Skip to contentSkip to conversation

usb-audio

DISCUSSION Antonio Willy Malara 3 participants 19 messages Oct 27, 2003 – Oct 31, 2003

lore
Share & tools
lore
#0 lore raw

When is it planned the fix of the usb-audio driver for big endian machines? I'm refferring particularly to that:

	/* FIXME: correct endianess and sign? */

(in function parse_audio_format_i_type (); at line 1990 in file alsa- kernel/usb/usbaudio.c version 1.67)

Willy

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#1 lore raw

At Mon, 27 Oct 2003 19:03:00 +0100,

Quoted 8 lines · Antonio Willy Malara

When is it planned the fix of the usb-audio driver for big endian machines? I'm refferring particularly to that:

	/* FIXME: correct endianess and sign? */

(in function parse_audio_format_i_type (); at line 1990 in file alsa- kernel/usb/usbaudio.c version 1.67)

could you give more information:
which program, which device and what format doesn't it work?

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#2 lore raw
Quoted 3 lines · Takashi Iwai
Quoted 1 line
	/* FIXME: correct endianess and sign? */

could you give more information:
which program, which device and what format doesn't it work?

the system is a powermac, the device is a Griffin iMic, the app is jack version 0.80, the output is:

Sorry. The audio interface "hw:1"doesn't support either of the two hardware sample formats that jack can use.

that is raised from here (sorry for the wordwrap):

if ((err = snd_pcm_hw_params_set_format (handle, hw_params,
SND_PCM_FORMAT_S32)) < 0) {
if ((err = snd_pcm_hw_params_set_format (handle,
hw_params, SND_PCM_FORMAT_S16)) < 0) {
jack_error ("Sorry. The audio interface \"%s\""
"doesn't support either of the two
hardware sample formats that jack can use.",

                                    device_name);
                        return -1;
                }
        }

(/drivers/alsa/alsa_driver.c line 293 from the jack 0.80 tree)

i made this test: if i change the SND_PCM_FORMAT_S32 to _S32_LE the app goes forward but alsalib will raise many broken pipe errors and got no sound

hope this will help

Quoted 1 line

Takashi

willy

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#3 lore raw

At Mon, 27 Oct 2003 20:53:08 +0100,

Quoted 11 lines · Antonio Willy Malara
Quoted 3 lines · Takashi Iwai
Quoted 1 line
	/* FIXME: correct endianess and sign? */

could you give more information:
which program, which device and what format doesn't it work?

the system is a powermac, the device is a Griffin iMic, the app is jack version 0.80, the output is:

Sorry. The audio interface "hw:1"doesn't support either of the two hardware sample formats that jack can use.

then it's a problem of JACK, not ALSA.
the hardware doesn't support 32bit integer but only 24bit packed in 3 bytes.

try plughw:1 instead of hw:1. (you might need to define the ctl for that, too.)

ciao,

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#4 lore raw

Am Dienstag, 28. Oktober 2003 11:11 schrieb Takashi Iwai:

Quoted 12 lines

At Mon, 27 Oct 2003 20:53:08 +0100,

Quoted 9 lines · Antonio Willy Malara
Quoted 4 lines · Takashi Iwai
Quoted 1 line
	/* FIXME: correct endianess and sign? */

could you give more information:
which program, which device and what format doesn't it work?

the system is a powermac, the device is a Griffin iMic, the app is jack version 0.80, the output is:

Quoted 6 lines
Quoted 2 lines

Sorry. The audio interface "hw:1"doesn't support either of the two hardware sample formats that jack can use.

then it's a problem of JACK, not ALSA.
the hardware doesn't support 32bit integer but only 24bit packed in 3 bytes.

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms, ...

My impression is that ALSA assumes that the connected interfaces _always_ allow for the same Endianess as the system. This, of course, isn't always so.
Mplayer:

Checking audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
AF_pre: af format: 2 bps, 2 ch, 44100 hz, big endian signed int
AF_pre: 44100Hz 2ch Signed 16-bit (Big-Endian)
alsa-init: requested format: 44100 Hz, 2 channels, Signed 16-bit (Big-Endian)
alsa-init: soundcard set to emi
alsa-init: format Signed 16-bit (Big-Endian) are not supported by hardware, trying default
alsa9: 44100 Hz/2 channels/4 bpf/65536 bytes buffer/Signed 16 bit Little Endian
AO: [alsa9] 44100Hz 2ch Signed 16-bit (Little-Endian) (2 bps)

alsaplayer either dies with "FIXME: f_unsynchronization is set.Please contact alsaplayer team." or produces static noise.

aplay does get this right, so it seems to use a different method of querying the device.

Have fun*

Niklas

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#5 lore raw

At Tue, 28 Oct 2003 11:41:07 +0100,

Quoted 14 lines · Niklas Werner

Am Dienstag, 28. Oktober 2003 11:11 schrieb Takashi Iwai:

Quoted 12 lines

At Mon, 27 Oct 2003 20:53:08 +0100,

Quoted 9 lines · Antonio Willy Malara
Quoted 4 lines · Takashi Iwai
Quoted 1 line
	/* FIXME: correct endianess and sign? */

could you give more information:
which program, which device and what format doesn't it work?

the system is a powermac, the device is a Griffin iMic, the app is jack version 0.80, the output is:

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

Sorry. The audio interface "hw:1"doesn't support either of the two hardware sample formats that jack can use.

then it's a problem of JACK, not ALSA.
the hardware doesn't support 32bit integer but only 24bit packed in 3 bytes.

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms, ...

did you use plughw instead of hw in all cases?
otherwise they won't work always.

Quoted 5 lines

My impression is that ALSA assumes that the connected interfaces _always_ allow for the same Endianess as the system. This, of course, isn't always so.
Mplayer:

Quoted 8 lines

Checking audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
AF_pre: af format: 2 bps, 2 ch, 44100 hz, big endian signed int
AF_pre: 44100Hz 2ch Signed 16-bit (Big-Endian)
alsa-init: requested format: 44100 Hz, 2 channels, Signed 16-bit (Big-Endian)
alsa-init: soundcard set to emi
alsa-init: format Signed 16-bit (Big-Endian) are not supported by hardware, trying default

so, mplayer likely uses hw layer.

mplayer has an option to specify the device name. it'd be better to define a new pcm in ~/.asoundrc such as

	pcm.emi26 {
		type plug
		slave.pcm "hw:0"
	}

(where "hw:0" should be changed to the corresponding one)

and run like

	% mplayer -ao alsa9:emi26 ...
Quoted 6 lines

alsa9: 44100 Hz/2 channels/4 bpf/65536 bytes buffer/Signed 16 bit Little Endian
AO: [alsa9] 44100Hz 2ch Signed 16-bit (Little-Endian) (2 bps)

alsaplayer either dies with "FIXME: f_unsynchronization is set.Please contact alsaplayer team." or produces static noise.

it sounds like a different one. but, at least, the above FIXME should be the alsaplayer problem.

Quoted 2 lines

aplay does get this right, so it seems to use a different method of querying the device.

then i guess using plug layer would solve the problem in most cases.

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#6 lore raw

Am Dienstag, 28. Oktober 2003 12:10 schrieb Takashi Iwai:

Quoted 7 lines
Quoted 4 lines

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms,
...

did you use plughw instead of hw in all cases?
otherwise they won't work always.

plughw doesn't work at all!

Quoted 8 lines

mplayer has an option to specify the device name. it'd be better to define a new pcm in ~/.asoundrc such as

	pcm.emi26 {
		type plug
		slave.pcm "hw:0"
	}
Quoted 6 lines

(where "hw:0" should be changed to the corresponding one)

and run like

	% mplayer -ao alsa9:emi26 ...

gives:
Checking audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
AF_pre: af format: 2 bps, 2 ch, 44100 hz, big endian signed int
AF_pre: 44100Hz 2ch Signed 16-bit (Big-Endian)
alsa-init: requested format: 44100 Hz, 2 channels, Signed 16-bit (Big-Endian)
alsa-init: soundcard set to emi26
alsa9: 44100 Hz/2 channels/2 bpf/32768 bytes buffer/Signed 16 bit Big Endian
AO: [alsa9] 44100Hz 2ch Signed 16-bit (Big-Endian) (2 bps) Building audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
Video: no video
Starting playback...

MPlayer interrupted by signal 11 in module: play_audio

and no sound, whereas accessing hw:1 directly at least produces output. playing through plug-hw with aplay works, though
alsaplayer doesn't like the card with plug-layer. (alsaplayer seems to have deeper problems, since its doesn't work on the internal snapper, as well.

(Oh: kernel 2.6-test8-benh from bitkeeper, if that is of any interest)

Have fun*

Niklas

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#7 lore raw

At Tue, 28 Oct 2003 12:48:17 +0100,

Quoted 10 lines · Niklas Werner

Am Dienstag, 28. Oktober 2003 12:10 schrieb Takashi Iwai:

Quoted 7 lines
Quoted 4 lines

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms,
...

did you use plughw instead of hw in all cases?
otherwise they won't work always.

plughw doesn't work at all!

for xmms, too?

Quoted 29 lines
Quoted 8 lines

mplayer has an option to specify the device name. it'd be better to define a new pcm in ~/.asoundrc such as

	pcm.emi26 {
		type plug
		slave.pcm "hw:0"
	}
Quoted 6 lines

(where "hw:0" should be changed to the corresponding one)

and run like

	% mplayer -ao alsa9:emi26 ...

gives:
Checking audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
AF_pre: af format: 2 bps, 2 ch, 44100 hz, big endian signed int
AF_pre: 44100Hz 2ch Signed 16-bit (Big-Endian)
alsa-init: requested format: 44100 Hz, 2 channels, Signed 16-bit (Big-Endian)
alsa-init: soundcard set to emi26
alsa9: 44100 Hz/2 channels/2 bpf/32768 bytes buffer/Signed 16 bit Big Endian
AO: [alsa9] 44100Hz 2ch Signed 16-bit (Big-Endian) (2 bps) Building audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
Video: no video
Starting playback...

Quoted 1 line

MPlayer interrupted by signal 11 in module: play_audio

it's segfault. likely a bug in the mplayer's alsa9 output routine.

to be sure, try to start mplayer on gdb and check what is broken. tracing via strace woule help, too.

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#8 lore raw

Am Dienstag, 28. Oktober 2003 13:13 wurde geschrieben:

Quoted 8 lines

At Tue, 28 Oct 2003 12:48:17 +0100,

Quoted 5 lines · Niklas Werner

Am Dienstag, 28. Oktober 2003 12:10 schrieb Takashi Iwai:

Quoted 4 lines
Quoted 3 lines

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms, ...

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

did you use plughw instead of hw in all cases?
otherwise they won't work always.

plughw doesn't work at all!

for xmms, too?

oops, sorry, forgot. xmms works! except for adjusting the PCM-Volume, such is life... (this works on Intel, btw)
Is there any special trick regarding the ctl.* for plughw? alsa-xmms is probably the newest one, I am using gentoo... i tried using diefferent mixer-card settings, but none did work.

Quoted 3 lines

to be sure, try to start mplayer on gdb and check what is broken. tracing via strace woule help, too.

gdb:

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 28218)]
0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2 (gdb) bt

#0  0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#1  0x0fd47b00 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#2  0x0fd4643c in snd_pcm_plugin_write_areas () from /usr/lib/
libasound.so.2
#3  0x0fd3ce54 in snd_pcm_write_areas () from /usr/lib/libasound.so.2
#4  0x0fd467bc in snd_pcm_plugin_writei () from /usr/lib/libasound.so.2
#5  0x0fd3615c in snd_pcm_writei () from /usr/lib/libasound.so.2
#6  0x10078718 in outputaudio ()
#7  0x1007837c in outputaudio ()
#8  0x1002ed74 in main ()

strace:

see attached file.

have fun*

Niklas

#9 lore raw

At Tue, 28 Oct 2003 13:42:15 +0100,

Quoted 11 lines · Niklas Werner

[1 <text/plain; us-ascii (7bit)>]
Am Dienstag, 28. Oktober 2003 13:13 wurde geschrieben:

Quoted 8 lines

At Tue, 28 Oct 2003 12:48:17 +0100,

Quoted 5 lines · Niklas Werner

Am Dienstag, 28. Oktober 2003 12:10 schrieb Takashi Iwai:

Quoted 4 lines
Quoted 3 lines

No, I don't think it is.
I get similar problems with my emi 2|6 and alsaplayer, mplayer, xmms, ...

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

did you use plughw instead of hw in all cases?
otherwise they won't work always.

plughw doesn't work at all!

for xmms, too?

oops, sorry, forgot. xmms works! except for adjusting the PCM-Volume, such is life... (this works on Intel, btw)

i don't know your emi26 device has PCM volume.
if not, assign another existing control in the config of alsa-xmms plugin.

Quoted 1 line

Is there any special trick regarding the ctl.* for plughw?

it has nothing to do with plughw abstraction.
it's the difference of hardware.

Quoted 11 lines

alsa-xmms is probably the newest one, I am using gentoo... i tried using diefferent mixer-card settings, but none did work.

Quoted 3 lines

to be sure, try to start mplayer on gdb and check what is broken. tracing via strace woule help, too.

gdb:

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 28218)]
0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2 (gdb) bt

Quoted 10 lines
#0  0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#1  0x0fd47b00 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#2  0x0fd4643c in snd_pcm_plugin_write_areas () from /usr/lib/
libasound.so.2
#3  0x0fd3ce54 in snd_pcm_write_areas () from /usr/lib/libasound.so.2
#4  0x0fd467bc in snd_pcm_plugin_writei () from /usr/lib/libasound.so.2
#5  0x0fd3615c in snd_pcm_writei () from /usr/lib/libasound.so.2
#6  0x10078718 in outputaudio ()
#7  0x1007837c in outputaudio ()
#8  0x1002ed74 in main ()

hmm, then something wrong in the converter routine... needs to take a deeper look.

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: The SF.net Donation Program. Do you like what SourceForge.net is doing for the Open Source Community? Make a contribution, and help us add new features and functionality. Click here: http://sourceforge.net/donate/

#10 lore raw

At Tue, 28 Oct 2003 14:28:57 +0100,

Quoted 18 lines · I wrote:
Quoted 7 lines

gdb:

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 28218)]
0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2 (gdb) bt

Quoted 10 lines
#0  0x0fd47d64 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#1  0x0fd47b00 in snd_pcm_linear_convert () from /usr/lib/libasound.so.2
#2  0x0fd4643c in snd_pcm_plugin_write_areas () from /usr/lib/
libasound.so.2
#3  0x0fd3ce54 in snd_pcm_write_areas () from /usr/lib/libasound.so.2
#4  0x0fd467bc in snd_pcm_plugin_writei () from /usr/lib/libasound.so.2
#5  0x0fd3615c in snd_pcm_writei () from /usr/lib/libasound.so.2
#6  0x10078718 in outputaudio ()
#7  0x1007837c in outputaudio ()
#8  0x1002ed74 in main ()
Quoted 2 lines

hmm, then something wrong in the converter routine... needs to take a deeper look.

i found a bug regarding the plugin but it must be another bug from the above problem.

segfault is a bit puzzling. could you try the attached patch to see the parameters? or, even better, check the values in gdb's backtrace.

also, what happens if running mplayer with mmap option, i.e.

% mplayer -ao alsa9:emi26:mmap

?

Takashi

#11 lore raw

Here we go again,

Am Dienstag, 28. Oktober 2003 19:25 schrieb Takashi Iwai:

Quoted 9 lines
Quoted 2 lines

hmm, then something wrong in the converter routine... needs to take a deeper look.

i found a bug regarding the plugin but it must be another bug from the above problem.

segfault is a bit puzzling. could you try the attached patch to see the parameters? or, even better, check the values in gdb's backtrace.

Building audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
Video: no video
Starting playback...
XXX linear_convert: dst=0x104d3dc0/0, src=0x7fffc020/0, ch=2, frames=16384, idx=35

===========
gdb:

Starting playback...
XXX linear_convert: dst=0x104d1f18/0, src=0x7fffcf40/0, ch=2,
frames=16384, idx= 35

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 14666)] snd_pcm_linear_convert (dst_areas=0x104d1f18, dst_offset=0,

    src_areas=0x7fffcf40, src_offset=0, channels=2, frames=16384, 
convidx=35)
    at plugin_ops.h:299

299 conv_1234_xx21: as_u16(dst) = bswap_16(as_u32c(src) >> 16); goto
CONV_EN D;
(gdb) bt
#0 snd_pcm_linear_convert (dst_areas=0x104d1f18, dst_offset=0,

    src_areas=0x7fffcf40, src_offset=0, channels=2, frames=16384, 
convidx=35)
    at plugin_ops.h:299
#1  0x0fd4326c in snd_pcm_linear_convert (dst_areas=0x104d1f18, 
dst_offset=0, 
    src_areas=0x7fffcf40, src_offset=0, channels=2, frames=16384, 
convidx=35)
    at pcm_linear.c:170
#2  0x0fd41838 in snd_pcm_plugin_write_areas (pcm=0x0, areas=0x7fffcf40, 
    offset=2, size=2147471168) at pcm_plugin.c:365
#3  0x0fd37dc4 in snd_pcm_write_areas (pcm=0x23, areas=0x104db86c, 
offset=2, 
    size=2147471168, func=0) at pcm.c:6206
#4  0x0fd41c44 in snd_pcm_plugin_writei (pcm=0x104d1f18, buffer=0x4, 
    size=2147471168) at pcm_plugin.c:436
#5  0x0fd30c6c in snd_pcm_writei (pcm=0x0, buffer=0x0, size=0)
    at pcm_local.h:368
#6  0x10078718 in outputaudio ()
#7  0x1007837c in outputaudio ()
#8  0x1002ed74 in main ()

==================================

Quoted 6 lines

also, what happens if running mplayer with mmap option, i.e.

% mplayer -ao alsa9:emi26:mmap

horrible, distorted sound: (sounds like computer is too slow to deliver the samples)
ALSA lib pcm_hw.c:494:(snd_pcm_hw_start) SNDRV_PCM_IOCTL_START failed: Die Dateizugriffsnummer ist ein schlechter Verfassung

meaning: "File descriptor in bad state" (whoever sneaked in this translation...)

Have fun

Niklas

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/

#12 lore raw

At Tue, 28 Oct 2003 20:18:35 +0100,

Quoted 28 lines · Niklas Werner

Here we go again,

Am Dienstag, 28. Oktober 2003 19:25 schrieb Takashi Iwai:

Quoted 9 lines
Quoted 2 lines

hmm, then something wrong in the converter routine... needs to take a deeper look.

i found a bug regarding the plugin but it must be another bug from the above problem.

segfault is a bit puzzling. could you try the attached patch to see the parameters? or, even better, check the values in gdb's backtrace.

Building audio filter chain for 44100Hz/2ch/16bit -> 44100Hz/2ch/16bit...
Video: no video
Starting playback...
XXX linear_convert: dst=0x104d3dc0/0, src=0x7fffc020/0, ch=2, frames=16384, idx=35

===========
gdb:

Starting playback...
XXX linear_convert: dst=0x104d1f18/0, src=0x7fffcf40/0, ch=2,
frames=16384, idx= 35

Quoted 8 lines

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 14666)] snd_pcm_linear_convert (dst_areas=0x104d1f18, dst_offset=0,

    src_areas=0x7fffcf40, src_offset=0, channels=2, frames=16384, 
convidx=35)
    at plugin_ops.h:299

299 conv_1234_xx21: as_u16(dst) = bswap_16(as_u32c(src) >> 16); goto
CONV_EN D;

hmm, it seems that a wrong label is used. the label should be conv_xx12_xx21 (= conv_labels[35]). something is really broken.

could you check stepwise the loop there?

Quoted 6 lines
Quoted 6 lines

also, what happens if running mplayer with mmap option, i.e.

% mplayer -ao alsa9:emi26:mmap

Quoted 7 lines

horrible, distorted sound: (sounds like computer is too slow to deliver the samples)
ALSA lib pcm_hw.c:494:(snd_pcm_hw_start) SNDRV_PCM_IOCTL_START failed: Die Dateizugriffsnummer ist ein schlechter Verfassung

meaning: "File descriptor in bad state" (whoever sneaked in this translation...)

(well, good to learn german, at least ;)

thanks for check.
can you hear a clear sound with aplay -M ?

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/

#13 lore raw

Am Mittwoch, 29. Oktober 2003 19:24 wurde geschrieben:

Quoted 8 lines

At Tue, 28 Oct 2003 20:18:35 +0100,

hmm, it seems that a wrong label is used. the label should be conv_xx12_xx21 (= conv_labels[35]). something is really broken.

could you check stepwise the loop there?

still checking (any tips on speeding this up (I'm trying "step 80000" atm?), seems to happen after the first set of frames (=16384) is processed.
the label stays wrong throughout the whole loop, the convidx stays 35

Quoted 3 lines

thanks for check.
can you hear a clear sound with aplay -M ?

Yes :-)

Niklas

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/

#14 lore raw

At Wed, 29 Oct 2003 23:26:07 +0100,

Quoted 8 lines · Niklas Werner

Am Mittwoch, 29. Oktober 2003 19:24 wurde geschrieben:

Quoted 6 lines

At Tue, 28 Oct 2003 20:18:35 +0100,

hmm, it seems that a wrong label is used. the label should be conv_xx12_xx21 (= conv_labels[35]). something is really broken.

Quoted 6 lines
Quoted 2 lines

could you check stepwise the loop there?

still checking (any tips on speeding this up (I'm trying "step 80000" atm?), seems to happen after the first set of frames (=16384) is processed.

in the first process, did it go to conv_xx12_xx21 properly?

Quoted 7 lines

the label stays wrong throughout the whole loop, the convidx stays 35

Quoted 3 lines

thanks for check.
can you hear a clear sound with aplay -M ?

Yes :-)

at least, something still good :)

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/

#15 lore raw

Am Donnerstag, 30. Oktober 2003 13:17 schrieb Takashi Iwai:

Quoted 10 lines

At Wed, 29 Oct 2003 23:26:07 +0100,

Quoted 7 lines · Niklas Werner

Am Mittwoch, 29. Oktober 2003 19:24 wurde geschrieben:

Quoted 6 lines

At Tue, 28 Oct 2003 20:18:35 +0100,

hmm, it seems that a wrong label is used. the label should be conv_xx12_xx21 (= conv_labels[35]). something is really broken.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

could you check stepwise the loop there?

still checking (any tips on speeding this up (I'm trying "step 80000" atm?), seems to happen after the first set of frames (=16384) is processed.

in the first process, did it go to conv_xx12_xx21 properly?

nope:

0x0fd422f4 184 goto *conv;
(gdb) si
299 conv_1234_xx21: as_u16(dst) = bswap_16(as_u32c(src) >> 16); goto
CONV_END;
(gdb) si
0x0fd424c0 299 conv_1234_xx21: as_u16(dst) = bswap_16
(as_u32c(src) >> 16); goto CONV_END;
(gdb) si
0x0fd424c4 299 conv_1234_xx21: as_u16(dst) = bswap_16
(as_u32c(src) >> 16); goto CONV_END;
(gdb) si
0x0fd424c8 299 conv_1234_xx21: as_u16(dst) = bswap_16
(as_u32c(src) >> 16); goto CONV_END;
(gdb) si
0x0fd424cc 299 conv_1234_xx21: as_u16(dst) = bswap_16
(as_u32c(src) >> 16); goto CONV_END;
(gdb) si
0x0fd4235c 230 conv_xxx1_xx10: as_u16(dst) =
(u_int16_t)as_u8c(src) << 8; goto CONV_END;
(gdb) si
0x0fd42360 230 conv_xxx1_xx10: as_u16(dst) =
(u_int16_t)as_u8c(src) << 8; goto CONV_END;
(gdb) si
190 dst += dst_step;

Niklas

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/

#16 lore raw

At Thu, 30 Oct 2003 13:42:14 +0100,

Quoted 18 lines · Niklas Werner

Am Donnerstag, 30. Oktober 2003 13:17 schrieb Takashi Iwai:

Quoted 16 lines

At Wed, 29 Oct 2003 23:26:07 +0100,

Quoted 12 lines · Niklas Werner

Am Mittwoch, 29. Oktober 2003 19:24 wurde geschrieben:

Quoted 7 lines

At Tue, 28 Oct 2003 20:18:35 +0100,

hmm, it seems that a wrong label is used. the label should be conv_xx12_xx21 (= conv_labels[35]). something is really broken.

could you check stepwise the loop there?

still checking (any tips on speeding this up (I'm trying "step 80000" atm?), seems to happen after the first set of frames (=16384) is processed.

Quoted 7 lines
Quoted 1 line

in the first process, did it go to conv_xx12_xx21 properly?

nope:

0x0fd422f4 184 goto *conv;
(gdb) si
299 conv_1234_xx21: as_u16(dst) = bswap_16(as_u32c(src) >> 16); goto

hmm, really weird.

meanwhile, i rewrote snd_pcm_linear_convert() without goto trick. could you try the attached patch?

Takashi

#17 lore raw

Am Donnerstag, 30. Oktober 2003 20:46 wurde geschrieben:

Quoted 7 lines

At Thu, 30 Oct 2003 13:42:14 +0100,

hmm, really weird.

meanwhile, i rewrote snd_pcm_linear_convert() without goto trick. could you try the attached patch?

... you don't really want to know...

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 22147)] 0x0fd41564 in snd_pcm_linear_conv_xx12_xx21 (

    src=0x104e4000 <Address 0x104e4000 out of bounds>, dst=0x30c28590 "", 
    src_step=4, dst_step=4, frames=7835) at pcm_linear.c:303
303     DEF_CONV(conv_xx12_xx21,l_conv_xx12_xx21);
(gdb) bt
#0  0x0fd41564 in snd_pcm_linear_conv_xx12_xx21 (
    src=0x104e4000 <Address 0x104e4000 out of bounds>, dst=0x30c28590 "", 
    src_step=4, dst_step=4, frames=7835) at pcm_linear.c:303
#1  0x0fd429f4 in snd_pcm_linear_convert (dst_areas=0x104d1f18, 
dst_offset=0, 
    src_areas=0x7fffcf10, src_offset=0, channels=2, frames=16384, 
convidx=0)
    at pcm_local.h:362
#2  0x0fd435a0 in snd_pcm_linear_write_areas (pcm=0x0, areas=0x1, 
    offset=2147471120, size=12, slave_areas=0x0, slave_offset=1140885572, 
    slave_sizep=0xfdb408c) at pcm_linear.c:722
#3  0x0fd3f808 in snd_pcm_plugin_write_areas (pcm=0x2, areas=0x7fffcf10, 
    offset=0, size=273489688) at pcm_plugin.c:365
#4  0x0fd35d94 in snd_pcm_write_areas (pcm=0x104db890, areas=0xfd41554, 
    offset=0, size=273489688, func=0) at pcm.c:6206
#5  0x0fd3fc14 in snd_pcm_plugin_writei (pcm=0x104db600, 
buffer=0x30c28590, 
    size=273489688) at pcm_plugin.c:436
#6  0x0fd2ec3c in snd_pcm_writei (pcm=0x0, buffer=0x0, size=0)
    at pcm_local.h:368
#7  0x10078718 in outputaudio ()
#8  0x1007837c in outputaudio ()
#9  0x1002ed74 in main ()

Niklas

Signature · 5 lines
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?   SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
#18 lore raw

At Fri, 31 Oct 2003 09:23:02 +0100,

Quoted 10 lines · Niklas Werner

Am Donnerstag, 30. Oktober 2003 20:46 wurde geschrieben:

Quoted 7 lines

At Thu, 30 Oct 2003 13:42:14 +0100,

hmm, really weird.

meanwhile, i rewrote snd_pcm_linear_convert() without goto trick. could you try the attached patch?

... you don't really want to know...

yes, i do want to know that :)
at least, it spots out the real segfault place.

Quoted 6 lines

Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 22147)] 0x0fd41564 in snd_pcm_linear_conv_xx12_xx21 (

    src=0x104e4000 <Address 0x104e4000 out of bounds>, dst=0x30c28590 "", 
    src_step=4, dst_step=4, frames=7835) at pcm_linear.c:303

303 DEF_CONV(conv_xx12_xx21,l_conv_xx12_xx21);

that's it: the source address is bogus.
i should have asked you to check that...

Takashi

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/