|
I'm building in Microsoft Visual Studio 2003, using C++ (so, it's v 7.1).
Here is my scenario, consider each situation be repeated ad infinitum:
Child thread performs "new A", parent thread performs "delete A" - no problem.
Parent thread performs "new B", child thread performs "delete B" - virtual memory grows uncontrollably over time, consuming my computer in an almost fiery death. The working set grows and gets trimmed over time as would be expected.
Some additional clarification - I can actually watch the VM size increase and decrease in the first situation (using both Task Manager and PerfMon). It appears to monotonically increase in the second situation. I'm just baffled why virtual memory is contained going the first direction, but not the second.
Any ideas?
|
|
|
|
|
Maybe it is spawning lots of threads? That would consume system resources as it is on the O/S level that manages this.
VM are not usually the problem in MT programs - usually it is odd crashes, when some var is not properly locked or we've screwed up some syncronization.
A cynic is a man who, when he smells flowers, looks around for a coffin.
-H.L. Mencken
|
|
|
|
|
Your Problem is somewhat incomprehensible. Give us some actual (tested) code.
Virtual memory is not something you have control over or should be concerned about! Your program only allocates memory! The operating system decides which part of all memory used should be virtual, your program has no control over that. If your monitor gives a sudden and sharp increase in virtual memory usage untill death, that means invariably that you ended up in an endless loop allocating memory. This has nothng to do with virtual memory anagement! Investigate all for, do,while loops in which memory is (indirectly) (m)allocate(d) Thats where the problem is!
Hint:
Debugging can be troublesome for beginners like me, but here'se a start
Put in the Header File: (Best in stdafx.h, at the bottom)
#define malloc(x) MyMalloc((x),__LINE__,__FILE__)
and write a function:-
void* MyMalloc(size_t Size,int LineNr,LPCSTR FileName){
CString Msg;
Msg.Format("malloc(%i) in File:%s, Line %i\n",FileName,LineNr);
size_t StrLen=Msg.GetLength();
size_t BufSize=StrLen+1>Size ?StrLen+1:Size;
void* pRes=malloc(BufSize+255);
if(pRes==NULL)return NULL:
strcpy(pRes,(LPCSTR)Msg);
return pRes;
}
Avoid writing anything to the objects you Use
Compile and Run the Program, when it dies, it gives a trace log in the bottom of the screen, of all blocks not freed"
Take it from there
LateNightsInNewry
|
|
|
|
|
N.B Its obvious, but before the above function, write:
#undef malloc. Else you have an Endless Loop in itself(think)
Also, write a similar function for calloc,new, etc
success
LateNightsInNewry
|
|
|
|
|
Disregard the question. I found my problem. It was an actual memory leak, due to lack of a call to delete in one unique case that wasn't being handled. Rookie mistake, my bad.
|
|
|
|
|
Just a quick question. When the exe is built as Multithreaded, should the dll's also be multithreaded or does it not matter?
|
|
|
|
|
all modules should be in parallel
Kuphryn
|
|
|
|
|
Yes, they should be. For such things as the C-Runtime library (the library that contains the implementation for things such as fopen and errno), there are slight differences in how some things are done. For example, in a single threaded CRTL DLL, memory management probably isn't locked so two threads don't try to allocate memory at the same time (due to the threads sharing the same core memory allocation tables). Even simple things such as errno and strtok are implemented differently in a single threaded CRTL.
Tim Smith
I'm going to patent thought. I have yet to see any prior art.
|
|
|
|
|
It depends. If you’re not sure just make all of them multithreaded.
If the main .EXE is multithreaded but only makes calls to a DLL using one of these threads and the DLL is single threaded itself then it need not be compiled with the multithreaded libraries.
Steve
|
|
|
|
|
Tell me if what I say below makes sense and is possible.
There is a program (matlab) that makes plots. Currently (as far as I can work it) it can either just display the plot or save it to the hardive. Well I want to display the plot on my separate application. Here are two ways I have done it - also, please note, the graph is real time data so it is refreshed every millisecond or so. When I was generating the graphs as displays (ie not save image to HD) I would - in my external app - take a screenshot of the plot and display it on my app. This was fast enough, etc but I doesn't work if the app is minimized or covered up. So, then I went towards the save image to HD method, this worked great, except it was kinda slow, especially if you're running it on a laptop harddrive.
So, currently both my solutions have their limitations. Here is my question. Ignoring whether the fact matlab even has the functionality to do this, I would like to save the plot to memory! Does this make sense? I just want to save the bytes somewhere in the RAM! Then I would like to read it in my external app and construct an image out of it. So, I guess serialize the image to memory then deserialize it to an image. Is this feasible?
Can anybody else think of an efficient way of gettting this plot from matlab?
P.S. The plot is hosted in a SunAwtCanvas (thanks to spy++)
/\ |_ E X E GG
|
|
|
|
|
Actually, I just found a function in matlab that lets you save the plot to the clipboard! So, I guess that's the easy way out...
/\ |_ E X E GG
|
|
|
|
|
Fundamental? Not to C or CPP development. Your question sounds like greek to me, I'm not the moderator of this forum, but I would think that posting Matlab questions here is stretching it a bit! I always viewed MatLab as 'Basic on Steroids', maybe I am a bit harsh here, but I do not think that you're in the right forum here, and that you may get better advice in a Matlab forum!
best of luck
LateNightsInNewry
|
|
|
|
|
typedef UpdateHandler* UpdateHandlerPtr;
I saw this code in a project file but I couldn't find what the definition of UpdateHandler is. It is not defined anywhere. Only thing I found on UpdateHandler is the code below
What does the syntax below do?
typedef void UpdateHandler(ObjectRootPtr, const RTI::AttributeHandleValuePairSet&);
|
|
|
|
|
Jay03 wrote: typedef void UpdateHandler(ObjectRootPtr, const RTI::AttributeHandleValuePairSet&);
That is the definition of UpdateHandler. This syntax is creating a function pointer. For more information, look up callback functions and function pointers.
If you decide to become a software engineer, you are signing up to have a 1/2" piece of silicon tell you exactly how stupid you really are for 8 hours a day, 5 days a week
Zac
|
|
|
|
|
Isn't it a declaration? It's interesting - I wonder how the compiler parsed out the fact that it is not typedef'ing void?
A cynic is a man who, when he smells flowers, looks around for a coffin.
-H.L. Mencken
|
|
|
|
|
The problem is that function pointer syntax is bizarre (sorry guys it is). In this example you are typdefing a function pointer that returns void and two parameters (ObjectRootPtr, const RTI::AttributeHandleValuePairSet&) to the name UpdateHandler. It does look wierd, but it works.
On those occasions I need function pointers, I always have to look at an example because I can never quite remember it.
|
|
|
|
|
And it's going to show up on one of the awful employment tests that they give us.
A cynic is a man who, when he smells flowers, looks around for a coffin.
-H.L. Mencken
|
|
|
|
|
From experience, I can tell you that it will if you ever apply to Amazon.com. Just an FYI.
If you decide to become a software engineer, you are signing up to have a 1/2" piece of silicon tell you exactly how stupid you really are for 8 hours a day, 5 days a week
Zac
|
|
|
|
|
Thanks,
This brought some clarity on function pointer to me and i think this is a better of doing it. I hope this method is supported in most all compilers.
typedef int FunctPtr(int,int);
....
main
{
FunctPtr *fp = somefunction; //This declaration is clearer that fp variable is a pointer.
fp(...);
}
typedef int (*FuctPtr1)(int, int); //sometimes looks confusing as what is happening.
and
FuctPtr1 c = somefunc;
-- modified at 4:03 Thursday 17th August, 2006
-- modified at 4:04 Thursday 17th August, 2006
-Prakash
|
|
|
|
|
Zac Howland wrote: This syntax is creating a function pointer
Won't it be typedef void UpdateHandlerPtr (ObjectRootPtr, const RTI::AttributeHandleValuePairSet&);
or typedef void (UpdateHandler*) (ObjectRootPtr,const RTI::AttributeHandleValuePairSet&);
|
|
|
|
|
typedef void UpdateHandler(ObjectRootPtr, const RTI::AttributeHandleValuePairSet&);
is syntaxically equivalent to writing
typedef void (*UpdateHandlerPtr)(ObjectRootPtr, const RTI::AttributeHandleValuePairSet&);
If you decide to become a software engineer, you are signing up to have a 1/2" piece of silicon tell you exactly how stupid you really are for 8 hours a day, 5 days a week
Zac
|
|
|
|
|
I Suggest you read the Book by Brian W. Kernighan & Dennis M. Ricthie: the C Programming Language This book is Much maligned, but actually very good! It spells out the Syntax and methods of the C Language, without going into the legalities, like the ANSI standard does! On page 122 of my copy it gives paragraph 5.12- Complicated Declarations. It explains there in lucid details how to read your declaration, and gives the code for a simple program which takes a declaration, and translates it back into 'english'
Success
LateNightsInNewry
|
|
|
|
|
Hello guys,
I am quite new to mfc and it seems there are tons of questions about it and little that I can get out of the documentation provided by microsoft, so hope someone with more exp can answer it for me.
Right now, I have several dialog bar in my app, all have been set according to how I want it to be, except, one thing that confuse me, when I drag one bar, I can't seem to reflect the mouse position onto the other bars, how do I obtain the mouse position while still dragging a dialog bar, from another dialog bar. thx...
LE
|
|
|
|
|
How about using GetCursorPos() which fills out a POINT structure
|
|
|
|
|
I did tried with GetCursorPos(), but fails and it seems that this would be the case,
When a user drag the dialog bar around, the mouse is captured by this bar, so any attempt to have the mouse sent any information to another dialog bar would be impossible.
I have written functions overriding those MouseDown, MouseUp, etc, to call GetCursorPos() but it still fail. Once the mouse is clicked and drag on a bar, these functions will not be called.
I wanted to detect the mouse, while it is dragging a bar, when it enters/over another bar...so the first bar would be dragged around, and the second bar will be the one doing the detection.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
