Skip to contentSkip to conversation

do apps need to care for xfer_align ???

DISCUSSION Jaroslav Kysela 2 participants 4 messages Sep 9, 2002 – Sep 10, 2002 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 21 lines · Paul Davis
Quoted 21 lines · Guilhem Tardy
Quoted 5 lines

Hi all,

While trying to understand snd_pcm_lib_write1(), I found that runtime->xfer_align is set to period_size in pcm_native.c, but wonder if it

is

Quoted 3 lines

just a default?

Paul Davis' example (at http://www.op.net/~pbd/alsa-audio.html#playex) doesn

't

Quoted 1 line

check for a particular transfer size, and my card supports playout of less t

han

Quoted 1 line

160 samples (i.e. the period size). Thus, I set to try playing out 512 bytes

at

Quoted 1 line

a time (i.e. 2048 samples of 2-bit ADPCM), which does NOT come to a multiple

of

Quoted 1 line

period size. The result is that only 1920 samples are played, and my app thi

nks

Quoted 2 lines

an error has occured.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Is it a general requirement for apps to send only audio in multiple of the period size?

Nope, applications might modify xfer_align to satisfy their requirements.

jaroslav, since even i don't understand what xfer_align is for, or even have an answer to the actual question, it might be good for you to expand on this a little.

It's software only limit for read/write transfers. The direct transfers (mmap) don't take care about this limit. The idea is to avoid using very small transfer chunks which are inefficient. Application might change this value from 1 (no control) to buffer_size (in samples), but values bigger than period_size hasn't any sense anyway.

The default value for xfer_align is period_size.

Jaroslav

-----
Jaroslav Kysela <[email protected]>
SuSE Linux http://www.suse.com
ALSA project http://www.alsa-project.org

-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old cell phone? Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390

#2 lore raw
Quoted 7 lines

It's software only limit for read/write transfers. The direct transfers (mmap) don't take care about this limit. The idea is to avoid using very small transfer chunks which are inefficient. Application might change this value from 1 (no control) to buffer_size (in samples), but values bigger than period_size hasn't any sense anyway.

The default value for xfer_align is period_size.

Since 'xfer_align' must be equal or greater than 'min_align', how can we get this value through the library?

Thanks,
Guilhem.

__________________________________________________
Do You Yahoo!?
Yahoo! Finance - Get real-time stock quotes
http://finance.yahoo.com

-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old cell phone? Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390

#3 lore raw
Quoted 10 lines · Guilhem Tardy
Quoted 7 lines

It's software only limit for read/write transfers. The direct transfers (mmap) don't take care about this limit. The idea is to avoid using very small transfer chunks which are inefficient. Application might change this value from 1 (no control) to buffer_size (in samples), but values bigger than period_size hasn't any sense anyway.

The default value for xfer_align is period_size.

Since 'xfer_align' must be equal or greater than 'min_align', how can we get this value through the library?

The physical size of xfer_align * frame_size_in_bits must be multiple by 8, of course. Adding bit copying in the kernel or library will cause a lot of overhead.

I've added this function to library:

snd_pcm_sframes_t snd_pcm_hw_params_get_min_align(const snd_pcm_hw_params_t *params)

Jaroslav

-----
Jaroslav Kysela <[email protected]>
Linux Kernel Sound Maintainer
ALSA Project http://www.alsa-project.org
SuSE Linux http://www.suse.com

-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old cell phone? Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390

#4 lore raw

Jaroslav,

I am working on a conferencing application, and need to deal with audio packets of various sizes (in sample).

Quoted 5 lines
Quoted 5 lines

It's software only limit for read/write transfers. The direct transfers (mmap) don't take care about this limit. The idea is to avoid using very small transfer chunks which are inefficient. Application might change this value from 1 (no control) to buffer_size (in samples), but values bigger than period_size hasn't any sense anyway.

The application writer who chooses to send audio in chunks of any size then must (and can) also set xfer_align to 1. Which is what I would do to avoid having:
1) audio systematically dropped at the end of each chunk, and
2) writei() return a value different from the number of frames passed to it
(which is usually treated as an error)

Quoted 3 lines

The physical size of xfer_align * frame_size_in_bits must be multiple by 8, of course. Adding bit copying in the kernel or library will cause a lot of overhead.

Agreed.
To me, this only means that period_size must be a multiple of min_align. Any number of samples given to write1() in alsa-kernel/core/pcm_lib.c should be supported already by the current code, except for the xfer_align check at the top of the function.

Quoted 4 lines

I've added this function to library:

snd_pcm_sframes_t snd_pcm_hw_params_get_min_align(const snd_pcm_hw_params_t *params)

That certainly is helpful, thank you.

Regards,
Guilhem.

__________________________________________________
Yahoo! - We Remember
9-11: A tribute to the more than 3,000 lives lost
http://dir.remember.yahoo.com/tribute

-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old cell phone? Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390