT class=A8 style="WIDTH: 62px" onclick="rp_register(document.rpsign);return false" type=button value=REGISTER><INPUT 
      type=hidden value=IN name=process><INPUT type=hidden 
      value=http%3A%2F%2Fwww.racingpost.co.uk%2Fnews%2Fnews_service.sd%3Fstyle%3D2 
      name=PARGS></TD>
    <TD width=126><INPUT class=A8 style="WIDTH: 126px" onclick="rp_forgotten(document.rpsign);return false" type=button value="FORGOTTEN PASSWORD"></TD></TR></TBODY></TABLE></FORM></BODY></HTML>

I am not too much when it comes to HTTP
Thanks,
Roy

hi, i know how to display database(.mdb) in a list control .But how to copy selected records in the list control into another table(in .mdb file) with the column ~field
Thanks

hero1995 wrote:
But how to copy selected records in the list control into another table...

Break this up into at least two smaller problems. Work on those individually.

"A good athlete is the result of a good and worthy opponent." - David Crow

"To have a respect for ourselves guides our morals; to have deference for others governs our manners." - Laurence Sterne

Hi,

Do anyone know the steps to use PlaySound to play a wave file by using Add resource MFC VC++ 2003?
Am using a button to play a wave file.

Can someone provide a step by step guide.
Thanks in advance

YUANGE wrote:
Do anyone know the steps to use PlaySound to play a wave file by using Add resource...

See here.

"A good athlete is the result of a good and worthy opponent." - David Crow

"To have a respect for ourselves guides our morals; to have deference for others governs our manners." - Laurence Sterne