|
Excuse me if I post in the wrong place but I'm having a very wierd problem.
I have an application written with MFC. Basically, it has around 10 worker threads doing smth continuously and post message to CMainFrame when done. The MainFrame will process these messages and update some list control on GUI. Simple, isnt' it?
Originally, I test and deploy my application on a Windows XP SP2 machine with CPU Celeron 2,4GHz. It runs well although it consumes 40-50% CPU all the time.
Now, I have a server with CPU Xeon 3GHz, 1GB RAM, SCSI hard disk with RAID 1 (seems to be an overkill hah). On this server, my application only consumes 5% CPU but it runs LIKE HELL. It takes ages for me to access menu items and do other things on the GUI. In other word, the GUI seems to be frozen regularly. I don't know what's going on !!!
I guess that because windows server 2003 has some mechanism to schedule and distribute CPU between applications. I have tried to config Processor scheduling & memory usage to best performance for program, it helps improves a bit, but still far from what's shown on the old Celeron computer !!! Setting process priority into "real time" in task manager also doesn't help.
I'm clueless now, any one please help me 
|
|
|
|
|
You really need to look at your implementation.
I don't know what your threads are doing but you need to make sure the UI thread gets some time
to do stuff. If you're posting too many messages to the UI thread from the worker threads
then the UI thread can't handle user input/output tasks as effectively.
Changing hardware platforms really can expose these problems 
Instead of using the UI message queue, you may want to look into queing messages yourself (in a
FIFO list, for example) and handling them during idle times on the UI thread.
Mark
"Go that way, really fast. If something gets in your way, turn."
|
|
|
|
|
thanks,
I have checked by installing windows server 2003 on my old Celeron computer and found that the problem was not because of the OS (windows 2003 or XP) but because of changing hardware platform.
I shall checked my implementation as your suggestion, thanks for that. But I wonder if there is a way to optimize code compiling for an Intel Xeon server (32 bit). There is an option "Target machine" in project properties but that seems to be not working since changing it cause my project not compilable.
|
|
|
|
|
Optimizing for specific hardware is way out of my league.
From what I've read, it's best to let the OS handle it unless one really know's what (s)he's
doing.
Most often, multithread problems are in the implementation.
Mark
"Go that way, really fast. If something gets in your way, turn."
|
|
|
|
|
For example:
<br />
<br />
BYTE b[16];<br />
BYTE b2[32];<br />
BYTE b3[64];<br />
<br />
vector<???> v;
<br />
v.push_back(?);<br />
v.push_back(?);<br />
v.push_back(?);<br />
If I declaration like this:
<br />
BYTE b[16];<br />
<br />
vector<BYTE*> v;<br />
v.push_back(b);<br />
The item what stored in vector is the point to b, but not the data;
|
|
|
|
|
You probably want to store BYTE *, which can then point to arrays of various sizes. Or, you want a vector of vectors of bytes, which is probably harder to work with.
Christian Graus - Microsoft MVP - C++
"I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
|
|
|
|
|
Perhaps the Boost's Any[^] library is what you're after.
Steve
|
|
|
|
|
|
vector<vector<BYTE> >?
Angle brackets get chopped off when not enclosed in pre tags.
[edit]Even weirder things happen with nested <>[/edit]
- S
50 cups of coffee and you know it's on!
|
|
|
|
|
how to obtain messages that have been posted/sent into a control?
SPY++ only shows messages that the control has sent out, not messages that it receives
help file from spy++
[QUOTE]
P The message was posted to the queue with the PostMessage function. No information is available concerning the ultimate disposition of the message.
S The message was sent with the SendMessage function. This means that the sender doesn’t regain control until the receiver processes and returns the message. The receiver can, therefore, pass a return value back to the sender.
s The message was sent, but security prevents access to the return value.
R Each ‘S’ line has a corresponding ‘R’ (return) line that lists the message return value. Sometimes message calls are nested, which means that one message handler sends another message. [/QUOTE]
|
|
|
|
|
Assuming that you have the source for the control, you can override the DefWindowProc() function and insert whatever message traps you want.
|
|
|
|
|
awah wrote: SPY++ only shows messages that the control has sent out, not messages that it receives
Incorrect. Spy++ shows the messages sent to a control. The 2nd value in the spy display is the window handle of the RECEIVING window. What you can't tell is where the message originated.
Judy
|
|
|
|
|
Dear all
how to do colour image quantization using VC++? who has code snippet? thanks a lot.
zhiyuan li
Li Zhiyuan
27/06/2007
|
|
|
|
|
|
while(getline(passage,nextword,'\n'))
{
int status = 0;
transform(nextword.begin(), nextword.end(), nextword.begin(), (int(*)(int))std::toupper);
int x=nextword.find(test1);
int y=nextword.find(test2);
int z=nextword.find(test3);
if(x!=std::string::npos)
{
status = 1;
}
if(y!=std::string::npos|| z!=std::string::npos)
{
status = 2;
}
string nextword;
ostringstream os(nextword);
string readInWord;
while (os>>readInWord)
{
if (status==0)
++wordCount[readInWord];
else if (status==1)
{
++wordCount[readInWord];
++single[readInWord];
}
else if (status==2)
{
++wordCount[readInWord];
++single[readInWord];
++compound[readInWord];
}
}
}
I am trying hard to get this code to work but seems to be something wrong. I am testing for 3 things and if present I am trying to update certain maps accordingly. I am getting some sort of error on the ostringstream os(nextword) not sure what this is or what it does. Someone put it in my code saying this would fix my problem but I have yet to figure it out.
|
|
|
|
|
andyspartyshop wrote: not sure what this is or what it does
Then why are you using it ? If you search this site for ostringstream, I have an article on it.
You're initialising a stringstream with a string which has not been initialised ( although it will initialise to "" ). What's the point of that ?
The code is pretty messy. At a minimum, change the if/else to a switch block and give x/y/z more meaningful names.
Christian Graus - Microsoft MVP - C++
"I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
|
|
|
|
|
Hi All,
Can any one tell me how to convert LPRSTR data type to string date type in VC++. I am doing in the following way.
LPTSTR testLpt; // this I am getting from some method
string testStr = string(testLpt);
I am getting type cast error.
Can any tell me how to do this?
Thanks in advance.
Regards,
JK
|
|
|
|
|
Sounds like your LPTSTR is Unicode (wchar_t).
Try a wstring instead.
You also don't need to construct a temporary string and copy it to another string.
You can construct it in one step:
LPTSTR testLpt; // this I am getting from some method
wstring testStr(testLpt);
"Go that way, really fast. If something gets in your way, turn."
|
|
|
|
|
LPTSTR can be either Unicode or not depending on whether UNICODE is defined. Using std::wstring is ok if you are sure LPTSTR is always unicode. But in this case it would be more correct to have LPWSTR. What to do in current case, is to use std::basic_string<TCHAR>:
typedef std::basic_string<TCHAR> tstring;
Now, conversions between LPTSTR and tstring will work regardless of being unicode or not.
--
=====
Arman
|
|
|
|
|
Thanks for the info
Mark
"Go that way, really fast. If something gets in your way, turn."
|
|
|
|
|
Hi Arman and Mark,
Thanks a lot. Its working fine with wstring for me.There are no compilation errors now.
But I am having one more problem.
I am using text resource file in my VC++ workspace.
I have included *.rc and *.h file. but I am not able to get the handle to resource file. Following is the sample code.
TCHAR sResName[5] = _T("#103");
TCHAR sRestype[12] = _T("MY_RESOURCE");
HRSRC hres = FindResource(NULL, sResName, sRestype);
HGLOBAL hbytes = LoadResource(NULL, hres);
LPVOID pdata = LockResource(hbytes);
LPBYTE sData = (LPBYTE)pdata;
LPTSTR sXml = (LPTSTR)sData;
wstring strXml(sXml);
In the above code i am not getting hres pointer.
The address of hres is null. Do u have any idea why I am not getting the pointer to resource file using FindResource(). I have included the resource file in the header files folder.
Is there any better way of doing the same?
Thanks in advance
JK
|
|
|
|
|
2 reasons I see for FindResource to fail -
1) A resource of type "MY_RESOURCE" with an integer ID of 103 isn't in the exe's resources.
2) The resource isn't compiled/linked into the exe file (since you've used NULL for hModule param)
Mark
"Go that way, really fast. If something gets in your way, turn."
|
|
|
|
|
Hi Mark,
Thanks for the reply.
I have checked the ID number its correct.
And I have included resource file in the workspace and I have included resource header in my *.cpp.
I could complie successfully. When I debug and see the return type of findresource() it is NULL.
Do u any idea how to fix this.
Thanks in advance.
JK
|
|
|
|
|
It looks like it should work.
Without seeing the resource script I can't imagine.
You could try AfxGetInstanceHandle() instead of NULL for the module handle.
Also make sure the type is a string and not a macro for an integer.
That's all I can think of off hand...
Mark
This episode brought to you by the number 5
|
|
|
|
|
i am very bad in strings and LPTSTR all that crap
so i switched to using the class CString
simply type
CString MyString;
if you want to put something into your Cstring, type
MyString = "this is my string"<br />
if you want to pass your CString into a LPTSTR field, use MyString.GetBuffer()
switch to CString today!
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
