[alsa-devel] period_time in slave definition

[Pavel Hofman]

Pavel Hofman napsal(a):
> Jaroslav Kysela napsal(a):
>> On Tue, 1 Mar 2011, Pavel Hofman wrote:
>>
>>> Jaroslav Kysela napsal(a):
>>>> It's correct, pcm:plug does not support buffer modifiers. Look to
>>>> "Automatic conversion plugin".
>>>>
>>>>> I checked the source code and the keywords seem to be ok
>>>>>
>>>>> pcm.c:
>>>>> static const char *const names[SND_PCM_HW_PARAM_LAST_INTERVAL + 1] = {
>>>>>     [SND_PCM_HW_PARAM_FORMAT] = "format",
>>>>>     [SND_PCM_HW_PARAM_CHANNELS] = "channels",
>>>>>     [SND_PCM_HW_PARAM_RATE] = "rate",
>>>>>     [SND_PCM_HW_PARAM_PERIOD_TIME] = "period_time",
>>>>>     [SND_PCM_HW_PARAM_PERIOD_SIZE] = "period_size",
>>>>>     [SND_PCM_HW_PARAM_BUFFER_TIME] = "buffer_time",
>>>>>     [SND_PCM_HW_PARAM_BUFFER_SIZE] = "buffer_size",
>>>>>     [SND_PCM_HW_PARAM_PERIODS] = "periods"
>>>>> };
>>>> It's just a table to translate all possibilites between index and name.
>>>> The plug plugins allow only this:
>>>>
>>>>     err = snd_pcm_slave_conf(root, slave, &sconf, 3,
>>>>                  SND_PCM_HW_PARAM_FORMAT, SCONF_UNCHANGED, &sformat,
>>>>                  SND_PCM_HW_PARAM_CHANNELS, SCONF_UNCHANGED, &schannels,
>>>>                  SND_PCM_HW_PARAM_RATE, SCONF_UNCHANGED, &srate);
>>>>
>>>>                     Jaroslav
>>>>
>>>
>>> Well, still I do not understand why we are talking about the plug
>>> plugin, when I am trying to redefine the slave via pcm_slave. The
>>> documentation at
>>> http://www.alsa-project.org/alsa-doc/alsa-lib/pcm_plugins.html says the
>>> pcm_slave can define period_time and buffer_time, while a slightly
>>> modified example from the wiki page fails:
>>>
>>> pcm_slave.slave_rate44100Hz {
>>>        pcm "hw:0,0"
>>>        rate 44100
>>> --->    period_time 1024
>>> }
>>>
>>> pcm.rate44100Hz {
>>>        type plug
>>>        slave slave_rate44100Hz
>>> }
>>>
>>> How does the slave definition relate to the plug plugin? Sorry for these
>>> basics but I am really trying to understand that, so far no success.
>> The above description is an universal description for all slaves.
>> Unfortunately, not all plugins accept the whole range of parameters.
>> You have to look to the snd_pcm_slave_conf() caller which parameters are
>> accepted.
>>
> 
> Thanks a lot for your clarification. Would it make sense to add these
> parameters to the plug plugin? I guess it would not be as simple as in
> dshare.c:
> 
> 	params.period_size = psize;
> 	params.buffer_size = bsize;
> 
> when opening the slave.
> 

One more question regarding the defaults. Older version of audacious
(e.g. in ubuntu karmic) uses a very simple code:

    snd_pcm_hw_params_alloca(&hwparams);
    snd_pcm_hw_params_any(pcm_handle, hwparams);
    snd_pcm_hw_params_set_access(pcm_handle, hwparams,
SND_PCM_ACCESS_RW_INTERLEAVED);
    snd_pcm_hw_params_set_format(pcm_handle, hwparams, afmt);
    snd_pcm_hw_params_set_channels(pcm_handle, hwparams, nch);
    snd_pcm_hw_params_set_rate(pcm_handle, hwparams, rate, 0);

    err = snd_pcm_hw_params(pcm_handle, hwparams);
    if (err < 0)
    {
        _ERROR("snd_pcm_hw_params failed: %s", snd_strerror(err));
        return -1;
    }


There is no buffer or period setup. I assume some alsa-lib default value
is used. When playing, the period size is set very low:

access: MMAP_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 96000 (96000/1)
period_size: 8
buffer_size: 8192

As a result, the sound card (ICH5) keeps throwing 12.000 interrupts per
second and CPU load is rather high. Does this "default" period_size
value make sense? This is the reason I am trying to specify it outside
of the application.

Thanks a lot for help.

Pavel.