🚀 go-pugleaf

Jonathan Benedicto wrote:
> Would allocating new 8192 byte wave buffers every 1/4 second affect a 
> computer's performance significantly ? I mean, if I kept allocating 
> and freeing those buffers, would excessive RAM fragmentation result ?
> 
> Jonathan
>  
> 
> 
I suspect this is what is causing your clicks. I statically allocate 
everything on the heap, by defining vast 10 MByte structures like this :-

#define BFLIM 100001
#define NBFS 100
char wvbfr[NBFS][BFLIM];
WAVEHDR wvhd[NBFS];

Also, are you using the correct lengths?  If you are using OGG format, every 
1/4 second buffer will be a different length to the next. The member 
wvhd[nxtbf].dwBufferLength must be set correctly each time.

-- 
Mark Jacobs
http://www.dkcomputing.co.uk

"Mark Jacobs" <http://www.jacobsm.com/mjmsg.htm?mj@critical> wrote in 
message news:428861c0@newsgroups.borland.com...
>
> I suspect this is what is causing your clicks. I statically allocate 
> everything on the heap, by defining vast 10 MByte structures like 
> this :-

Actually, if I do create a new buffer every time I need it, then the 
clicks disappear. It's just when I re-use one of those buffers that 
they come.

> Also, are you using the correct lengths?  If you are using OGG 
> format, every 1/4 second buffer will be a different length to the 
> next. The member wvhd[nxtbf].dwBufferLength must be set correctly 
> each time.

I set the dwBufferLength every time I fill the buffer.

Jonathan 

Jonathan Benedicto wrote:
> Actually, if I do create a new buffer every time I need it, then the 
> clicks disappear. It's just when I re-use one of those buffers that 
> they come.

So, why not do that, and free the buffer after use. If allocating a new one 
every time and freeing it afterwards, means you get clickless playback, then 
use that. But I must admit, I am mystified by your problem.
-- 
Mark Jacobs
http://www.dkcomputing.co.uk

"Mark Jacobs" <http://www.jacobsm.com/mjmsg.htm?mj@critical> wrote in 
message news:4288b9ba@newsgroups.borland.com...
>
> So, why not do that, and free the buffer after use. If allocating a 
> new one every time and freeing it afterwards, means you get 
> clickless playback, then use that. But I must admit, I am mystified 
> by your problem.

I didn't want to do that, because I thought that it might cause RAM 
fragmentation, especially if the system worked for a couple of hours.

Jonathan 

>> I suspect this is what is causing your clicks. I statically 
>> allocate everything on the heap, by defining vast 10 MByte 
>> structures like this :-
>
> Actually, if I do create a new buffer every time I need it, then the 
> clicks disappear. It's just when I re-use one of those buffers that 
> they come.
>
>> Also, are you using the correct lengths?  If you are using OGG 
>> format, every 1/4 second buffer will be a different length to the 
>> next. The member wvhd[nxtbf].dwBufferLength must be set correctly 
>> each time.
>
> I set the dwBufferLength every time I fill the buffer.

Do you "new"/"delete" your buffers? I have found situations where 
Borland's memory manager returns memory blocks that are improperly 
aligned (as far as some WinAPI functions are concerned). You might try 
using GlobalAlloc()/GlobalFree() to see if your clicks go away.
Just my $.02.

-- 
Bruce 

" Bruce Salzman" <bruce@nospam.org> wrote in message 
news:4288c053@newsgroups.borland.com...
>
> Do you "new"/"delete" your buffers? I have found situations where 
> Borland's memory manager returns memory blocks that are improperly 
> aligned (as far as some WinAPI functions are concerned). You might 
> try using GlobalAlloc()/GlobalFree() to see if your clicks go away.
> Just my $.02.

The crazy thing is that when I do use new / delete, I don't get the 
clicks. It is just when I re-use a buffer that was previously 
allocated that the clicks come.

But thank you for your input.

Jonathan 

Jonathan Benedicto wrote:
> The crazy thing is that when I do use new / delete, I don't get the 
> clicks. It is just when I re-use a buffer that was previously 
> allocated that the clicks come.

 From what you have said so far, it looks like one of the entities in the wave 
header is not being set correctly, and I suspect it's the dwLoops member. I 
set dwLoops, dwBufferLength and lpData, every 1/4 second (buffer-full). This 
is what I use :-
...
   sounding=true;
   wvhd[nxtbf].lpData=wvbfr[nxtbf];
   wvhd[nxtbf].dwBufferLength=bfln[nxtbf];
   wvhd[nxtbf].dwLoops=0;
   wh=waveOutPrepareHeader(phwo,&wvhd[nxtbf],sizeof(WAVEHDR));
   if (wh==MMSYSERR_NOERROR)
     waveOutWrite(phwo,&wvhd[nxtbf],sizeof(WAVEHDR));
   else ShowMessage("waveOutPrepareHeader failed");
   sounding=false;
...
HTH
-- 
Mark Jacobs
http://www.dkcomputing.co.uk

"Mark Jacobs" <http://www.jacobsm.com/mjmsg.htm?mj@critical> wrote in 
message news:4289ab9f@newsgroups.borland.com...
>
> From what you have said so far, it looks like one of the entities in 
> the wave header is not being set correctly, and I suspect it's the 
> dwLoops member. I set dwLoops, dwBufferLength and lpData, every 1/4 
> second (buffer-full). This is what I use :-

Thank you for continuing to provide ideas. I think that you may be 
right, because yesterday I tried a last resort:

char *TempBufferPtr = WaveHeader->lpData;
ZeroMemory(TempBufferPtr, 8192);
ZeroMemory(WaveHeader, sizeof(WAVEHDR));
WaveHeader->lpData = TempBufferPtr;

And it seems that the clicks are just about silenced. I can still hear 
them when I turn the speakers up, but that could be my sound card, 
speakers or mike. So I'm going to leave it at this.

Thank you again for your help.

Jonathan