hw_rule

hw_rule

[Giuliano Pochini]

I wrote this rule:


static int hw_rule_format_set_channels(snd_pcm_hw_params_t *params,
snd_pcm_hw_rule_t *rule) {
  snd_interval_t *c=hw_param_interval(params,
SNDRV_PCM_HW_PARAM_CHANNELS);
  snd_mask_t *f=hw_param_mask(params, SNDRV_PCM_HW_PARAM_FORMAT);
  snd_interval_t ch;

  printk("Rule f=%x %x %x\n", f->bits[0], f->bits[1], f->bits[2]);
  memset(&ch, 0, sizeof(ch));

  // S32_BE is mono only
  if (f->bits[0]==SNDRV_PCM_FMTBIT_S32_BE) {
    ch.min=1;
    ch.max=1;
    ch.integer=1;
    return(snd_interval_refine(c, &ch));
  }
  // U8 and S16_LE are stereo only
  if (f->bits[0]==SNDRV_PCM_FMTBIT_U8 ||
      f->bits[0]==SNDRV_PCM_FMTBIT_S16_LE) {
    ch.min=2;
    ch.max=2;
    ch.integer=1;
    return(snd_interval_refine(c, &ch));
  }
  // S32_LE supports any number of channels.
  return(0);
}


It seems to work, but since it's based on other drivers (no docs...),
there are some thing I don't know:
- Is f->bits[0] correct ?  All formats supported by the card are < 31. I
suppose formats >=32 go in bits[1], isn't it ?
- what are ch.openmin, openmax, empty ?  They are taken into account
because if I don't clear the structure the rule doesn't work anymore.
- How many stupid bugs are there ? :)


Bye.




-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com

Re: hw_rule

[Takashi Iwai]

At 08 May 2003 23:19:28 +0000,
Giuliano Pochini wrote:
> 
> 
> I wrote this rule:
> 
> 
> static int hw_rule_format_set_channels(snd_pcm_hw_params_t *params,
> snd_pcm_hw_rule_t *rule) {
>   snd_interval_t *c=hw_param_interval(params,
> SNDRV_PCM_HW_PARAM_CHANNELS);
>   snd_mask_t *f=hw_param_mask(params, SNDRV_PCM_HW_PARAM_FORMAT);
>   snd_interval_t ch;
> 
>   printk("Rule f=%x %x %x\n", f->bits[0], f->bits[1], f->bits[2]);
>   memset(&ch, 0, sizeof(ch));
> 
>   // S32_BE is mono only
>   if (f->bits[0]==SNDRV_PCM_FMTBIT_S32_BE) {
>     ch.min=1;
>     ch.max=1;
>     ch.integer=1;
>     return(snd_interval_refine(c, &ch));
>   }
>   // U8 and S16_LE are stereo only
>   if (f->bits[0]==SNDRV_PCM_FMTBIT_U8 ||
>       f->bits[0]==SNDRV_PCM_FMTBIT_S16_LE) {
>     ch.min=2;
>     ch.max=2;
>     ch.integer=1;
>     return(snd_interval_refine(c, &ch));
>   }
>   // S32_LE supports any number of channels.
>   return(0);
> }
> 
> 
> It seems to work, but since it's based on other drivers (no docs...),
> there are some thing I don't know:
> - Is f->bits[0] correct ?  All formats supported by the card are < 31. I
> suppose formats >=32 go in bits[1], isn't it ?

yes.  but over 32 is rarely used.

> - what are ch.openmin, openmax, empty ?  They are taken into account
> because if I don't clear the structure the rule doesn't work anymore.

openmin/openmax are the minimum and maximum value but the
configuration cannot be exactly these values.
empty is set if the configuration is empty.

> - How many stupid bugs are there ? :)

i see only subtle issues.
it's better to use snd_interval_any() instead of memset().
the resultant interval_t would accept all values.

the constraint for the another direction would be needed, too,
that is, the constraint for the channels which checks the acceptable
formats from the current values.


ciao,

Takashi


-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com

Re: hw_rule

[Abramo Bagnara]

Takashi Iwai ha scritto:
> At 08 May 2003 23:19:28 +0000,
> Giuliano Pochini wrote:
> 
>>
> 
>>- what are ch.openmin, openmax, empty ?  They are taken into account
>>because if I don't clear the structure the rule doesn't work anymore.
> 
> 
> openmin/openmax are the minimum and maximum value but the
> configuration cannot be exactly these values.

Not exactly: openmin/openmax are the flags that, when set to true, means 
that min/max are not valid value of the numeric interval.

They're sensible only for non integer intervals.

As an example the interval 44100.1 - 48000.4 is represented as an 
interval 44100 - 48001 with both boundaries open.

-- 
Abramo Bagnara                       mailto:abramo.bagnara@libero.it

Opera Unica                          Phone: +39.546.656023
Via Emilia Interna, 140
48014 Castel Bolognese (RA) - Italy



-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com

Re: hw_rule

[Giuliano Pochini]

On ven, 2003-05-09 at 14:27, Takashi Iwai wrote:
> > - How many stupid bugs are there ? :)
>
> i see only subtle issues.
> it's better to use snd_interval_any() instead of memset().
> the resultant interval_t would accept all values.

Done, tnx.

rme9652.c::snd_rme9652_hw_rule_*() functions do not initialize the
structures. They may fail, unless I'm missing something.

> the constraint for the another direction would be needed, too,
> that is, the constraint for the channels which checks the acceptable
> formats from the current values.

Yes, of course.


Bye.




-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com