Skip to contentSkip to conversation

Channel map API patches: to be 3.7 or not to be?

DISCUSSION Takashi Iwai 4 participants 14 messages Sep 10, 2012 – Sep 17, 2012

lore
Share & tools
lore
#0 lore raw

Hi,

does anyone have concern if I push the current channel map API patches (in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for any other purposes.)

thanks,

Takashi

#1 lore raw

2012-9-11 上午1:43 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 11 lines

Hi,

does anyone have concern if I push the current channel map API patches (in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

Quoted 8 lines

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for any other purposes.)

How about the channel map of the usb mono playback device and those sound card which can pan mono to stereo speakers and headphone?

Pulseaudio developer has just changed the downmixing method

The mono playback is not left anymore

https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/416190
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#2 lore raw

At Tue, 11 Sep 2012 16:10:43 +0800,

Quoted 16 lines · Raymond Yau

2012-9-11 上午1:43 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 14 lines

Hi,

does anyone have concern if I push the current channel map API patches (in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

Quoted 8 lines
Quoted 5 lines

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for any other purposes.)

How about the channel map of the usb mono playback device and those sound card which can pan mono to stereo speakers and headphone?

If the device specifies which speaker position the mono channel is for, then you can map it properly. Unless it's really specified by the device, the mono stream is UNKNOWN position in general.

Quoted 5 lines

Pulseaudio developer has just changed the downmixing method

The mono playback is not left anymore

https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/416190

Takashi
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#3 lore raw

At Tue, 11 Sep 2012 10:22:23 +0200,

Quoted 25 lines · Takashi Iwai

At Tue, 11 Sep 2012 16:10:43 +0800,

Quoted 22 lines · Raymond Yau

2012-9-11 上午1:43 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 19 lines

Hi,

does anyone have concern if I push the current channel map API patches (in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for any other purposes.)

Quoted 6 lines
Quoted 2 lines

How about the channel map of the usb mono playback device and those sound card which can pan mono to stereo speakers and headphone?

If the device specifies which speaker position the mono channel is for, then you can map it properly. Unless it's really specified by the device, the mono stream is UNKNOWN position in general.

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

Takashi
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#4 lore raw
Quoted 7 lines · Takashi Iwai

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

What would it be used for?

Labeling surround channels is necessary for reordering them or for doing up/downmixing. A "mono" channel, however, does not appear to have any semantic difference from a single "unknown" channel.

Regards,
Clemens

#5 lore raw

At Tue, 11 Sep 2012 14:19:49 +0200,

Quoted 10 lines · Clemens Ladisch
Quoted 7 lines · Takashi Iwai

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

Quoted 5 lines

What would it be used for?

Labeling surround channels is necessary for reordering them or for doing up/downmixing. A "mono" channel, however, does not appear to have any semantic difference from a single "unknown" channel.

Right, that's the current status. OTOH, labeling it as "mono" makes clear that it's an unaligned mono stream. i.e. it shows the purpose of the stream by itself.

Well, I'm not for introducing yet another definition. Just stumbled on it, and would like to know opinions from others.

thanks,

Takashi

#6 lore raw

At Tue, 11 Sep 2012 14:24:09 +0200,

Quoted 15 lines · Takashi Iwai

At Tue, 11 Sep 2012 14:19:49 +0200,

Quoted 12 lines · Clemens Ladisch
Quoted 7 lines · Takashi Iwai

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

What would it be used for?

Quoted 7 lines
Quoted 3 lines

Labeling surround channels is necessary for reordering them or for doing up/downmixing. A "mono" channel, however, does not appear to have any semantic difference from a single "unknown" channel.

Right, that's the current status. OTOH, labeling it as "mono" makes clear that it's an unaligned mono stream. i.e. it shows the purpose of the stream by itself.

Looking at PulseAudio and gstreamer codes, they seem to have explicit MONO channel definition.

typedef enum pa_channel_position {
PA_CHANNEL_POSITION_INVALID = -1,
PA_CHANNEL_POSITION_MONO = 0,
....

typedef enum {
GST_AUDIO_CHANNEL_POSITION_INVALID = -1,

/* Main front speakers. Mono and left/right are mututally exclusive! */ GST_AUDIO_CHANNEL_POSITION_FRONT_MONO,
....

So, defining MONO seems rather standard. Now I'm inclined to define SND_CHMAP_MONO...

Takashi

#7 lore raw
Quoted 11 lines · Takashi Iwai
Quoted 5 lines
Quoted 4 lines · Clemens Ladisch
Quoted 1 line · Takashi Iwai

is it useful to add SND_CHMAP_MONO, or is it just redundant?

What would it be used for?

Looking at PulseAudio and gstreamer codes, they seem to have explicit MONO channel definition.
[...]
So, defining MONO seems rather standard. Now I'm inclined to define SND_CHMAP_MONO...

Okay; if they want it, they can get it ...

Regards,
Clemens

#8 lore raw
Quoted 7 lines · Clemens Ladisch
Quoted 6 lines · Takashi Iwai
Quoted 5 lines
Quoted 4 lines · Clemens Ladisch
Quoted 1 line · Takashi Iwai

is it useful to add SND_CHMAP_MONO, or is it just redundant?

What would it be used for?

Quoted 7 lines
Quoted 5 lines

Looking at PulseAudio and gstreamer codes, they seem to have explicit MONO channel definition.
[...]
So, defining MONO seems rather standard. Now I'm inclined to define SND_CHMAP_MONO...

Okay; if they want it, they can get it ...

Speaking as a PulseAudio developer, "wanting" is a bit strong word. I initially wanted it, but I don't anymore think that it's necessary. I don't think it will do any harm either, though...

Signature · 1 line
Tanu
#9 lore raw
In reply to Takashi Iwai

2012-9-11 下午8:14 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 12 lines

At Tue, 11 Sep 2012 10:22:23 +0200,

Quoted 9 lines · Takashi Iwai

At Tue, 11 Sep 2012 16:10:43 +0800,

Quoted 6 lines · Raymond Yau

2012-9-11 上午1:43 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 4 lines

Hi,

does anyone have concern if I push the current channel map API

patches

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

(in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for

any

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines

other purposes.)

How about the channel map of the usb mono playback device and those

sound

Quoted 6 lines
Quoted 5 lines
Quoted 1 line

card which can pan mono to stereo speakers and headphone?

If the device specifies which speaker position the mono channel is for, then you can map it properly. Unless it's really specified by the device, the mono stream is UNKNOWN position in general.

Quoted 9 lines

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

How about those sound card which use multi plugin for multi channels (e.g. emu10k1 , c0106 and cs4630) as they can also provide a rear or clfe device for independent stereo playback ?

It seem the branch still miss the patch for au88x0 and ymfpci which need to check the sdac bit of ac97 codec for adding the 4 channels map

Most of 2.1 external speaker set only have green jack and those creative 4.1 fps have green and black jacks where the volume of subwoofer is controlled by the volume knob of the subwoofer

https://bugs.launchpad.net/ubuntu/+source/alsa-driver/+bug/1047543

However some of those 2.1 internal speaker in notebook may need stereo channel map (e.g. stac920x hda codec using mono widget)

Some notebook (e.g. acer aspire 4930g) hda codec) may need a special channel map for the internal 2.1 speaker but a stereo channel map for the external 2.1 speaker and some Asus laptops have propertiary jack for the external subwoofer

http://git.kernel.org/?p=linux/kernel/git/tiwai/sound.git;a=commit;h=786c51f9168cfd2d49250c6e5e60035cbb2fd5a1

Does those 3stack-6ch laptop/desktop have those 5.1 channel map before the input jack retasked as output?

If yes, does the retasking is performed automatically by the set channel map function?

http://git.kernel.org/?p=linux/kernel/git/tiwai/sound-unstable.git;a=commit;h=9c9a5175e65b2667001e6a3b6dedddebeee82aa2
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#10 lore raw

At Wed, 12 Sep 2012 09:35:20 +0800,

Quoted 50 lines · Raymond Yau

2012-9-11 下午8:14 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 12 lines

At Tue, 11 Sep 2012 10:22:23 +0200,

Quoted 9 lines · Takashi Iwai

At Tue, 11 Sep 2012 16:10:43 +0800,

Quoted 6 lines · Raymond Yau

2012-9-11 上午1:43 於 "Takashi Iwai" <[email protected]> 寫道:

Quoted 4 lines

Hi,

does anyone have concern if I push the current channel map API

patches

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

(in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for

any

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines

other purposes.)

How about the channel map of the usb mono playback device and those

sound

Quoted 6 lines
Quoted 5 lines
Quoted 1 line

card which can pan mono to stereo speakers and headphone?

If the device specifies which speaker position the mono channel is for, then you can map it properly. Unless it's really specified by the device, the mono stream is UNKNOWN position in general.

Quoted 9 lines

BTW, this reminds me of an open question:
is it useful to add SND_CHMAP_MONO, or is it just redundant?

It's nothing but indicating a mono channel without any channel position, so I supposed SND_CHMAP_UNKNOWN being sufficient. OTOH, SND_CHMAP_MONO would give a clear sign of mono streams while SND_CHMAP_UNKNOWN could be used for any other exceptional purposes.

Quoted 3 lines

How about those sound card which use multi plugin for multi channels (e.g. emu10k1 , c0106 and cs4630) as they can also provide a rear or clfe device for independent stereo playback ?

The case of emu10k1 is a bit difficult, as the routing is completely dynamic and determined in alsa-lib plugin. So, my plan for this case is to provide the chmap override in the alsa-lib config definition for emu10k1 surrounds.

In other cases, the PCM stream is mostly bound with certain channels, and multi plugin can combine all these chmaps. ctxfi already provides independent stereo streams like RL/RR, combined via multi plugin for 4.0 or 5.1.

Quoted 2 lines

It seem the branch still miss the patch for au88x0 and ymfpci which need to check the sdac bit of ac97 codec for adding the 4 channels map

Feel free to send a patch.

Quoted 9 lines

Most of 2.1 external speaker set only have green jack and those creative 4.1 fps have green and black jacks where the volume of subwoofer is controlled by the volume knob of the subwoofer

https://bugs.launchpad.net/ubuntu/+source/alsa-driver/+bug/1047543

However some of those 2.1 internal speaker in notebook may need stereo channel map (e.g. stac920x hda codec using mono widget)

Quoted 6 lines

Some notebook (e.g. acer aspire 4930g) hda codec) may need a special channel map for the internal 2.1 speaker but a stereo channel map for the external 2.1 speaker and some Asus laptops have propertiary jack for the external subwoofer

http://git.kernel.org/?p=linux/kernel/git/tiwai/sound.git;a=commit;h=786c51f9168cfd2d49250c6e5e60035cbb2fd5a1

The _only_ interest for now is how to map the channels for multi-channel streams, and whether the API proposal can cover it or not. When the hardware binds a boost speaker internally, it's not what applications care. It's nothing but a driver setup.

So, the only interesting topic in the above would be the proper 2.1 mapping (i.e. the lack of FC channel) for ASUS laptops. And this case doesn't conflict with the API proposal.

Quoted 2 lines

Does those 3stack-6ch laptop/desktop have those 5.1 channel map before the input jack retasked as output?

It depends on the driver implementation. API doesn't restrict the behavior. I myself suppose the query could be static in such a case while get/set should reflect to the current jack state.

Quoted 4 lines

If yes, does the retasking is performed automatically by the set channel map function?

http://git.kernel.org/?p=linux/kernel/git/tiwai/sound-unstable.git;a=commit;h=9c9a5175e65b2667001e6a3b6dedddebeee82aa2

The operation is allowed in theory, feel free to implement. But I don't think we'd need it.

Takashi
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#11 lore raw
Quoted 1 line
Quoted 1 line

How about those sound card which use multi plugin for multi channels

(e.g.

Quoted 1 line
Quoted 1 line

emu10k1 , c0106 and cs4630) as they can also provide a rear or clfe

device

Quoted 7 lines
Quoted 1 line

for independent stereo playback ?

The case of emu10k1 is a bit difficult, as the routing is completely dynamic and determined in alsa-lib plugin. So, my plan for this case is to provide the chmap override in the alsa-lib config definition for emu10k1 surrounds.

But the routing of device 0 "hw" is not as same as front device since the driver pan the output to both front and rear jacks (upmixing).

Does all pcm devices support channel map ?

How about those 16 channels multichannel playback and capture devices ?

card 0: Live [SB Live! Platinum [CT4760P]], device 2: emu10k1 efx [Multichannel Capture/PT Playback]
Subdevices: 8/8

card 0: Live [SB Live! Platinum [CT4760P]], device 3: emu10k1 [Multichannel Playback]
Subdevices: 1/1

**** List of CAPTURE Hardware Devices ****

card 0: Live [SB Live! Platinum [CT4760P]], device 2: emu10k1 efx [Multichannel Capture/PT Playback]
Subdevices: 1/1

BTW. It seem that the patch for USB audio is still missing as some USB card support 5.1/7.1

#12 lore raw

At Mon, 17 Sep 2012 18:46:32 +0800,

Quoted 7 lines · Raymond Yau
Quoted 2 lines
Quoted 2 lines

How about those sound card which use multi plugin for multi channels

(e.g.

Quoted 1 line
Quoted 1 line

emu10k1 , c0106 and cs4630) as they can also provide a rear or clfe

device

Quoted 2 lines
Quoted 1 line

for independent stereo playback ?

Quoted 7 lines
Quoted 5 lines

The case of emu10k1 is a bit difficult, as the routing is completely dynamic and determined in alsa-lib plugin. So, my plan for this case is to provide the chmap override in the alsa-lib config definition for emu10k1 surrounds.

But the routing of device 0 "hw" is not as same as front device since the driver pan the output to both front and rear jacks (upmixing).

The hw device doesn't provide any map. It's only for "front", "rear" whatever in strict channel position definitions.

Quoted 1 line

Does all pcm devices support channel map ?

No. The channel map is an optional feature.

Quoted 10 lines

How about those 16 channels multichannel playback and capture devices ?

card 0: Live [SB Live! Platinum [CT4760P]], device 2: emu10k1 efx [Multichannel Capture/PT Playback]
Subdevices: 8/8

card 0: Live [SB Live! Platinum [CT4760P]], device 3: emu10k1 [Multichannel Playback]
Subdevices: 1/1

Quoted 5 lines

**** List of CAPTURE Hardware Devices ****

card 0: Live [SB Live! Platinum [CT4760P]], device 2: emu10k1 efx [Multichannel Capture/PT Playback]
Subdevices: 1/1

Such devices are not supported. If any, they can be mapped using DRIVER_SPEC flag.

Quoted 2 lines

BTW. It seem that the patch for USB audio is still missing as some USB card support 5.1/7.1

Feel free to submit a patch :)

Takashi

#13 lore raw
In reply to Takashi Iwai

At Mon, 10 Sep 2012 19:43:33 +0200,

Quoted 11 lines · Takashi Iwai

Hi,

does anyone have concern if I push the current channel map API patches (in sound-unstable git tree topic/tlv-chmap branch) for linux-next, i.e. for inclusion to 3.7 kernel?

I don't worry too much about the kernel API. This can be refactored later. But the kernel ABI must retain, so we won't change in an incompatible way any longer once after it's merged to the upstream.

Quoted 7 lines

So, if anyone sees a flaw in the kernel ABI definition using control elements, let me know.

(But note that the API can't be perfect for all generic purposes.
It's designed to be "good enough" for a big lack for multi-channel streaming. Of course, this doesn't discourage to use the API for any other purposes.)

FYI, as I haven't got any big NO-NO, now the topic branch for chmap is merged to for-next branch. I'm going to merge user-space side, alsa-lib and alsa-utils patches.

Takashi