On Wed, May 04, 2011 at 11:12:50PM +0800, Lu Guanqun wrote:
On Wed, May 04, 2011 at 10:46:00PM +0800, Dimitris Papastamos wrote:
On Wed, May 04, 2011 at 09:44:58PM +0800, Lu Guanqun wrote:
- case SND_SOC_BIAS_OFF:
snd_soc_write(codec, UPD9976_VREFPLL, 0);snd_soc_write(codec, UPD9976_VAUDIOCNT, 0x24);break;- }
Why not snd_soc_update_bits()? These should normally be DAPM widgets.
There's no DAPM widgets bound to these two registers. So I'm afraid it's OK to use snd_soc_write when it's been powered off totally.
That's an orthogonal thing, the usual reason for pushing back on this stuff is that people have open coded a read/modify/write cycle which only updates a subset of bits. In this case I think the writes are fine as you're setting the full register to a specific value.
+static int upd9976_codec_probe(struct snd_soc_codec *codec) +{
- upd9976_set_bias_level(codec, SND_SOC_BIAS_OFF);
- return 0;
+}
Why SND_SOC_BIAS_OFF and not SND_SOC_BIAS_STANDBY?
Try to use as little power as possible.
You should set idle_bias_off in the CODEC driver if you're doing this otherwise DAPM will just bring the CODEC up to _STANDBY at runtime.