C / C++ / MFC

Re: MeasureString and MeasureCharacterRanges (Page Faults)

1-Dec-08 5:55

GDI+ MeasureString and MeasureCharacterRanges work nice with small strings. An occasional page fault and some additional memory usage but nothing to be alarmed about.

When I push the limits by using large strings and the above two functions numerous times to perform various measurements, I get my fair share of page faults and a reasonable increase in memory usage. Nothing that I wouldn't expect while increasing the demands on the GDI+ library.

However, once I go back to being reasonable and only working with small strings again, the excessive amount of page faults continue to occur. It's like GDI+ can't get itself back together once the memory requirements are reduced.

I've tried "GdiPlusShutdown" and "GdiPlusStartup" just to see if that would get GDI+ to reset itself but the excessive page faulting continues after that. (I didn't expect this to work or be a reasonable solution anyway but it was worth trying to gain more information on the problem)

Has anyone noticed this excessive page faulting (sometimes I can get it up into the 1,000's per refresh cycle) that continues after GDI+ workloads are reduced to simple single character measurements again? Does anyone know how to get GDI+ page faults back down again?

Randor 1-Dec-08 6:19

Randor

1-Dec-08 6:19

Very interesting, can you post a snippit that causes a page fault?

Thanks,
-David Delaune

Re: MeasureString and MeasureCharacterRanges (Page Faults)

bob169721-Dec-08 7:13

Re: MeasureString and MeasureCharacterRanges (Page Faults)

1-Dec-08 7:13

Unfortunately I can't. Too much code in three different classes.

It's a windowless GDI+ edit control that draws on the parent windows client area. It is mainly when I set the GDI+ edit control to multiline and it takes up the metric equivalent of an 8 1/2 x 11 inch piece of paper using the equivalent of a 10 point font size and filling it completely up. The calculations for text selection and custom soft breaking utilize MeasureString and MeasureCharacterRanges in various places. Regions are used to derive the text selection rects.

Most of my usage of the two functions resembles the samples in MSDN.

Nemanja Trifunovic1-Dec-08 9:50

Nemanja Trifunovic

1-Dec-08 9:50

bob16972 wrote:
Has anyone noticed this excessive page faulting (sometimes I can get it up into the 1,000's per refresh cycle) that continues after GDI+ workloads are reduced to simple single character measurements again? Does anyone know how to get GDI+ page faults back down again?

Unfortunatelly, this is an inherent problem with GDI+. Do a search for "GDI+ page faults" and you'll see it all over the internet.

The problem seems to be that GDI+ bitmaps use memory mapped files.

Programming Blog

utf8-cpp

Re: MeasureString and MeasureCharacterRanges (Page Faults)

bob169721-Dec-08 10:30

Recive TCP packets without TX [modified]

1-Dec-08 10:30

Thanks. I had been using MSN for my search engine and I wasn't getting too much. You reminded me that I hadn't used google in a while. I got a few more links there.

I like Bob Powell but in one of the links his comment is, 'To cure the "problem" buy more RAM'. One of my machines has 4GB and nothing else is running except the services so I'm guessing that isn't the problem.

I would agree with some comments that they're probably "soft" page faults mostly but there are enough (or an occasional hard fault gets mixed in there) that it definitely slows my calculations down once the excessive page faults kick in.

I changed the code to "chunk" the calculations which has made it less common to pass the threshold that triggers the massive faulting but I haven't been successful at completely eliminating it and my code is now something not even a mother would be proud of. :: Hack, Hack, Hack :: Cry | :((

mehdir211-Dec-08 5:00

mehdir21

1-Dec-08 5:00

I want to connect to a HUB_SWITCH who has some computers,and recive TCP packets.But i want use only RX wires(recive packets without have any TX wire).Can i do this and implement it in C++ (DOS OS)?

modified on Monday, December 1, 2008 11:12 AM

Re: Recive TCP packets without TX

Randor 1-Dec-08 6:30

Randor

1-Dec-08 6:30

You cannot implement IPv4-TCP without both TX/RX pairs. If you are designing a system to meet military or industrial standards where the regulations state "Device Communication should be read-only" then I would suggest serial RS-232 with the TX wires physically cut.

Btw this should probably have been posted in the Hardware forum[^].

Best Wishes,
-David Delaune

Instead of using nested if-else

krishnakumartm1-Dec-08 4:30

krishnakumartm

1-Dec-08 4:30

Hi
i have requirement like doing 10 steps to complete task
After completion of one only i can enter into another
I used nested if- else to do that like

if (true)
{
if (true)
{
if (true)
{
return true;
}
else
{
}
}
else
{
}
}
else
{
}
is there any better way to handle these type of situation

Thanks in Advance

----------------------------
KRISHNA KUMAR T M

Hamid_RT1-Dec-08 4:34

Hamid_RT

1-Dec-08 4:34

Yes

switch(i)
{
case 0:
break;
case 1:
break;
default:
break;
}

Code-o-mat1-Dec-08 5:30

Code-o-mat

1-Dec-08 5:30

How about:

if (expression1 &&
    expression2 &&
    expression3 &&
    expression4 &&
    ...) ... //success
else //failure...

or an uglier aproach:

try {
  if (!expression1) throw 1;
  if (!expression2) throw 2;
  if (!expression3) throw 3;
  ...
} catch (int where)
{
  //Do whatever here, "where" will even show you where the whole thing failed
}

> The problem with computers is that they do what you tell them to do and not what you want them to do. <

ssclaire1-Dec-08 7:55

ssclaire

1-Dec-08 7:55

The argument is "are guard clauses in methods okay?" And how do you handle resource cleanup if one of the guard clauses fail? I haven't seen a convincing consensus either way.

A specific example: Let's say you have 10 resources to allocate (like opening files, networks, etc) -- for brevity here, let's just use four resources. All resources have to be allocated for the method to continue.

So what we are trying to avoid is something like this (yuck):

isMethodSuccessful = false;        // assume this isn't going to work
resource1 = OpenMyFile(myFile);
if (resource1 == SUCCESS)
{
    resource2 = OpenSomeNetworkConnection(networkID)
    if (resource2 == OKAY)
    {
        resource3 = CreateNewFile(outputFile);
        if (resource3 == SUCCESS)
        {
            resource4 = ConnectToSomething(mySomething);
            if (resource4)
            {
                //
                // Do all my logic here
                //
                isMethodSuccessful = true;      // return success
            }
            else
            {
                 DeleteFile(resource3);    // do cleanup
                 CloseNetwork(resource2);
                 CloseFile(resource1);
             }
         }
         else
         {
              CloseNetwork(resource2);   // do cleanup
              CloseFile(resource1);
         }
     }
     else
     {
         CloseFile(resource1);    // do cleanup
     }

     return (isMethodSuccessful);
}

Guard clauses will convert that mess into something like this:

resource1 = OpenMyFile(myFile);
if (resource1 != SUCCESS)
{
    return failure;
}

resource2 = OpenSomeNetworkConnection(networkID)
if (resource2 != OKAY)
{
    CloseFile(resource1);  // do cleanup
    return failure;
}

resource3 = CreateNewFile(outputFile);
if (resource3 != SUCCESS)
{
    CloseNetwork(resource2);   // do cleanup
    CloseFile(resource1);
    return failure;
}

resource4 = ConnectToSomething(mySomething);
if (!resource4)
{
    DeleteFile(resource3);    // do cleanup
    CloseNetwork(resource2);
    CloseFile(resource1);
    return failure;
}

// 
// Put all your logic here
//

return (success);

...which is far from ideal (i.e. more yuck). So what's the best code logic in this situation?

CPallini1-Dec-08 8:09

CPallini

1-Dec-08 8:09

For instance, the C language goto statement allows to handle well the above scenario.
Smile | :)

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

This is going on my arrogant assumptions. You may have a superb reason why I'm completely wrong.
-- Iain Clarke

[My articles]

Nemanja Trifunovic1-Dec-08 9:18

Nemanja Trifunovic

1-Dec-08 9:18

Actually, your "yuck" example is much better than the "more yack" one. The former is just yacky, and the later is dangerous (think introducing changes to the function and remembering to copy/paste all the resource cleanup).

The only robust and non-yacky way to handle this situation is use of The RAII programming idiom[^]

In this case, the destructors will make sure that all your resources are freed at the end of the scope - even if you throw an exception.

Programming Blog

utf8-cpp

Timothy Baldwin2-Dec-08 12:47

Timothy Baldwin

2-Dec-08 12:47

ssclaire wrote:
And how do you handle resource cleanup if one of the guard clauses fail?

Use RAII, see http://en.wikipedia.org/wiki/Resource_Acquisition_Is_Initialization[^].

CPallini1-Dec-08 5:47

CPallini

1-Dec-08 5:47

Usually this can be handled (as already suggested) with try-catch blocks or (using plain C) with the (in)famous goto.
You may also schedule a set of operation to be invoked inside a loop (for instance with an array of function pointers) and use break to interrupt the iteration whenever needed (i.e. on any operation failure).
Smile | :)

David Crow1-Dec-08 10:17

David Crow

1-Dec-08 10:17

krishnakumartm wrote:
is there any better way to handle these type of situation

Yes, like:

if (ret)
    ret = foo1();
 
if (ret)
    ret = foo2();
 
if (ret)
    ret = foo3();
 
if (ret)
    ret = foo4();

If any of the function calls fail, none of the remaining ones will execute.

"Love people and use things, not love things and use people." - Unknown

"The brick walls are there for a reason...to stop the people who don't want it badly enough." - Randy Pausch

bob169721-Dec-08 16:29

1-Dec-08 16:29

Tons of COM examples and winsock examples use this approach in methods which would normally result in deep nested branch statements...

void CSomeClass::SomeMethod(void)
{
if (FALSE) { return; }

// got here 1

if (FALSE) { return; }

// got here 2

if (FALSE) { return; }

// got here 3

// ...
}

sheshidar1-Dec-08 18:15

sheshidar

1-Dec-08 18:15

if (true)
	{
	if (true)
		{
		if (true)
			{
			return true;
			}
		else
			{
			}
		}
	else
		{
		}
	}
else
	{
	}

Do you need to do something in each else block?
if yes than follow the same, else you can use

if((condition1)&&(condition2)&&(condition3))
	return true;
else
	return false;

Using "extern"

krishnakumartm1-Dec-08 4:05

krishnakumartm

1-Dec-08 4:05

Hi
in my application i am using extern to make application's theApp to access by other

is there any problem if use extern

Thanks in advance

----------------------------
KRISHNA KUMAR T M

Code-o-mat1-Dec-08 4:08

Code-o-mat

1-Dec-08 4:08

I doubt there're any problems using extern for that but you could also use AfxGetApp(), i think that is the preferred method.

> The problem with computers is that they do what you tell them to do and not what you want them to do. <

krishnakumartm1-Dec-08 4:17

krishnakumartm

1-Dec-08 4:17

Hi thanks for reply
already iam using AfxGetApp()
i don't know wether can i able to declare like that or not

----------------------------
KRISHNA KUMAR T M

Code-o-mat1-Dec-08 4:53

Code-o-mat

1-Dec-08 4:53

Pardon?

> The problem with computers is that they do what you tell them to do and not what you want them to do. <