problems: flushing during snd_pcm_wait()

problems: flushing during snd_pcm_wait()

[James Courtier-Dutton]

Hello
I have an application that has many different threads.
The sound card's PCM buffer is full during playback, and one thread is 
currently in snd_pcm_wait() waiting for enough space to appear in the 
buffer before doing the next snd_pcm_write().
A different thread wants to flush the buffer. There are two types of 
flush: -
1) Immeadiately empty the buffer.
2) Stop accepting snd_pcm_write() commands until the current buffer has 
drained.

Can someone please tell me which API calls I should make to get the 
buffer to flush properly.
I am currently doing things like: -
int flush() {
 <snip...>
      if ((result=snd_pcm_reset(this->audio_fd)) < 0) {
        printf("audio_alsa_out: Reset call failed err=%d\n",result);
      }
      if ((result=snd_pcm_drain(this->audio_fd)) < 0) {
        printf("audio_alsa_out: Drain call failed err=%d\n",result);
      }
      if ((result=snd_pcm_prepare(this->audio_fd)) < 0) {
        printf("audio_alsa_out: Prepare call failed err=%d\n",result);
      }
}

But this is causing all sorts of problems, one of which is that the 
application hangs, and has to be killed with -9, and the audio hardware 
ring buffer is left playing a repeated sound even with: -
/proc/asound/card0/pcm0p/sub0# cat status
closed

Cheers
James





-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: problems: flushing during snd_pcm_wait()

[Takashi Iwai]

At Fri, 11 Oct 2002 01:25:19 +1000,
James Courtier-Dutton wrote:
> 
> Hello
> I have an application that has many different threads.
> The sound card's PCM buffer is full during playback, and one thread is 
> currently in snd_pcm_wait() waiting for enough space to appear in the 
> buffer before doing the next snd_pcm_write().
> A different thread wants to flush the buffer. There are two types of 
> flush: -
> 1) Immeadiately empty the buffer.

snd_pcm_drop() ?

> 2) Stop accepting snd_pcm_write() commands until the current buffer has 
> drained.

snd_pcm_drain() ?


Takashi


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: problems: flushing during snd_pcm_wait()

[James Courtier-Dutton]

Takashi Iwai wrote:

>At Fri, 11 Oct 2002 01:25:19 +1000,
>James Courtier-Dutton wrote:
>  
>
>>Hello
>>I have an application that has many different threads.
>>The sound card's PCM buffer is full during playback, and one thread is 
>>currently in snd_pcm_wait() waiting for enough space to appear in the 
>>buffer before doing the next snd_pcm_write().
>>A different thread wants to flush the buffer. There are two types of 
>>flush: -
>>1) Immeadiately empty the buffer.
>>    
>>
>
>snd_pcm_drop() ?
>
>  
>
>>2) Stop accepting snd_pcm_write() commands until the current buffer has 
>>drained.
>>    
>>
>
>snd_pcm_drain() ?
>
>
>Takashi
>
>  
>
Thankyou, I will use snd_pcm_drop(), but as a side note, what actually 
does "snd_pcm_reset()" do.
Just resetting delay to 0 does not make much sense to me.

Cheer




-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: problems: flushing during snd_pcm_wait()

[Jaroslav Kysela]

On Fri, 11 Oct 2002, James Courtier-Dutton wrote:

> Takashi Iwai wrote:
> 
> >At Fri, 11 Oct 2002 01:25:19 +1000,
> >James Courtier-Dutton wrote:
> >  
> >
> >>Hello
> >>I have an application that has many different threads.
> >>The sound card's PCM buffer is full during playback, and one thread is 
> >>currently in snd_pcm_wait() waiting for enough space to appear in the 
> >>buffer before doing the next snd_pcm_write().
> >>A different thread wants to flush the buffer. There are two types of 
> >>flush: -
> >>1) Immeadiately empty the buffer.
> >>    
> >>
> >
> >snd_pcm_drop() ?
> >
> >  
> >
> >>2) Stop accepting snd_pcm_write() commands until the current buffer has 
> >>drained.
> >>    
> >>
> >
> >snd_pcm_drain() ?
> >
> >
> >Takashi
> >
> >  
> >
> Thankyou, I will use snd_pcm_drop(), but as a side note, what actually 
> does "snd_pcm_reset()" do.
> Just resetting delay to 0 does not make much sense to me.

It drops all samples in the ring buffer (thus reseting delay to 0). Note 
that everybody are welcome to improve the current documentation 
(preferably with patches).

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project  http://www.alsa-project.org
SuSE Linux    http://www.suse.com



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: problems: flushing during snd_pcm_wait()

[James Courtier-Dutton]

>
>
>>>      
>>>
>>Thankyou, I will use snd_pcm_drop(), but as a side note, what actually 
>>does "snd_pcm_reset()" do.
>>Just resetting delay to 0 does not make much sense to me.
>>    
>>
>
>It drops all samples in the ring buffer (thus reseting delay to 0). Note 
>that everybody are welcome to improve the current documentation 
>(preferably with patches).
>
>						Jaroslav
>  
>
Ok, so snd_pcm_drop drops all samples in the buffer, so I would assume 
this sets the delay to 0.
so snd_pcm_reset drops all sample in the buffer, so as you say is sets 
delay to 0.
So my question is "What is the difference between the 2 functions?"

Cheers
James





-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: problems: flushing during snd_pcm_wait()

[Jaroslav Kysela]

On Fri, 11 Oct 2002, James Courtier-Dutton wrote:

> >>Thankyou, I will use snd_pcm_drop(), but as a side note, what actually 
> >>does "snd_pcm_reset()" do.
> >>Just resetting delay to 0 does not make much sense to me.
> >>    
> >>
> >
> >It drops all samples in the ring buffer (thus reseting delay to 0). Note 
> >that everybody are welcome to improve the current documentation 
> >(preferably with patches).
> >
> >						Jaroslav
> >  
> >
> Ok, so snd_pcm_drop drops all samples in the buffer, so I would assume 
> this sets the delay to 0.
> so snd_pcm_reset drops all sample in the buffer, so as you say is sets 
> delay to 0.
> So my question is "What is the difference between the 2 functions?"

reset() doesn't stop the stream. It's mainly useable, if you have streams 
linked together with hardware or if an application wants to act very 
realtime and has xrun detection disabled.

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project  http://www.alsa-project.org
SuSE Linux    http://www.suse.com



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf