Re: Shared Memory Question

New Message Reply Date view Thread view Subject view Author view

Angus Dorbie (dorbie++at++sgi.com)
Tue, 20 Jul 1999 01:44:56 -0700

The shared structure approach can work but you should set the DCS
position in the app after you send the position in the vec. You also
defeat the purpose of having the pointer in shared memory by newing off
the heap. That's never going to work. Eliminate the indirection in the
struct declaration adding the vec data to the shared structure.

The second suggestion you make is also worrying. The stack will have
completely different data. The pointer could work provided you asign it
before the pfConfig AND allocate it from shared memory. I suspect you
don't.

Joaquín Casillas wrote:
>
> Steve Woolsey escribió:
> >
> > I have an app I am doing in C++ that is forking a motif process and
> > another process to calculate the movements for a number of object. I am
> > trying to get the location of the object to pass back to the draw
> > process, so I can update the pfDCS's. I have tried three different
> > approaches, with no success.
> >
> > 1. I tried putting a pfVec3 into the Shared struct.
> >
> > typedef struct {
> > pfPipeWindow *pw;
> > pfChannel *chan;
> > pfScene *scene;
> > int exitFlag;
> > pfDCS **objectDCS;
> > pfVec3 *objectLocation;

No, just use pfVec3 objectLocation; and access members with . instead of
->

> > int numObjects;
> > }SharedData;
> >
> > Then I tried declaring the pointer with:
> >
> > Shared->objectLocation = new pfVec3[numObjects];
> >

This won't work, the new is going to create unshared data, so the
pointer will be shared but will point to invalid memory in other
processes.

> > 2. Tried declaring a pfDataPool in the main process, and then attaching
> > the movement process.
> >
> > pfDataPool *pool;
> > pfVec3 *objectLocation;
>
> Steve, another approach may be to fork a compute process that calculates
> the movements of the objects and then use the mechanism of
> pfFlux/pfEngine/pfFCS (use pfFCS's instead of pfDCS's for the moving
> objects).

The problems seem considerably more basic than this. There are
fundamental problems with his notion of how shared memory should be
created w.r.t. the fork.

Cheers,Angus.

-- 
Never express yourself more clearly than you think.
                                       Neils Bohr 


For advanced 3D graphics Performer + OpenGL based examples and tutors:
http://www.dorbie.com/

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Tue Jul 20 1999 - 01:45:00 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.