Problem with free up memory! - C# Discussion Boards

7-Oct-08 23:24

Hi everybody,

I'm developing a network-based application. It is running on two stations. One station is sending 4separate & simultaneous voice streams
to the other station which in turn plays those streams. The problem is that memory consumption is high(4 MB/s) at the second station.
I've used Memoryobject.Dispose() to free up some memory but it seemed useless. Is there any other way to free up memory? Or am I using the Dispose method inappropriately?

Here is my code to play the sound streams:

<br />
private void Receive()<br />
        {<br />
            try<br />
            {<br />
                /*BufferDescription*/<br />
                BufferDescription playbackBufferDescription = new BufferDescription();<br />
                playbackBufferDescription.BufferBytes = waveFormat.AverageBytesPerSecond / 5;<br />
                playbackBufferDescription.Format = waveFormat;<br />
<br />
                bbStop = false;  <br />
                IPEndPoint remoteEP = new IPEndPoint(IPAddress.Any, 0);<br />
<br />
                byte[] byteData;<br />
                byte[] byteDecodedData;<br />
<br />
                /*While the voice stream is being recieved */<br />
                while (!bbStop)   <br />
                {<br />
		    //Receive data.<br />
		    byteData = udpClient.Receive(ref remoteEP);<br />
                    byteDecodedData = new byte[byteData.Length * 2];<br />
                    <br />
                    //I've used G711 for voice stream compression<br />
                    //G711 compresses the data by 50%, so I allocate a buffer of double<br />
                    //the size to store the decompressed data.<br />
                    //byteDecodedData = new byte[byteData.Length * 2];<br />
<br />
                    //Decompress data using the proper vocoder.<br />
                        ALawDecoder.ALawDecode(byteData, out byteDecodedData);<br />
<br />
                    //Play the data received to the user.<br />
                    playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device);<br />
                    playbackBuffer.Write(0, byteDecodedData, LockFlag.None);<br />
                    playbackBuffer.Play(0, BufferPlayFlags.Default);<br />
<br />
                    first = false;<br />
                    //byteData = null;<br />
                    playbackBuffer=null;<br />
                    //byteDecodedData = null;<br />
                }<br />
                <big>playbackBuffer.Dispose();</big>   //Here is the problem! It doesn't work! :confused:<br />
            }<br />
            catch (Exception ex)<br />
            {<br />
                MessageBox.Show(ex.Message, "Error on Receive", MessageBoxButtons.OK, MessageBoxIcon.Error);<br />
            }<br />
            <br />
        }

Thanks in forward

Simon P Stevens8-Oct-08 0:11

8-Oct-08 0:11

private void Receive()
{
    try
    {
        /*BufferDescription*/
        BufferDescription playbackBufferDescription = new BufferDescription();
        playbackBufferDescription.BufferBytes = waveFormat.AverageBytesPerSecond / 5;
        playbackBufferDescription.Format = waveFormat;

        bbStop = false;
        IPEndPoint remoteEP = new IPEndPoint(IPAddress.Any, 0);

        byte[] byteData;
        byte[] byteDecodedData;

        /*While the voice stream is being recieved */
        while (!bbStop)
        {
            //Receive data.
            byteData = udpClient.Receive(ref remoteEP);
            byteDecodedData = new byte[byteData.Length * 2];

            //I've used G711 for voice stream compression
            //G711 compresses the data by 50%, so I allocate a buffer of double
            //the size to store the decompressed data.
            //byteDecodedData = new byte[byteData.Length * 2];

            //Decompress data using the proper vocoder.
            ALawDecoder.ALawDecode(byteData, out byteDecodedData);

            //Play the data received to the user.
** HERE ==> playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device);
            playbackBuffer.Write(0, byteDecodedData, LockFlag.None);
            playbackBuffer.Play(0, BufferPlayFlags.Default);

            first = false;
            //byteData = null;
            playbackBuffer = null;
            //byteDecodedData = null;
        }
        playbackBuffer.Dispose(); //Here is the problem! It doesn't work! Confused
    }
    catch (Exception ex)
    {
        MessageBox.Show(ex.Message, "Error on Receive", MessageBoxButtons.OK, MessageBoxIcon.Error);
    }
}

You are creating a new playback buffer every time you loop round the while loop (See annotation above). You are then only disposing it at the end (after you've set it to null, so I'm surprised you don't get a null ref exception. (Also, if an exception occurs, the playback buffer will never get disposed)

You should either be re-using the same playback buffer all the time to prevent having to create a new one each time, or you should be properly disposing of it each time.

You can use the "using" construct to ensure it gets disposed of. This will automatically call dispose on the playback buffer object when it passes the closing brace, and also ensures it will get disposed even if an exception occurs:

//Play the data received to the user.
using (playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device))
{
    playbackBuffer.Write(0, byteDecodedData, LockFlag.None);
    playbackBuffer.Play(0, BufferPlayFlags.Default);
}

Without seeing the rest of your code it's hard to tell, but if there is a way to just re-use the same playback buffer without creating a new one each time round the loop, that would be a better solution.

[By the way, If you use the <pre> "code block" tags instead, for large blocks of code it will be easier to read]

Simon

leppie8-Oct-08 1:53

leppie

8-Oct-08 1:53

Good spot

xacc.ide - now with TabsToSpaces support
IronScheme - 1.0 beta 1 - coming soon
((lambda (x) `((lambda (x) ,x) ',x)) '`((lambda (x) ,x) ',x))

Ha ha ha ha ha ha ha ha8-Oct-08 1:54

8-Oct-08 1:54

Thanks Simon for your useful answer. I tested what you offered, but anytime I tried, I lost the voices.

1.By using:

//Play the data received to the user.
using (playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device))
{
    playbackBuffer.Write(0, byteDecodedData, LockFlag.None);
    playbackBuffer.Play(0, BufferPlayFlags.Default);
}

There were no voice being played anymore.

2.By transferring:

playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device);

to outside of the while statement, the playing voice was cut just like 1010101010101 where the "1" character indicates parts where voice output was present. &therefore the "0" character represents the absence of voice output.

3.By disposing the playbackBuffer each time inside the while statement, again there exists no voice.

Note: the only part for playing voice is the routine above, &the rest is not in connection with the playing problem. So I don't think the rest code will be more helpful to send... Any other suggestions? I appreciate that!

Simon P Stevens8-Oct-08 3:06

8-Oct-08 3:06

I've never used directsound (I'm assuming that is what you are using), so I'm not entirely sure about all of this.

With 1 - The problem is likely to be that you are disposing of the buffer before it has a chance to do any playback.

With 2 - The problem is likely to be that you are calling play() while the buffer is still busy playing the previous bit (I'm assuming the play method is asynchronous). Hence it cuts.

What you probably need to do is remove the call to Play(). You need to track how much is in the buffer. When there is more than 5 seconds of data in the buffer, call play, and set a flag to prevent it calling play again. When the buffer runs out, clear the flag, and allow play to be called again when the buffer fills back up to the 5 second mark. This way you prevent play from being called repeatedly while the buffer is still playing, and you also prevent it from being called when there is only a small amount of data in the buffer.

if (!isPlaying && timeInBuffer > 5)
{
    buffer.Play();
    isPlaying = true;
}
else
{
    isPlaying = // Check if buffer has finished playing
    timeInBuffer = // Check the amount of time left in the buffer
}

Simon

Ha ha ha ha ha ha ha ha8-Oct-08 3:51

8-Oct-08 3:51

Thanks again for the answer
But I don't consider this solution as an appropriate one, as it plays the buffer whenever it's full, but my application already plays the buffer without any problem or any cut in the voice, the problem is high consumption of memory, which is not likely to be solved with this solution.
And yes, I'm using DirectSound. It seems I have to place the

playbackBuffer = new SecondaryBuffer(playbackBufferDescription, device)

to prevent the voice to be cut up.
Anyway, I'll try to call play after some loops of filling the buffer. But the problem of disposing still remains unsolved and memory use is still too high.
Thanks again for any other help tips

Simon P Stevens8-Oct-08 4:42

8-Oct-08 4:42

No, your missing the point.

What you are doing is creating a new buffer every time you loop round. The sound processor is then queuing all of these buffers and playing them in order. Thus, you are getting a clear uninterrupted sound, but at the same time you are creating hundreds of buffer objects that are just forming a longer and longer queue.

What you should be doing is adding the data onto the end of the same buffer each time you go round the loop. This way the sound processor will continually play out of the same buffer as it will never run out of data because you are continually adding more data into the buffer.

My solution will stop the jittering because I am suggesting that you only need to call play once at the beginning. Provided that your source continues to send, you will never need to call play again. The 5 second counter is there to prevent play being called before there is anything in the buffer to begin with. And the flag is there so that play isn't called again. You will only need to call play again if the buffer runs out and stops (maybe you have network trouble, or the source stops for some reason, so if this happens, then flag is reset, and the code will wait until the buffer has 5 seconds in it before calling play() again)

Simon

DaveyM698-Oct-08 5:58

DaveyM69

8-Oct-08 5:58

what Simon's referring to is called a 'circular buffer' and is used in many communication protocols where data needs to be processed (played in your case) immediately, before the end of the stream is reached or where the end point is not known.

All live audio monitoring applications utilize this approach. Most use seperate record (receive) and playback buffers too to avoid possible data loss.

Dave
BTW, in software, hope and pray is not a viable strategy. (Luc Pattyn)
Visual Basic is not used by normal people so we're not covering it here. (Uncyclopedia)

Simon P Stevens8-Oct-08 22:58

Re: Problem with free up memory! [modified]

8-Oct-08 22:58

DaveyM69 wrote:
'circular buffer'

Ahh, yes...This is what I'm talking about. For some reason the word just never came to my head. Nice one.

Simon

Mark Salsbery8-Oct-08 8:07

Mark Salsbery

8-Oct-08 8:07

serima nazarian wrote:
my application already plays the buffer without any problem or any cut in the voice

You obviously haven't let it play long enough.

The way it's coded now, aside from the memory leak, you're going to
have problems with clock drift and network jitter.

Let it go for several minutes - you'll see what I mean Smile | :)

Good luck.

Edit:<br />
Actually, looking at the code again, you may not have too much problem<br />
with jitter or drift, since you're using a new secondary buffer for every<br />
receive.  There will probably be a problem if the GC gets around to collecting<br />
any secondary buffers that are still playing asynchronously, however.

Mark Salsbery
Microsoft MVP - Visual C++
Java | [Coffee]

modified on Wednesday, October 8, 2008 2:15 PM

Ha ha ha ha ha ha ha ha9-Oct-08 20:50