|
Checking to see if anyone as any good comment to this scenario...
Using VS2005 to convert a server app to 64 bit. Running it on a Xeon ~3 GHz with 2 GB RAM etc. Since it is Xeon, we are talking EM64T and not AMD which might be the cause of this strange result:
In debug compilation, x64 wins with 30%
In release build, x86 wins with about 10% which seems pretty odd but there's nothing I can do about it.
I've tried to check just about any project configuration option I can possibly think of, but the same result. Obviously something is different between release and debug, but is it VS2005 or am I missing something? The same project config in all cases too. Could it be that VS2005 optimizes it's release code for AMD64 and that EM64T is different is the cause of the difference?
TIA
|
|
|
|
|
There is a C/C++ command line option: /favor:AMD64, but it is not enabled by default, nor would it explain your results.
http://msdn2.microsoft.com/en-US/library/ms173505(VS.80).aspx[^]
Burtraskkutaren wrote: In debug compilation, x64 wins with 30%
In release build, x86 wins with about 10%
Could it be that VS2005 optimizes it's release code for AMD64 and that EM64T is different is the cause of the difference?
There isn't enough information about your code/compiler settings to say.
My own informal tests with my code shows that my x86 and x64 builds perform about the same (i.e. release x86 ~= x64, debug x86 ~= x64).
...cmk
Save the whales - collect the whole set
|
|
|
|
|
Yep, I know, it really shouldn’t be that much of a difference but still it is. And the +30% in debug performance is completely accurate, no doubt about it. I’m going to check out the suggestions above as well as testing of different amount of simultaneous connected clients today. A little more info about the test environment:
The application is a kind of a webbserver, working with more or less large XML files. This actual version is single threaded, using MFC sockets. That means that CPU load is ~25%, Dual Xeon with HTT.
Testing is done through ACT (Microsoft Application Center Test) where a simulated surf action is recorded. Of course, same test in both cases. The biggest amount of testing has been done with 50 clients, where the results are those I’ve already mentioned. During testing in debug mode, CPU load is pretty even throughout the test but when running in release it seems to differ a bit more with periods of much less load. Preliminary testing yesterday, which I’m going to follow up now this morning, seemed to show that more normal performance was obtained if the number of clients in ACT was reduced to 1.
I’ll try to return later today with those results, the compiler settings isn’t so much to argue about since I’ve even tried to have the same settings in debug/release without any real measurable difference.
Release 64: /O2 /Ob1 /I "." /I "..\\" /I "..\..\private\include" /I "..\..\public\include" /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /D "_VC80_UPGRADE=0x0710" /D "_MBCS" /GF /FD /EHsc /MT /Gy /Yu"stdafx.h" /Fp".\R/ErrorService.pch" /Fo".\R/" /Fd".\R/" /W3 /nologo /c /errorReport:prompt
release 32: /O2 /Ob1 /I "." /I "..\\" /I "..\..\private\include" /I "..\..\public\include" /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /D "_VC80_UPGRADE=0x0710" /D "_MBCS" /GF /FD /EHsc /MT /Gy /Yu"stdafx.h" /Fp".\R/ErrorService.pch" /Fo".\R/" /Fd".\R/" /W3 /nologo /c /errorReport:prompt
Debug 64: /Od /I "." /I "..\\" /I "..\..\private\include" /I "..\..\public\include" /D "WIN32" /D "_DEBUG" /D "_WINDOWS" /D "_VC80_UPGRADE=0x0710" /D "_MBCS" /FD /EHsc /RTC1 /MTd /Yu"stdafx.h" /Fp".\D/ErrorService.pch" /Fo"x64\Debug\\" /Fd".\D/" /W3 /nologo /c /Wp64 /ZI /errorReport:prompt
Debug 32: /Od /I "." /I "..\\" /I "..\..\private\include" /I "..\..\public\include" /D "WIN32" /D "_DEBUG" /D "_WINDOWS" /D "_VC80_UPGRADE=0x0710" /D "_MBCS" /FD /EHsc /RTC1 /MTd /Yu"stdafx.h" /Fp".\D/ErrorService.pch" /Fo".\D/" /Fd".\D/" /W3 /nologo /c /Wp64 /ZI /errorReport:prompt
|
|
|
|
|
All-in-all i would find it hard to measure a web-server type app, where caching can play such a big part of the performance ... and there are so many places where data gets cached.
You say the version being tested is single-threaded, so i take it the network code is using select or wait on mutexes as opposed to IOCP. Maybe there is some difference in the x64 winsock library, or the x64 driver for your network card.
I would first try a console based numerical comp app and compare those runs first to try and narrow down what subsystem (e.g. networking, GUI, OpenGL, ...) is causing the difference.
I would (have) bite the buttlet and go unicode (_UNICODE instead of _MBCS) as (almost) all internal OS data is transmitted using unicode.
You may also want to look at:
http://msdn2.microsoft.com/en-us/library/9yb4317s(VS.80).aspx[^]
In particular you have /D WIN32 for your x64 builds, the bottom of the above MSDN page states "Values of WIN32 are replaced by WIN64" ... might make a difference in some of the win headers (doubt it though).
I also ended up blowing away the sln the upgrade wizard created, wasn't happy with it.
In general the Property Manager hasn't been very good in setting up 32/64bit builds, i ended up having to edit the sln and vsproj files directly to get what i needed.
For reference here are (short version) settings for one of my apps:
Release 64:
/Ox /Ob1 /Oi /Ot /GT /GL /D "WIN64" /D "NDEBUG" /D "_WINDOWS" /D "_UNICODE" /D "UNICODE" /GF /FD /MD /GS- /fp:fast /GR- /W4 /nologo /c /Wp64 /Zi /TP /errorReport:prompt /favor:AMD64
Release 32:
/Ox /Ob1 /Oi /Ot /GT /GL /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /D "_UNICODE" /D "UNICODE" /GF /FD /MD /GS- /arch:SSE2 /fp:fast /GR- /W4 /nologo /c /Wp64 /Zi /TP /errorReport:prompt
Debug 64:
/Od /D "WIN64" /D "_DEBUG" /D "_WINDOWS" /D "_UNICODE" /D "UNICODE" /GF /FD /MDd /GS- /fp:fast /GR- /W4 /nologo /c /Wp64 /Zi /TP /errorReport:prompt /favor:AMD64
Debug 32:
/Od /D "WIN32" /D "_DEBUG" /D "_WINDOWS" /D "_UNICODE" /D "UNICODE" /GF /FD /MDd /GS- /fp:fast /GR- /W4 /nologo /c /Wp64 /Zi /TP /errorReport:prompt
Good-luck
...cmk
Save the whales - collect the whole set
|
|
|
|
|
A few answers so far...
/favor:EM64T only performed worse.
Network code, well haven't dug into it deeply enough but it's all into MFC, and how they have solved it I don't know.
Unicode, well maybe later since it generates a couple of hundred compile errors right now 
/D Win32, seems like it gives slightly better performance than /Win64...
I did test yesterday with some special variant of the code, which generated more reasonable results (+10-20% for 64bit in release) but I'll have to keep on digging deeper to find the possible solution to it. One theory of mine is that there might be some conversions / larger datatypes that messes things up although it would be most logical if it would generate same performance debug/release...
It might be possible that it is the Winsock library, I doubt that it is the NIC because NIC drivers have to be 64-bit in a 64-bit OS to work. All tests are done in Win2003Server x64...
Keeping on searching for an answer 
|
|
|
|
|
Hi there,
I would like to create a message-only window to handle messages from external applications. But how do I go about doing that? I couldn't possibly run a message loop in a DLL after I have created a window, can I?
// after CreateWindow is successful
while ((bRet = GetMessage(&msg, NULL, NULL, NULL)) != 0)
{
if (bRet == -1)
{
// some code ...
}
else
{
TranslateMessage(&msg);
DispatchMessage(&msg);
}
}
|
|
|
|
|
if i am correct are you trying to hook the other applications..???
and it is not possible to write the message loop into the win32 DLL as its drawback of dll
Knock out "T" from CAN'T
You 'CAN' if you think you 'CAN'
|
|
|
|
|
Hi all,
We have a class CBitmapButton which is derived from cButton. We need to enable the tool tip for this class. We have written the code for the same.
We have followed all the steps required to enable a tool tip for this button. However we are facing a problem while displaying the tool tip. The tool tip is seen at first when the mouse is moved to this button. If we hold the mouse on that button, after some time, the tool tip vanishes and then the tool tip is never seen even if the mouse is moved to the next button and brought back to this button .
Please suggest some solution to the above problem.
Any kind of help would be appreciated.
Thanks
amit
|
|
|
|
|
can you give he short code snippet what you are doing so that someone can find out where you are gong wrong
Cheers
"Peace of mind through Technology"
|
|
|
|
|
|
Hi all,
Thanks for the help. I tried the following things one by one and checked the output
1. Capture the mouse event and used TrackMouseEvent
2. Called InitToolTip function in PreTranslate before RelayEvent
3. Set the flag to TTF_SUBCLASS while creating the tool tip control.
However, I am still not getting the tool tip after the time out.
I shall post my code in a short while. Please look at it and let me know what is wrong in it
Thanks again for the help provided.
Regards,
amit
|
|
|
|
|
Given below are the steps for enabling the tool tip for the CBitmapButton
1. The class CTipBitmapButton is derived from CButton. This has teh owner's draw property set to TRUE.
2. An object of CToolTipCtrl m_ToolTipCtrl is created.
3. m_ToolTipCtrl is initialised in the constructor
4. The tool tip is initialised by using following code:
m_ToolTipCtrl = new CToolTipCtrl;
m_ToolTipCtrl->Create(this);
m_ToolTipCtrl->Activate(TRUE);
m_ToolTipCtrl->SetFont(GetFont(), FALSE);
5. New function DisplayToolTip is called in the DrawItem function of the class CTipBitmapButton
6. The function DisplayToolTip has the following code
a. Delete the tool tip if it exists
b. Set the tool tip using SetToolTipText function
c. Set the delay time to TTDT-INITIAL
7. The function SetToolTipText has the following code
a. Initialise the tool tip
b. If the tooltip tool count is 0, then m_ToolTipCtrl->AddTool(this, rect, 1)
c. m_ToolTipCtrl->UpdateTipText(tooltext,this,1)
d. m_ToolTipCtrl->Activate(TRUE);
8. In PreTranslateMessage
m_ToolTipCtrl->RelayEvent(pMsg);
With this code, the tool tip does not come after the time out. Please suggest.
Any kind of help would be appreciated.
Thanks,
amit
|
|
|
|
|
Given below are the steps for enabling the tool tip for the CBitmapButton
Please note that this code is the original code written before trying out options of TrackMouseEvent and TTF_SUBCLASS
1. The class CTipBitmapButton is derived from CButton. This has teh owner's draw property set to TRUE.
2. An object of CToolTipCtrl m_ToolTipCtrl is created.
3. m_ToolTipCtrl is initialised in the constructor
4. The tool tip is initialised by using following code:
m_ToolTipCtrl = new CToolTipCtrl;
m_ToolTipCtrl->Create(this);
m_ToolTipCtrl->Activate(TRUE);
m_ToolTipCtrl->SetFont(GetFont(), FALSE);
5. New function DisplayToolTip is called in the DrawItem function of the class CTipBitmapButton
6. The function DisplayToolTip has the following code
a. Delete the tool tip if it exists
b. Set the tool tip using SetToolTipText function
c. Set the delay time to TTDT-INITIAL
7. The function SetToolTipText has the following code
a. Initialise the tool tip
b. If the tooltip tool count is 0, then m_ToolTipCtrl->AddTool(this, rect, 1)
c. m_ToolTipCtrl->UpdateTipText(tooltext,this,1)
d. m_ToolTipCtrl->Activate(TRUE);
8. In PreTranslateMessage
m_ToolTipCtrl->RelayEvent(pMsg);
With this code, the tool tip does not come after the time out. Please suggest.
Any kind of help would be appreciated.
Thanks,
amit
|
|
|
|
|
Hi all,
I am sending you the code snippet i have written.
Hope you will get better idea with this.
Given below are the steps for enabling the tool tip for the CBitmapButton
1. The class CTipBitmapButton is derived from CButton. This has teh owner's draw property set to TRUE.
2. An object of CToolTipCtrl m_ToolTipCtrl is created.
3. m_ToolTipCtrl is initialised in the constructor
4. The tool tip is initialised by using following code:
m_ToolTipCtrl = new CToolTipCtrl;
m_ToolTipCtrl->Create(this);
m_ToolTipCtrl->Activate(TRUE);
m_ToolTipCtrl->SetFont(GetFont(), FALSE);
5. New function DisplayToolTip is called in the DrawItem function of the class CTipBitmapButton
6. The function DisplayToolTip has the following code
a. Delete the tool tip if it exists
b. Set the tool tip using SetToolTipText function
c. Set the delay time to TTDT-INITIAL
7. The function SetToolTipText has the following code
a. Initialise the tool tip
b. If the tooltip tool count is 0, then m_ToolTipCtrl->AddTool(this,
rect, 1)
c. m_ToolTipCtrl->UpdateTipText(tooltext,this,1)
d. m_ToolTipCtrl->Activate(TRUE);
8. In PreTranslateMessage
m_ToolTipCtrl->RelayEvent(pMsg);
With this code, the tool tip does not come after the time out. Please suggest.
amit
|
|
|
|
|
If you are manually passing mouse movement messages to the tooltip, be sure that you are doing it from the parent window of the CBitmapButton, and not from the button control itself, or use _TrackMouseEvent(...) to indicate to the tooltip control when you have moved off of the button (send it a POP message).
My guess is that the tooltip is not getting messages when you move off of the control, so it still thinks the mouse is over the button (standing still), and will not reset its state to display itself again.
Alternately, you could use a little-known (or little-used) feature of the ToolTip control, the TTF_SUBCLASS flag when creating the tooltip to implement "fire-and-forget" tooltip support, where you do not have to handle passing messages around.
Peace!
-=- James
If you think it costs a lot to do it right, just wait until you find out how much it costs to do it wrong!
Avoid driving a vehicle taller than you and remember that Professional Driver on Closed Course does not mean your Dumb Ass on a Public Road!
DeleteFXPFiles & CheckFavorites (Please rate this post!)
|
|
|
|
|
|
how can i monitor all register sub keys in the selected root regiter key.
and display the register key name,value ans status in list control.
how can i do it?
|
|
|
|
|
|
You can use :
char lszValue[255];
CString str ;
HKEY hKey;
LONG returnStatus;
DWORD dwType=REG_SZ;
DWORD dwSize=255;
returnStatus = RegOpenKeyEx(HKEY_LOCAL_MACHINE, "Key path separeted by \\", 0L, KEY_ALL_ACCESS, &hKey);
if (returnStatus == ERROR_SUCCESS)
{
returnStatus = RegQueryValueEx(hKey,"value you want to read", NULL, &dwType,(LPBYTE)&lszValue, &dwSize);
if(returnStatus == ERROR_SUCCESS)
{
str = (LPCTSTR)lszValue;
}
}
RegCloseKey(hKey);
Cheers
"Peace of mind through Technology"
|
|
|
|
|
kiranin wrote: how can i monitor all register sub keys in the selected root regiter key.
By using RegNotifyChangeKeyValue().
"Let us be thankful for the fools. But for them the rest of us could not succeed." - Mark Twain
"There is no death, only a change of worlds." - Native American Proverb
|
|
|
|
|
Plz Can U explain the procedure......how i willl do it.?
|
|
|
|
|
See here.
"Let us be thankful for the fools. But for them the rest of us could not succeed." - Mark Twain
"There is no death, only a change of worlds." - Native American Proverb
|
|
|
|
|
PabloN wrote: ...when i move my groupbox the radio button won't.
This is normal behavior with VS v6. Has that changed with VS 2005?
"Let us be thankful for the fools. But for them the rest of us could not succeed." - Mark Twain
"There is no death, only a change of worlds." - Native American Proverb
|
|
|
|
|
I build a dialog based MFC program.
in the file "StdAfx.h" I add the following code,which is used to define a global variable of type CString.
#ifndef jjjjjjjjjjjjjjjjjjj
#define jjjjjjjjjjjjjjjjjjj
CString StrGlobal;
#endif
and then I use it in another file "newfile.cpp"
#include "stdafx.h"
extern CString StrGlobal;
void fun()
{
StrGlobal = _T("KDKDK");
}
when I build the program, following link error shows up:
Linking...
StdAfx.obj : error LNK2005: "class CString StrGlobal" (?StrGlobal@@3VCString@@A) already defined in newfile.obj
why dose this happen? and how can I fix it ?
Thank you very much!!!
-------------------
I am learning C++ and English
|
|
|
|
|
You've got the extern in the wrong file: It should be in the .h file.
Steve
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
