Skip to contentSkip to conversation

ASoC: omap-mcbsp: fixes for some underflow problems

PATCH In review Grazvydas Ignotas 4 participants 20 messages Mar 8, 2012 – Mar 12, 2012

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover ASoC: omap-mcbsp: fixes for some underflow problems

With these series I'm trying to deal with underflow problems I was complaining about earlier. This mostly tries to get minimum period size large enough for things to work.

To illustrate the problems I wrote a testcase with alsalib, hopefully it shows the problem more clearly:
http://notaz.gp2x.de/misc/alsa_period_test.c

Applies on top of mcbsp move series.

Grazvydas Ignotas (3):
ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments
ASoC: omap-mcbsp: place size constrain on period, not buffer
ASoC: omap-mcbsp: make minimum period size larger than FIFO

 sound/soc/omap/omap-mcbsp.c |   31 +++++++++++++++++++++----------
 1 files changed, 21 insertions(+), 10 deletions(-)
#1 lore raw

PATCH 1/3 ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

We are setting SNDRV_PCM_HW_PARAM_BUFFER_SIZE based on SNDRV_PCM_HW_PARAM_CHANNELS, not vice versa. This bug didn't have much impact because the rules are evaluated multiple times by the core, and intended value got set eventually.

---
 sound/soc/omap/omap-mcbsp.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
sound/soc/omap/omap-mcbsp.c +2 -2
index 10eb645..207365c 100644
@@ -142,10 +142,10 @@ static int omap_mcbsp_dai_startup(struct snd_pcm_substream *substream,
 		* smaller buffer than the FIFO size to avoid underruns
 		*/
 		snd_pcm_hw_rule_add(substream->runtime, 0,
-				    SNDRV_PCM_HW_PARAM_CHANNELS,
+				    SNDRV_PCM_HW_PARAM_BUFFER_SIZE,
 				    omap_mcbsp_hwrule_min_buffersize,
 				    mcbsp,
-				    SNDRV_PCM_HW_PARAM_BUFFER_SIZE, -1);
+				    SNDRV_PCM_HW_PARAM_CHANNELS, -1);
 
 		/* Make sure, that the period size is always even */
 		snd_pcm_hw_constraint_step(substream->runtime, 0,
Signature · 1 line
1.7.0.4
#2 lore raw

Re: [PATCH 1/3] ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

Quoted 2 lines · Grazvydas Ignotas

We are setting SNDRV_PCM_HW_PARAM_BUFFER_SIZE based on SNDRV_PCM_HW_PARAM_CHANNELS, not vice versa.

The intention is to set the buffer size based on the channels here. This is needed because of the McBSP internal FIFO operation. It is word based. As example on McBSP2 has 1280 word FIFO. In mono stream it is 1280 sample long.
In stereo it is 640 (stereo) sample long.
In four channel mode it is 320 (4 channel) sample long.

Quoted 26 lines

This bug didn't
have much impact because the rules are evaluated multiple times by the core, and intended value got set eventually.

---
 sound/soc/omap/omap-mcbsp.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
sound/soc/omap/omap-mcbsp.c +2 -2
index 10eb645..207365c 100644
@@ -142,10 +142,10 @@ static int omap_mcbsp_dai_startup(struct snd_pcm_substream *substream,
 		* smaller buffer than the FIFO size to avoid underruns
 		*/
 		snd_pcm_hw_rule_add(substream->runtime, 0,
-				    SNDRV_PCM_HW_PARAM_CHANNELS,
+				    SNDRV_PCM_HW_PARAM_BUFFER_SIZE,
 				    omap_mcbsp_hwrule_min_buffersize,
 				    mcbsp,
-				    SNDRV_PCM_HW_PARAM_BUFFER_SIZE, -1);
+				    SNDRV_PCM_HW_PARAM_CHANNELS, -1);
 
 		/* Make sure, that the period size is always even */
 		snd_pcm_hw_constraint_step(substream->runtime, 0,
Signature · 1 line
Péter
#3 lore raw

Re: [PATCH 1/3] ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

Quoted 10 lines · Peter Ujfalusi
Quoted 2 lines · Grazvydas Ignotas

We are setting SNDRV_PCM_HW_PARAM_BUFFER_SIZE based on SNDRV_PCM_HW_PARAM_CHANNELS, not vice versa.

The intention is to set the buffer size based on the channels here. This is needed because of the McBSP internal FIFO operation. It is word based. As example on McBSP2 has 1280 word FIFO. In mono stream it is 1280 sample long.
In stereo it is 640 (stereo) sample long.
In four channel mode it is 320 (4 channel) sample long.

That's all fine, but the argument order is wrong, take a look at snd_pcm_hw_rule_add definition:

/**
* snd_pcm_hw_rule_add - add the hw-constraint rule
* @var: the variable to evaluate
* @dep: the dependent variables
int snd_pcm_hw_rule_add(struct snd_pcm_runtime *runtime, unsigned int cond,
int var,
snd_pcm_hw_rule_func_t func, void *private, int dep, ...)

So var should be SNDRV_PCM_HW_PARAM_BUFFER_SIZE, dep SNDRV_PCM_HW_PARAM_CHANNELS, but it is the opposite now.

Quoted 28 lines
Quoted 27 lines

This bug didn't
have much impact because the rules are evaluated multiple times by the core, and intended value got set eventually.

---
 sound/soc/omap/omap-mcbsp.c |    4 ++--

 1 files changed, 2 insertions(+), 2 deletions(-)
sound/soc/omap/omap-mcbsp.c +2 -2
index 10eb645..207365c 100644
@@ -142,10 +142,10 @@ static int omap_mcbsp_dai_startup(struct snd_pcm_substream *substream,
               * smaller buffer than the FIFO size to avoid underruns
               */
               snd_pcm_hw_rule_add(substream->runtime, 0,
-                                 SNDRV_PCM_HW_PARAM_CHANNELS,
+                                 SNDRV_PCM_HW_PARAM_BUFFER_SIZE,
                                   omap_mcbsp_hwrule_min_buffersize,
                                   mcbsp,
-                                 SNDRV_PCM_HW_PARAM_BUFFER_SIZE, -1);
+                                 SNDRV_PCM_HW_PARAM_CHANNELS, -1);
 
               /* Make sure, that the period size is always even */
               snd_pcm_hw_constraint_step(substream->runtime, 0,
Signature · 5 lines
Gražvydas
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#4 lore raw

Re: [PATCH 1/3] ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

Quoted 7 lines · Grazvydas Ignotas

That's all fine, but the argument order is wrong, take a look at snd_pcm_hw_rule_add definition:

/**
* snd_pcm_hw_rule_add - add the hw-constraint rule
* @var: the variable to evaluate
* @dep: the dependent variables

True.

Signature · 1 line
Péter
Acked-by on v1 1/3
#5 lore raw

Re: [PATCH 1/3] ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

In reply to Grazvydas Ignotas
Quoted 6 lines · Grazvydas Ignotas

We are setting SNDRV_PCM_HW_PARAM_BUFFER_SIZE based on SNDRV_PCM_HW_PARAM_CHANNELS, not vice versa. This bug didn't have much impact because the rules are evaluated multiple times by the core, and intended value got set eventually.

Acked-by on v1 1/3
#6 lore raw

Re: [PATCH 1/3] ASoC: omap-mcbsp: fix snd_pcm_hw_rule_add arguments

In reply to Grazvydas Ignotas
Quoted 4 lines · Grazvydas Ignotas

We are setting SNDRV_PCM_HW_PARAM_BUFFER_SIZE based on SNDRV_PCM_HW_PARAM_CHANNELS, not vice versa. This bug didn't have much impact because the rules are evaluated multiple times by the core, and intended value got set eventually.

#7 lore raw

PATCH 2/3 ASoC: omap-mcbsp: place size constrain on period, not buffer

In reply to Grazvydas Ignotas

Placing a constraint on buffer size is not enough, programs are still free to use periods that are smaller than FIFO, which causes transmit underflows as soon as the stream starts if program operates in period-sized writes, ending up in endless loop of underruns.

---
 sound/soc/omap/omap-mcbsp.c |   18 +++++++++---------
 1 files changed, 9 insertions(+), 9 deletions(-)
sound/soc/omap/omap-mcbsp.c +9 -9
index 207365c..b9d1272 100644
@@ -93,11 +93,11 @@ static void omap_mcbsp_set_threshold(struct snd_pcm_substream *substream)
 		omap_mcbsp_set_rx_threshold(mcbsp, words);
 }
 
-static int omap_mcbsp_hwrule_min_buffersize(struct snd_pcm_hw_params *params,
-				    struct snd_pcm_hw_rule *rule)
+static int omap_mcbsp_hwrule_min_period_size(struct snd_pcm_hw_params *params,
+					     struct snd_pcm_hw_rule *rule)
 {
-	struct snd_interval *buffer_size = hw_param_interval(params,
-					SNDRV_PCM_HW_PARAM_BUFFER_SIZE);
+	struct snd_interval *period_size = hw_param_interval(params,
+					SNDRV_PCM_HW_PARAM_PERIOD_SIZE);
 	struct snd_interval *channels = hw_param_interval(params,
 					SNDRV_PCM_HW_PARAM_CHANNELS);
 	struct omap_mcbsp *mcbsp = rule->private;
@@ -109,7 +109,7 @@ static int omap_mcbsp_hwrule_min_buffersize(struct snd_pcm_hw_params *params,
 
 	frames.min = size / channels->min;
 	frames.integer = 1;
-	return snd_interval_refine(buffer_size, &frames);
+	return snd_interval_refine(period_size, &frames);
 }
 
 static int omap_mcbsp_dai_startup(struct snd_pcm_substream *substream,
@@ -138,12 +138,12 @@ static int omap_mcbsp_dai_startup(struct snd_pcm_substream *substream,
 	 */
 	if (mcbsp->pdata->buffer_size) {
 		/*
-		* Rule for the buffer size. We should not allow
-		* smaller buffer than the FIFO size to avoid underruns
+		* Rule for the period size. We should not allow
+		* smaller period than the FIFO size to avoid underruns
 		*/
 		snd_pcm_hw_rule_add(substream->runtime, 0,
-				    SNDRV_PCM_HW_PARAM_BUFFER_SIZE,
-				    omap_mcbsp_hwrule_min_buffersize,
+				    SNDRV_PCM_HW_PARAM_PERIOD_SIZE,
+				    omap_mcbsp_hwrule_min_period_size,
 				    mcbsp,
 				    SNDRV_PCM_HW_PARAM_CHANNELS, -1);
 
Signature · 1 line
1.7.0.4
Acked-by on v1 2/3
#8 lore raw

Re: [PATCH 2/3] ASoC: omap-mcbsp: place size constrain on period, not buffer

Quoted 4 lines · Grazvydas Ignotas

Placing a constraint on buffer size is not enough, programs are still free to use periods that are smaller than FIFO, which causes transmit underflows as soon as the stream starts if program operates in period-sized writes, ending up in endless loop of underruns.

#9 lore raw

PATCH 3/3 ASoC: omap-mcbsp: make minimum period size larger than FIFO

In reply to Grazvydas Ignotas

With a program operating in minimum sized periods, first write will cause a transfer that will fill mcbsp FIFO quickly, and there will be no more data to DMA before program manages to do another write. As the core considers this as underflow condition, we may get many underflows at the start.
Increase minimum period size by half to deal with this problem.

---
 sound/soc/omap/omap-mcbsp.c |   11 +++++++++++
 1 files changed, 11 insertions(+), 0 deletions(-)
sound/soc/omap/omap-mcbsp.c +11 -0
index b9d1272..b373a0b 100644
@@ -108,6 +108,17 @@ static int omap_mcbsp_hwrule_min_period_size(struct snd_pcm_hw_params *params,
 	size = mcbsp->pdata->buffer_size;
 
 	frames.min = size / channels->min;
+
+	/*
+	 * If period is not larger than FIFO, it may be transfered faster than
+	 * program operating in period sizes is able to send 2 periods, so
+	 * underrun condition can be triggered at the beginning of stream.
+	 * This is because underrun is triggered as soon as DMA has no more
+	 * data to send, and we get this from the start if period is too small.
+	 * To deal with this, set period larger than FIFO size.
+	 */
+	frames.min = frames.min + frames.min / 2;
+
 	frames.integer = 1;
 	return snd_interval_refine(period_size, &frames);
 }
Signature · 1 line
1.7.0.4
Acked-by on v1 3/3
#10 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

sound/soc/omap/omap-mcbsp.c static int omap_mcbsp_hwrule_min_period_size(struct snd_pcm_hw_params *params, · line 120

Quoted 1 line · Grazvydas Ignotas
sound/soc/omap/omap-mcbsp.c +1 -0
+	frames.min = frames.min + frames.min / 2;

Call me picky but I'd like some brackets here as I had to think for a few seconds if that was (frames.min + frames.min) / 2. But

#11 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

In reply to Grazvydas Ignotas
Quoted 6 lines · Grazvydas Ignotas

With a program operating in minimum sized periods, first write will cause a transfer that will fill mcbsp FIFO quickly, and there will be no more data to DMA before program manages to do another write. As the core considers this as underflow condition, we may get many underflows at the start.
Increase minimum period size by half to deal with this problem.

Not entirely sure about the last two patch...
The implications will be big for existing user space since we are going to limit the period size in this way (on OMAP3 McBSP2):
48K stereo: 20ms min
44.1K stereo: 21.768ms min
8K stereo: 120ms min
8K mono: 240ms min

As I recall n900 (+ MeeGo) uses 5ms period sizes in call cases on McBSP2.

So I'm a bit reluctant to ack these (it is not an issue on other McBSP ports, and on OMAP4, but OMAP3 McBSP2 is problematic IMHO).

Jarkko: What do you think?

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

Signature · 1 line
Péter
Quoted 28 lines
---
 sound/soc/omap/omap-mcbsp.c |   11 +++++++++++
 1 files changed, 11 insertions(+), 0 deletions(-)
sound/soc/omap/omap-mcbsp.c +11 -0
index b9d1272..b373a0b 100644
@@ -108,6 +108,17 @@ static int omap_mcbsp_hwrule_min_period_size(struct snd_pcm_hw_params *params,
 	size = mcbsp->pdata->buffer_size;
 
 	frames.min = size / channels->min;
+
+	/*
+	 * If period is not larger than FIFO, it may be transfered faster than
+	 * program operating in period sizes is able to send 2 periods, so
+	 * underrun condition can be triggered at the beginning of stream.
+	 * This is because underrun is triggered as soon as DMA has no more
+	 * data to send, and we get this from the start if period is too small.
+	 * To deal with this, set period larger than FIFO size.
+	 */
+	frames.min = frames.min + frames.min / 2;
+
 	frames.integer = 1;
 	return snd_interval_refine(period_size, &frames);
 }
#12 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 16 lines · Peter Ujfalusi
Quoted 6 lines · Grazvydas Ignotas

With a program operating in minimum sized periods, first write will cause a transfer that will fill mcbsp FIFO quickly, and there will be no more data to DMA before program manages to do another write. As the core considers this as underflow condition, we may get many underflows at the start.
Increase minimum period size by half to deal with this problem.

Not entirely sure about the last two patch...
The implications will be big for existing user space since we are going to limit the period size in this way (on OMAP3 McBSP2):
48K stereo: 20ms min
44.1K stereo: 21.768ms min
8K stereo: 120ms min
8K mono: 240ms min

Quoted 1 line

As I recall n900 (+ MeeGo) uses 5ms period sizes in call cases on McBSP2.

I would guess they are using something larger, at least larger write quantities with this kind of period, otherwise it would keep overflowing there too.

Quoted 2 lines

So I'm a bit reluctant to ack these (it is not an issue on other McBSP ports, and on OMAP4, but OMAP3 McBSP2 is problematic IMHO).

Maybe we could use MCBSPLP_THRSH2_REG XTHRESHOLD to artificially limit MCBSP2 fifo depth?

Quoted 4 lines

Jarkko: What do you think?

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

We are already doing that, however we encourage people to install their own distros on pandora, and now they have to drag whole ALSA configuration for things to work. IMHO device driver like this must be able to work without any special userspace configuration.

Signature · 5 lines
Gražvydas
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#13 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 3 lines · Grazvydas Ignotas

I would guess they are using something larger, at least larger write quantities with this kind of period, otherwise it would keep overflowing there too.

They are using PulseAudio as with Maemmo (the product software).

Quoted 5 lines
Quoted 2 lines

So I'm a bit reluctant to ack these (it is not an issue on other McBSP ports, and on OMAP4, but OMAP3 McBSP2 is problematic IMHO).

Maybe we could use MCBSPLP_THRSH2_REG XTHRESHOLD to artificially limit MCBSP2 fifo depth?

It is not possible. The McBSP FIFO/sDMA implementation (in HW) does not allow it, unfortunately.
We are using the Threshold settings in dma_op_mode == threshold:
echo threshold > /sys/devices/platform/omap/omap-mcbsp.2/dma_op_mode

It is good for achieving lover power consumption during audio activity.

Quoted 9 lines
Quoted 4 lines

Jarkko: What do you think?

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

We are already doing that, however we encourage people to install their own distros on pandora, and now they have to drag whole ALSA configuration for things to work. IMHO device driver like this must be able to work without any special userspace configuration.

We have two issues here:
if you are using the element mode (dma_op_mode) you might want to have bigger period size than the FIFO size. However if you are using the threshold mode you will be fine with a period size between fifo_size/2, and fifo_size.
In threshold mode we are sending the data to McBSP FIFO in bursts, and the burst size is the same as the McBSP threshold config. The McBSP will request new burst when there is free space for the next burst in the FIFO. So we can fill the FIFO to 3/4 for example, and the next burst will happen when the FIFO reaches 1/4 fill level.

You might want to try this as well to lover the period_size constraint.

But let's wait for Jarkko's take on this.

Signature · 5 lines
Péter
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#14 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 13 lines · Peter Ujfalusi
Quoted 3 lines · Grazvydas Ignotas

I would guess they are using something larger, at least larger write quantities with this kind of period, otherwise it would keep overflowing there too.

They are using PulseAudio as with Maemmo (the product software).

Quoted 5 lines
Quoted 2 lines

So I'm a bit reluctant to ack these (it is not an issue on other McBSP ports, and on OMAP4, but OMAP3 McBSP2 is problematic IMHO).

Maybe we could use MCBSPLP_THRSH2_REG XTHRESHOLD to artificially limit MCBSP2 fifo depth?

Quoted 6 lines

It is not possible. The McBSP FIFO/sDMA implementation (in HW) does not allow it, unfortunately.
We are using the Threshold settings in dma_op_mode == threshold:
echo threshold > /sys/devices/platform/omap/omap-mcbsp.2/dma_op_mode

It is good for achieving lover power consumption during audio activity.

Maybe enable it by default then?

Quoted 10 lines
Quoted 9 lines
Quoted 4 lines

Jarkko: What do you think?

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

We are already doing that, however we encourage people to install their own distros on pandora, and now they have to drag whole ALSA configuration for things to work. IMHO device driver like this must be able to work without any special userspace configuration.

Quoted 5 lines

We have two issues here:
if you are using the element mode (dma_op_mode) you might want to have bigger period size than the FIFO size. However if you are using the threshold mode you will be fine with a period size between fifo_size/2, and fifo_size.

Threshold mode doesn't seem to help, I still need to go over FIFO size. From what I can see in the code it's setting the threshold to period size, so I still get underflow condition after first write, as first write goes to FIFO, DMA shows that it has nothing left to send after first write finishes.

I bet PulseAudio is setting start threshold to something way above FIFO size, which makes it work here. However you can't expect that from every program that uses ALSA directly..

Quoted 5 lines

In threshold mode we are sending the data to McBSP FIFO in bursts, and the burst size is the same as the McBSP threshold config. The McBSP will request new burst when there is free space for the next burst in the FIFO. So we can fill the FIFO to 3/4 for example, and the next burst will happen when the FIFO reaches 1/4 fill level.

That would work if I did first write larger than period (when it's below FIFO size) or set large enough start_threshold. However if you look at alsalib samples, they always write in period size quantities, so it's not surprising other programs do that too (alsalib samples work here though as they do set high start threshold).

Quoted 7 lines

You might want to try this as well to lover the period_size constraint.

But let's wait for Jarkko's take on this.

--
Péter

Signature · 5 lines
Gražvydas
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#15 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 10 lines · Grazvydas Ignotas
Quoted 9 lines · Peter Ujfalusi
Quoted 7 lines · Grazvydas Ignotas
Quoted 2 lines

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

We are already doing that, however we encourage people to install their own distros on pandora, and now they have to drag whole ALSA configuration for things to work. IMHO device driver like this must be able to work without any special userspace configuration.

I agree. What's the use case where the problem shows up? Or actually hw parameters in that use case so can it be repeated using plain aplay/arecord?

Quoted 6 lines
Quoted 5 lines

We have two issues here:
if you are using the element mode (dma_op_mode) you might want to have bigger period size than the FIFO size. However if you are using the threshold mode you will be fine with a period size between fifo_size/2, and fifo_size.

Quoted 6 lines

Threshold mode doesn't seem to help, I still need to go over FIFO size. From what I can see in the code it's setting the threshold to period size, so I still get underflow condition after first write, as first write goes to FIFO, DMA shows that it has nothing left to send after first write finishes.

Period size smaller than FIFO size has used to work in element mode too but I think I have always used in my tests multiple periods. What I'm thinking if we have a boundary condition with 2-3 periods and where buffer size is near FIFO size.

I think instead of limiting minimum period size to FIFO size we should find that boundary condition. Like if min buffer > FIFO + 1 period or something like that in order to keep possible to use small periods.

Signature · 1 line
Jarkko
#16 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 6 lines · Jarkko Nikula
Quoted 5 lines · Grazvydas Ignotas
Quoted 4 lines · Peter Ujfalusi
Quoted 3 lines · Grazvydas Ignotas
Quoted 2 lines

If this is a real issue in Pandora, could it be solved with adding /etc/asound.conf to the filesystem, and limit the period size via that?

Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

We are already doing that, however we encourage people to install their own distros on pandora, and now they have to drag whole ALSA configuration for things to work. IMHO device driver like this must be able to work without any special userspace configuration.

I agree. What's the use case where the problem shows up?

When alsa program doesn't set high enough start_threshold and period size, or when using OSS with smaller period (start_threshold is period size there).

Quoted 3 lines

Or actually hw
parameters in that use case so can it be repeated using plain aplay/arecord?

aplay -D hw:0 -f s16_le -c 2 --period-size=512 --start-delay=50000 /dev/urandom

note that this will only show problems on OMAP3 McBSP2 due to it's long FIFO.

Quoted 12 lines
Quoted 6 lines
Quoted 5 lines

We have two issues here:
if you are using the element mode (dma_op_mode) you might want to have bigger period size than the FIFO size. However if you are using the threshold mode you will be fine with a period size between fifo_size/2, and fifo_size.

Quoted 6 lines

Threshold mode doesn't seem to help, I still need to go over FIFO size. From what I can see in the code it's setting the threshold to period size, so I still get underflow condition after first write, as first write goes to FIFO, DMA shows that it has nothing left to send after first write finishes.

Quoted 4 lines

Period size smaller than FIFO size has used to work in element mode too but I think I have always used in my tests multiple periods. What I'm thinking if we have a boundary condition with 2-3 periods and where buffer size is near FIFO size.

I'm not sure what boundary means in ALSA, but ASoC never touches it or start_threshold.

Quoted 3 lines

I think instead of limiting minimum period size to FIFO size we should find that boundary condition. Like if min buffer > FIFO + 1 period or something like that in order to keep possible to use small periods.

I was testing some hacks that force runtime->start_threshold to a bit above FIFO size with good results, however none of ASoC code ever touches it, so I guess it could be considered wrong thing to do, but it works with everything I tested so far, including OSS emulation.

Signature · 5 lines
Gražvydas
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#17 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 10 lines · Grazvydas Ignotas
Quoted 3 lines · Jarkko Nikula

I think instead of limiting minimum period size to FIFO size we should find that boundary condition. Like if min buffer > FIFO + 1 period or something like that in order to keep possible to use small periods.

I was testing some hacks that force runtime->start_threshold to a bit above FIFO size with good results, however none of ASoC code ever touches it, so I guess it could be considered wrong thing to do, but it works with everything I tested so far, including OSS emulation.

That's more due to nothing ever particularly needing it before than due to it being a bad idea.

#18 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

In reply to Grazvydas Ignotas
Quoted 3 lines · Grazvydas Ignotas

aplay -D hw:0 -f s16_le -c 2 --period-size=512 --start-delay=50000 /dev/urandom

note that this will only show problems on OMAP3 McBSP2 due to it's long FIFO.

There's something else here since:
aplay -D hw:0 -fs16_le -c2 --period-size=320 --buffer-size=972 \
--start-delay=81000 -v /dev/urandom

Works correctly, however
aplay -D hw:0 -fs16_le -c2 --period-size=324 --buffer-size=972 \
--start-delay=81000 -v /dev/urandom

produces underruns.
In both cases the start_threshold is set to 648 (640 is the FIFO McBSP2 size). Both should be fine, but period_size=324 underruns.

So the issue is not that simple as it seams...

Signature · 5 lines
Péter
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#19 lore raw

Re: [PATCH 3/3] ASoC: omap-mcbsp: make minimum period size larger than FIFO

Quoted 13 lines · Peter Ujfalusi
Quoted 3 lines · Grazvydas Ignotas

aplay -D hw:0 -f s16_le -c 2 --period-size=512 --start-delay=50000 /dev/urandom

note that this will only show problems on OMAP3 McBSP2 due to it's long FIFO.

There's something else here since:
aplay -D hw:0 -fs16_le -c2 --period-size=320 --buffer-size=972 \
--start-delay=81000 -v /dev/urandom

Works correctly, however
aplay -D hw:0 -fs16_le -c2 --period-size=324 --buffer-size=972 \
--start-delay=81000 -v /dev/urandom

Quoted 5 lines

produces underruns.
In both cases the start_threshold is set to 648 (640 is the FIFO McBSP2 size). Both should be fine, but period_size=324 underruns.

So the issue is not that simple as it seams...

Seems to be affected --stop-delay, increasing it helps here, maybe core stops it for too long.

Signature · 5 lines
Gražvydas
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel