Takashi Iwai <tiwai@suse.de> writes:

> Thanks for the patch.  But this looks suboptimal, unfortunately, since
> it keeps the amp always on, and more badly, it would block the power
> save of the widget root node.

Thanks very much for your feedback; I wasn't sure precisely how this
code worked and tried to make a change that was as close as I could
manage to existing examples.

> Can just using gpio_mute_led_mask=0x18 and gpio_led=0 (also drop
> AC_VERB_SET_GPIO_DATA in gpio_init[]) work instead?  If GPIO4 is the
> the amp, we can associate it with the master mute control together
> with the mute LED.  The only concern would be the possible click
> noise, but it doesn't happen on most machines.

It's not quite that simple; the GPIO4 value is inverted from the mute
LED value (the amp is powered up when GPIO4 is set).

What I've done is to make the amp powered only when a headphone is
plugged in, and then removed the code which was disabling power saving,
which lets everything (including the amp) get turned back off when the
device goes idle.

Here's a second version of the patch.