To disable EDMA for MCASP audio in ti sdk 8

I'm forwarding this to the SW team. By the way, see this wiki too: processors.wiki.ti.com/.../Linux_Core_Audio_User%27s_Guide

Hi Srini,

Remove the CONFIG_SND_EDMA_SOC from your defconfig file.
Also try reverting audio driver sources to the older ones (from SDK7). You can use a diff tool, i.e. meld, to see differences between audio drivers in SDK7.0 & SDK8.0 and quickly make the changes.

Best Regards,
Yordan

Dear Yordan,

Thanks a lot for your suggestions will definitely start it reverting to sdk 7

If am using 8K sampling rate, How can I verify that am receiving full 8000 samples in the mcasp buffers??

Yordan, could you please let me know Is it possible that how can I print & verify in mcasp buffer, that I have received 16 bit 8000 samples, could you pls let me know with what base address & the offset I can verify inorder to ensure am getting 8000 samples

Could you please help me out how can I print these Samples values using printk,

If it is unable to print full 8000 samples, Am eager to verify the first 10 samples (base address) & the last 10 samples (base address + end of offset ie., end address of 8K buffer) inorder to ensure that am receiving all 8000 samples

After I verify mcasp buffer, could you please let me know how this can be verified in alsa layer inorder to ensure that am receiving & transmitting full 8000 samples


Could you please, help me out in this, If I verify this, I think would be able to fix in ti sdk 8 itself

Kindly do the needful as early as possible

Awaiting for your replies,
Many Thanks in advance again

Srini said:

could you please let me know Is it possible that how can I print & verify in mcasp buffer, that I have received 16 bit 8000 samples, could you pls let me know with what base address & the offset I can verify inorder to ensure am getting 8000 samples

 

Have you tried reading the RBUF_0 to RBUF_5 registers? Check AM335x TRM, Section 22.4.1 MCASP Registers. 

MCASP0 Data regs base address is: 0x4600_0000 to 0x463F_FFFF       

MCASP1 Data regs base address is: 0x4640_0000 to 0x467F_FFFF

MCASP0 CFG regs base address is: 0x4803_8000 to 0x4803_9FFF

MCASP1 CFG regs base address is: 0x4803_C000 to 0x4803_DFFF

Best Regards, 

Yordan

Dear Yordan,

Thanks a lot for your replies & sorry to push you hard will verify your inputs & keep you updated for sure

Am verifying all these things because am seeing lots of overrun & broken pipes in alsa application

A. As per TRM page number 4593, does the event AXEVT gets generated every 16 bits of data transfers??

B .Am using 16 bit 8 Khz sampling rate, I2S mode , could you please let me know what is the correct value that needs to be configured in tx-num-evt & rx-num-evt????

Currently am configuring as per below is this correct???

tx-num-evt = <1>;

rx-num-evt = <1>;

If I set the tx-num-evt = <1>; & rx-num-evt = <1>;

I see that means for one DMA transfers am I transfering 32 bit of data or 16 bit of data????? please correct me if my understanding is correct

C. Could you please let me know in what order of  the sequence  the below files gets called by ALSA ??

1. davinci-evm.c

2. davinci-mcasp.c

3. davinci-i2s.c

4. davinci-pcm.c

5. davinci-vcif.c

6.edma-pcm.c

D. Am trying to calculate buffer_size in alsa application, could you please let me know how this is calculated ie., buffersize = period_size * period (am using 16 bit 8Khz sampling rate) & set in ALSA application

I hope that the period_size & period are MCASP dependent, could you please let me know how to calculate period_size & periods w.r.t MCASP

E.In davinci-pcm.c, In the below structures, am unable to understand how this values are derived & set

Could you please explain: How are the following values derived:

    .buffer_bytes_max = 128 * 1024,
    .period_bytes_min = 32,
    .period_bytes_max = 8 * 1024,
    .periods_min = 16,
    .periods_max = 255,

The size available for McASP is 4MB

static struct snd_pcm_hardware pcm_hardware_playback = {

.info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER |
SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME|
SNDRV_PCM_INFO_BATCH),
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 32,
.period_bytes_max = 8 * 1024,
.periods_min = 16,
.periods_max = 255,
.fifo_size = 0,
};

static struct snd_pcm_hardware pcm_hardware_capture = {
.info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER |
SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_PAUSE |
SNDRV_PCM_INFO_BATCH),
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 32,
.period_bytes_max = 8 * 1024,
.periods_min = 16,
.periods_max = 255,
.fifo_size = 0,
};

Dear Yordan could you please provide your guidance w.r.t the above queries, Hope that would be useful in resolving all the issues w.r.t XRUNS, broken pipes, etc..,

Once again really appreciate for your prompt responses w.r..t this thread

Kindly do the needful as early as possible

Many Many Thanks in advance

Dear Yordan,
Thanks a lot for your replies & sorry to push you hard will verify your inputs & keep you updated for sure

Am verifying all these things because am seeing lots of overrun & broken pipes in alsa application

A. As per TRM page number 4593, does the event AXEVT gets generated every 16 bits of data transfers??

B .Am using 16 bit 8 Khz sampling rate, I2S mode , could you please let me know what is the correct value that needs to be configured in tx-num-evt & rx-num-evt????
Currently am configuring as per below is this correct???
tx-num-evt = <1>;
rx-num-evt = <1>;
If I set the tx-num-evt = <1>; & rx-num-evt = <1>;
I see that means for one DMA transfers am I transfering 32 bit of data or 16 bit of data????? please correct me if my understanding is correct

C. Could you please let me know in what order of the sequence the below files gets called by ALSA ??
1. davinci-evm.c
2. davinci-mcasp.c
3. davinci-i2s.c
4. davinci-pcm.c
5. davinci-vcif.c
6.edma-pcm.c

D. Am trying to calculate buffer_size in alsa application, could you please let me know how this is calculated ie., buffersize = period_size * period (am using 16 bit 8Khz sampling rate) & set in ALSA application
I hope that the period_size & period are MCASP dependent, could you please let me know how to calculate period_size & periods w.r.t MCASP

E.In davinci-pcm.c, In the below structures, am unable to understand how this values are derived & set

Could you please explain: How are the following values derived:
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 32,
.period_bytes_max = 8 * 1024,
.periods_min = 16,
.periods_max = 255,
The size available for McASP is 4MB
static struct snd_pcm_hardware pcm_hardware_playback = {
.info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER |
SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME|
SNDRV_PCM_INFO_BATCH),
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 32,
.period_bytes_max = 8 * 1024,
.periods_min = 16,
.periods_max = 255,
.fifo_size = 0,
};
static struct snd_pcm_hardware pcm_hardware_capture = {
.info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER |
SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_PAUSE |
SNDRV_PCM_INFO_BATCH),
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 32,
.period_bytes_max = 8 * 1024,
.periods_min = 16,
.periods_max = 255,
.fifo_size = 0,
};

Dear Yordan could you please provide your guidance w.r.t the above queries, Hope that would be useful in resolving all the issues w.r.t XRUNS, broken pipes, etc..,

Once again really appreciate for your prompt responses w.r..t this thread

Kindly do the needful as early as possible
Many Many Thanks in advance

Dear Yordan,

As am still trying to fix this issue with EDMA feature  ie., -EPIPE error in alsa application

Could you please provide any inputs w.r.t the above queries ie., A, B, C D & E points, Hope your inputs will help me to resolve all the above issues

Mainly could you please address the below query ie.,

Am trying to calculate buffer_size in alsa application, could you please let me know how this is calculated ie., buffersize = period_size * period (am using 16 bit 8Khz sampling rate) & set in ALSA application
I hope that the period_size & period are MCASP dependent, could you please let me know how to calculate period_size & periods w.r.t MCASP

In point no. C, I dont think so we are using davinci-pcm.c, davinci-i2s.c & davinci-vcif.c for MCASP & for UDa1345ts codec

Could you please let me know how this values are derived in edma-pcm.c

.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 512,
.period_bytes_max = 64 * 1024,
.periods_min = 2,
.periods_max = 19, /* Limit by edma dmaengine driver */

static const struct snd_pcm_hardware edma_pcm_hardware = {
.info = SNDRV_PCM_INFO_MMAP |
SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME |
SNDRV_PCM_INFO_NO_PERIOD_WAKEUP |
SNDRV_PCM_INFO_INTERLEAVED,
.buffer_bytes_max = 128 * 1024,
.period_bytes_min = 512,
.period_bytes_max = 64 * 1024,
.periods_min = 2,
.periods_max = 19, /* Limit by edma dmaengine driver */
};

 

Kindly do the needful as early as possible

Thanks in advance

Dear Yordan,

Am was going through reference manaul as per below section it says that it is unable to access RBUF's could you please clearly  let me know how to print these RBUF sample values

As still am using ti sdk 8 with EDMA feature, could you please let me know

I don't see the MACRO DAVINCI_MCASP_RXBUF_REG 0x280 being used anywhere in the code to print the samples , could you please let me know


Could you please help me out where exactly & how to add the printk's in davinci-mcasp.c, could you please add & send across the patch Yordan

Kindly do the needful as early as possible, as am trying to isolate this issue from driver side

Many Thanks in advance once again