I'm trying to do a single-instance filter, which would be able to access some common global data across different applications which instantiate it. Could anyone advise me on the architecture to achieve that? For example, now I want my "Bouncing Ball" to run exactly the same (i.e the ball at the same position) when I run it in Graph Edit and in my application simultaneously.
I've tried modifying the CreateInstance() inside the BouncingBall, so that it becomes a singleton design, but that only works when I open two BouncingBall in GraphEdit (i.e. from the same app). The moment I open it in my app it would create and return another instance of Bouncing Ball.
> I'm trying to do a single-instance filter, which would be > able to access some common global data across different > applications which instantiate it. Could anyone advise me > on the architecture to achieve that? > For example, now I want my "Bouncing Ball" to run exactly > the same (i.e the ball at the same position) when I run > it in Graph Edit and in my application simultaneously.
> I've tried modifying the CreateInstance() inside the > BouncingBall, so that it becomes a singleton design, but > that only works when I open two BouncingBall in GraphEdit > (i.e. from the same app). The moment I open it in my app > it would create and return another instance of Bouncing > Ball.
This is not really a DirectShow question. Anyway...
You need to use a shared memory buffer and a pair of named global mutex and event for synchronization. It's up to how to generate the data and synchronized the processes.
You can create the mutex and event objects using CreateMutex() and CreateEvent() and giving them a name (better use unique names, e.g. string that contain some GUID, like your own CLSID).
You can create a shared memory buffer using the memory-mapped file APIs (CreateFileMapping(); again, better choose a unique name) or enclosing the external buffer declaration in a data_seg() #pragma directive (which actually instructs the compiler to create a memory-mapped file for you):
> > I'm trying to do a single-instance filter, which would be > > able to access some common global data across different > > applications which instantiate it. Could anyone advise me > > on the architecture to achieve that? > > For example, now I want my "Bouncing Ball" to run exactly > > the same (i.e the ball at the same position) when I run > > it in Graph Edit and in my application simultaneously.
> > I've tried modifying the CreateInstance() inside the > > BouncingBall, so that it becomes a singleton design, but > > that only works when I open two BouncingBall in GraphEdit > > (i.e. from the same app). The moment I open it in my app > > it would create and return another instance of Bouncing > > Ball.
> This is not really a DirectShow question. Anyway...
> You need to use a shared memory buffer and a pair of named > global mutex and event for synchronization. It's up to how > to generate the data and synchronized the processes.
> You can create the mutex and event objects using > CreateMutex() and CreateEvent() and giving them a name > (better use unique names, e.g. string that contain some > GUID, like your own CLSID).
> You can create a shared memory buffer using the > memory-mapped file APIs (CreateFileMapping(); again, better > choose a unique name) or enclosing the external buffer > declaration in a data_seg() #pragma directive (which > actually instructs the compiler to create a memory-mapped > file for you):
> > On Fri, 10 Dec 2004 12:13:00 +0800, haipt wrote:
> > <snip> > >> your answer Allesandro.
> > Alessandro: still better than Alessandra <g>
> I always told my mother "Alessandro" was too long a name and > she should have chosen a shorter and simpler one :-))
> --
> // Alessandro Angeli > // MVP :: Digital Media > // a dot angeli at psynet dot net
I'm sorry to misspell your name, Alessandro (a careful programmer should never misspell anything actually :) ). Anyway, your suggested method works really well in my case. Initially i declared like this: #pragma data_seg("SHARED") char buf[1000]; #pragma data_seg() #pragma comment(linker, "/section:SHARED,RWS") (i'm using VC.NET 2003)
Then the two processes did see the same buf's global address, but somehow whenever one updated it, the other could not see the update. After a good few hour tracing the problem, i realized that it's due to my buf's not initialized. So I changed it to char buf[1000]="", and from there on it works perfectly for me. I've got no explanation except the guess that it's due to some linker's internal decision.
haipt wrote: > After a good few hour tracing the problem, i realized > that it's due to my buf's not initialized. So I changed > it to char buf[1000]="", and from there on it works > perfectly for me. I've got no explanation except the > guess that it's due to some linker's internal decision.
I don't think that's the case: when you declare an array pointer but you do not initialize it, you're only reserving space for the pointer while, when you initialize it, you are reserving space for the array but not the pointer. In the first case, your processes were sharing the pointer's memory location but not the actual buffers which were allocated outside shared memory unless you did some memory-mapped file "magic" of you own. This scenario was also dangerous because one process would end up using the other's pointer value which was not necessarily a valid pointer in its own memory space.
> So thanks again, Alessandro.
You're welcome :-)
--
// Alessandro Angeli // MVP :: Digital Media // a dot angeli at psynet dot net
|
