Socket and multi-core troubles

Question

I have a 'hybrid' application as its refered to in the docs
that I use to generate 3D scenes with the sandy flash library, and
use those as motion graphics on top of a live video source for
recording and broadcasting.
The Camera sends hello to the flash socket, the flash app,
draws the scene, the socket sends a byte array to the camera where
it is alpha blended to into the live camera.
On a single core system, there are no issues.
On a dual core system, I do not get the whole byte array.
Some how it seems in the middle of getting the bitmap to the array,
something stops the writing and the socket is flushed anyway.
I am receiving all 307200 bytes to the camera, but image is
not fully drawn to the byte array. It results in a random
quartering of the image, depending on how many bytes are actually
sent.
another behavior noted is when I drag the window, the socket
still responds to the c++ code calling it, still sends 307200
bytes, all '0''s, while nothing is drawn to bytes in the getPixels
call.
I can live with this behavior, but not the former.
quote:
//b bit mapdata, bytes is byte array
b.draw(m_iDrawable);
bytes=b.getPixels(new Rectangle(0,0,320,240));
writeBytes(bytes,0,307200);
flush();
Is there any info regarding multi core processors
interrupting drawing or socket communication? Even passing static
bitmapdata that is never changes or displayed on the stage is
interupted.
The socket seems to be broken on the multicore machine.
The c++ application that is creating the broadcast/file does
receive the full frame buffer size, other wise it would hang until
the full 307200 bytes are recieived. It appears that flash stops
writing, and continues in the next call.
There appears to be no handle to gauge how much data is
actually passed to the byte array.
I can set up live demos if desired.
Single core machine and dual core machine running exact same
software, and then stream the output graphics over rtmp.
Help! Suggestions! Limiting the app to singe core machines is
not an option!

Answer 1

not Vista on the newer machine, right? (since Vista has
TCP/IP stack problems)
No, thanks.... Not vista. this is the working solution.
quote:
// in the enterframe event...
do{}while(vidOut.m_bSharedState)
// in the render event
do{}while(vidOut.m_bSharedState)
//this is in the function triggering the whole affair
//Whether clone or lock is still nessecary I dont know
if(fillBuffer)return;
fillBuffer=true;
if(!vidOut)return;
do{}while(vidOut.m_bSharedState)
vidOut.readBytes(new ByteArray(),0,vidOut.bytesAvailable);
vidOut.draw(this.sprite());
vidOut.flush();
do{}while(vidOut.m_bSharedState)
fillBuffer=false;
//and this is the video out object's relevant function
m_bSharedState=true;
b=new BitmapData(320,240,true,0);
b.lock();
b.draw(p_iDrawable);
var bmc:BitmapData=b.clone();
bytes=bmc.getPixels(rec);
bytes.position=0;
writeBytes(bytes,0,307200);
b.unlock(rec);
m_bSharedState=false;
Currently running at 15 fps encoding sesions with 0 drops as
it should be.
I'll throw up a temporary demo today in a bit.
Thank you for your response!
Solved