At Sun, 23 Nov 2008 09:30:40 +0800, Vincent Petry wrote:

Hi,

I have noticed that from one model to another in ALC888 (maybe others as well), the nodes for Int speakers, HP out, are not always the same. But in the current driver model in patch_realtek.c we always hard-code the node numbers, for example 0x14 for int speakers, 0x15 for hp out, etc.

Note that this is what the preset model assumes -- the same pin configuration.

It would be good to have it this way:

1. Add a spec attribute that contains the output nids in a given order

or a map, like this for my laptop: alc888_acer_aspire_jack_nids = { {0x14, NID_INT_SPEAKER}, {0x15, NID_HP_OUT}, {0x18, NID_MIC_IN}, {0x1a, NID_LINE_IN}, {} };

For another laptop, that has built-in subwoofer: alc888_fujitsu_xa2530_jack_nids = { { {0x14, NID_INT_SPEAKER}, {0x15, NID_INT_SPEAKER}, // subwoofer, also considered as int speaker for the jack sense muting {0x17, NID_LINE_OUT}, // this is also side {0x18, NID_MIC_IN}, {0x1a, NID_LINE_IN}, {} };

This way we could then use a function to access the NID number of each one of them, for example get_jack_nid(..., NID_INT_SPEAKER).

The mapping is nice. A similar mechanism is found in patch_sigmatel.c.

But I'd suggest to avoid the name NID_*, because this sounds like a widget NID although it doesn't mean so in this case: it's the role of each pin. Thus, something like ALC_PIN_INT_SPEAKER would be more appropriate.

2. Once this is implemented, we could add a detection routine that would

build the above map automatically, based on the spec info returned by the codec.

3. Then I think we should add some special "detection" model that has no

hard-coded values that anyone could use by default, if no suitable model def exists.

Such a model can be hardcoded indeed. No need to introduce a model over a model at this point. If any better way comes in sight, we can go forward. But let's reduce the codes step by step.

thanks,

Takashi