Problem with buffer/period sizes.
I have code that looks like the following: -
err = snd_pcm_hw_params_get_buffer_size_min(params, &buffer_size_min); err = snd_pcm_hw_params_get_buffer_size_max(params, &buffer_size_max); dir=0;
err = snd_pcm_hw_params_get_period_size_min(params, &period_size_min,&dir); dir=0;
err = snd_pcm_hw_params_get_period_size_max(params, &period_size_max,&dir);
<snip> some calculations happen here
err = snd_pcm_hw_params_set_buffer_size_near(audio_fd, params, &buffer_size); dir=0;
err = snd_pcm_hw_params_set_period_size_near(audio_fd, params, &period_size, &dir);
For a snd_intel8x0 driver with INTEL ICH5 chip with PCI IDs of 0x8086, 0x24d5
Displays: -
buffer_size_min/max = 64 / 16384 period_size_min/max = 16 / 16384
We then try to set: -buffer_size = 16384period_size = 2048
And it accepts those settings, and works.
For a snd_intel8x0 driver with NVIDIA NFORCE chip with PCI IDs of 0x10de, 0x01b1
Displays: -
buffer_size_min/max = 67 / 15052 period_size_min/max = 7 / 15053
We then try to set: -buffer_size = 15048period_size = 1881
But, this sets a period size = buffer_size = 15048 which then fails to work due to not having 2 periods or more.
What is causing the NFORCE chip to suggest such strange values, and then to not accept those values.
Cheers
James
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. SourceForge.net hosts over 70,000 Open Source Projects. See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
Quoted 34 lines · James Courtier-Dutton
I have code that looks like the following: -
err =
snd_pcm_hw_params_get_buffer_size_min(params, &buffer_size_min); err =snd_pcm_hw_params_get_buffer_size_max(params, &buffer_size_max); dir=0;
err =snd_pcm_hw_params_get_period_size_min(params, &period_size_min,&dir); dir=0;
err =snd_pcm_hw_params_get_period_size_max(params, &period_size_max,&dir);<snip> some calculations happen here
err =
snd_pcm_hw_params_set_buffer_size_near(audio_fd, params, &buffer_size);
dir=0;
err =snd_pcm_hw_params_set_period_size_near(audio_fd, params, &period_size, &dir);For a
snd_intel8x0driver with INTEL ICH5 chip with PCI IDs of0x8086,0x24d5
Displays: -
buffer_size_min/max = 64 / 16384 period_size_min/max = 16 / 16384We then try to set: -
buffer_size= 16384period_size= 2048
And it accepts those settings, and works.For a
snd_intel8x0driver with NVIDIA NFORCE chip with PCI IDs of0x10de,0x01b1
Displays: -
buffer_size_min/max = 67 / 15052 period_size_min/max = 7 / 15053
Quoted 8 lines
We then try to set: -
buffer_size= 15048period_size= 1881
But, this sets a period size =buffer_size= 15048 which then fails to work due to not having 2 periods or more.What is causing the NFORCE chip to suggest such strange values, and then to not accept those values.
It appears that the NFORCE chip is clocked with 44.1kHz. Thus alsa-lib is trying to do the resampling from 44.1kHz to 48kHz. Unfortunately, we're fighting with some rounding errors so that we cannot guarantee that all hw parameter combinations will work.
Because this situation is quite common for cheap motherboards where the hardware engineers are so clever to spare one clock crystal thus make the audio chips work with absolutely non-standard frequencies, I suggest to find a working combination of values and use it as last resort.
Note that we will solve these hardware dependand things with the new "ordinary" PCM. We will guarantee that you'll have always working setup for given hardware there.
Jaroslav
-----
Jaroslav Kysela <[email protected]>
Linux Kernel Sound Maintainer
ALSA Project, SuSE Labs
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. SourceForge.net hosts over 70,000 Open Source Projects. See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
Quoted 57 lines · Jaroslav Kysela
Quoted 34 lines · James Courtier-Dutton
I have code that looks like the following: -
err =
snd_pcm_hw_params_get_buffer_size_min(params, &buffer_size_min); err =snd_pcm_hw_params_get_buffer_size_max(params, &buffer_size_max); dir=0;
err =snd_pcm_hw_params_get_period_size_min(params, &period_size_min,&dir); dir=0;
err =snd_pcm_hw_params_get_period_size_max(params, &period_size_max,&dir);<snip> some calculations happen here
err =
snd_pcm_hw_params_set_buffer_size_near(audio_fd, params, &buffer_size);
dir=0;
err =snd_pcm_hw_params_set_period_size_near(audio_fd, params, &period_size, &dir);For a
snd_intel8x0driver with INTEL ICH5 chip with PCI IDs of0x8086,0x24d5
Displays: -
buffer_size_min/max = 64 / 16384 period_size_min/max = 16 / 16384We then try to set: -
buffer_size= 16384period_size= 2048
And it accepts those settings, and works.For a
snd_intel8x0driver with NVIDIA NFORCE chip with PCI IDs of0x10de,0x01b1
Displays: -
buffer_size_min/max = 67 / 15052 period_size_min/max = 7 / 15053Quoted 8 lines
We then try to set: -
buffer_size= 15048period_size= 1881
But, this sets a period size =buffer_size= 15048 which then fails to work due to not having 2 periods or more.What is causing the NFORCE chip to suggest such strange values, and then to not accept those values.
It appears that the NFORCE chip is clocked with 44.1kHz. Thus alsa-lib is trying to do the resampling from 44.1kHz to 48kHz. Unfortunately, we're fighting with some rounding errors so that we cannot guarantee that all hw parameter combinations will work.
Because this situation is quite common for cheap motherboards where the hardware engineers are so clever to spare one clock crystal thus make the audio chips work with absolutely non-standard frequencies, I suggest to find a working combination of values and use it as last resort.
Quoted 6 lines
Note that we will solve these hardware dependand things with the new "ordinary" PCM. We will guarantee that you'll have always working setup for given hardware there.
Jaroslav
Thanks. In the mean time, I found a solution be reversing the order in which I set period and buffer sizes, setting period size first, as suggested by yourselves in previous mailing list items.
Cheers
James
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. SourceForge.net hosts over 70,000 Open Source Projects. See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php