Skip to contentSkip to conversation

AV out of sync with "-ao alsa" in gmplayer, mplayer works ok]

Fwd: Re: Lee Revell 4 participants 7 messages Dec 29, 2004 – Jan 19, 2005

lore
Share & tools
lore

Lee Revell

author
#0 lore raw

OK, I trolled the mplayer dev list, and got this response. ;-)

Are Reimar's statements about memory leaks valid?

Lee

-------- Forwarded Message --------
From: Reimar Döffinger <[email protected]>
Reply-To: [email protected]
To: [email protected]
Subject: Re: [MPlayer-dev-eng] AV out of sync with "-ao alsa" in gmplayer, mplayer works ok
Date: Wed, 29 Dec 2004 09:41:01 +0100
Hi,

Quoted 3 lines · Lee Revell

Using 1.0pre6, "gmplayer -ao alsa" results in the sound and video being out of sync. "mplayer -ao alsa" works fine, as does "-ao oss" with either mplayer or gmplayer.

Are you sure this isn't caused either by different settings in gmplayer or because the computer is too slow? Please provide a full -v log.

Quoted 7 lines

This is quite annoying, as I was expecting to FINALLY have a gmplayer release usable with ALSA. Previous versions were unusable due to that stupid "/dev/mixer" bug.

What's the deal? Why is it so hard for the mplayer developers to get ALSA support right? I hack on ALSA a lot, so if there is some problem with the ALSA API, we would like to hear about it on the mailing list.

Yes, please fix the huge memory leak with ALSA output, especially when changing the volume (it is cause by the call to snd_mixer_load, but I couldn't find an ALSA function that frees that memory again snd_mixer_free has absolutely no effect :-( ).
Also the problem is more that the Gui is not well supported.

Quoted 1 line

Or do you all just use OSS still?

The OSS interface is actually a lot nicer, so yes it is supported better.

Quoted 4 lines

From the users perspective things like this make ALSA look bad, while the problem is really with gmplayer's shoddy ALSA support.

From what I saw of ALSA as a developer it is bad, but that's a bit of

personal opinion of course. What I refer to is very slow software mixing, no way to use it without getting into synchronization hell because it uses threads,
hundreds of function calls to initialize it, and those memory leaks that
to my best knowledge seem to come from the alsa lib (my best guess is
that this is because like in conf.c, in _snd_config_make pointers
are overwritten without being freed first).
What makes it a bit difficult to use it that it's quite difficult to
find the function that is supposed to free things again, they e.g.
often aren't located directly below the allocation function in the
headers, and that there is a snd_mixer_load function but no snd_mixer_unload as I would expect it.
All these things together made me give up on ALSA.

Greetings,
Reimar Döffinger

_______________________________________________
MPlayer-dev-eng mailing list
[email protected]
http://mplayerhq.hu/mailman/listinfo/mplayer-dev-eng

-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now.
http://productguide.itmanagersjournal.com/

#1 lore raw
Quoted 3 lines · Lee Revell

OK, I trolled the mplayer dev list, and got this response. ;-)

Are Reimar's statements about memory leaks valid?

No. I regularly run my player under valgrind, and it reports no ALSA related leaks. I also run the same player for weeks on end, and the memory usage stays below 4MB, after opening, playing a bit, and closing the sound device thousands of times.

Quoted 13 lines

Hi,

Quoted 3 lines · Lee Revell

Using 1.0pre6, "gmplayer -ao alsa" results in the sound and video being out of sync. "mplayer -ao alsa" works fine, as does "-ao oss" with either mplayer or gmplayer.

Are you sure this isn't caused either by different settings in gmplayer or because the computer is too slow? Please provide a full -v log.

Quoted 4 lines

This is quite annoying, as I was expecting to FINALLY have a gmplayer release usable with ALSA. Previous versions were unusable due to that stupid "/dev/mixer" bug.

Quoted 5 lines
Quoted 3 lines

What's the deal? Why is it so hard for the mplayer developers to get ALSA support right? I hack on ALSA a lot, so if there is some problem with the ALSA API, we would like to hear about it on the mailing list.

Yes, please fix the huge memory leak with ALSA output,

Never seen that one. Could be a leak in mplayer's ALSA support, of course.

Quoted 8 lines

especially when changing the volume (it is cause by the call to snd_mixer_load, but I couldn't find an ALSA function that frees that memory again snd_mixer_free has absolutely no effect :-( ). Also the problem is more that the Gui is not well supported.

Quoted 1 line

Or do you all just use OSS still?

The OSS interface is actually a lot nicer, so yes it is supported better.

LOL

Quoted 5 lines
Quoted 4 lines

From the users perspective things like this make ALSA look bad, while the problem is really with gmplayer's shoddy ALSA support.

From what I saw of ALSA as a developer it is bad, but that's a bit of

personal opinion of course. What I refer to is very slow software

Who said what here?

Quoted 2 lines

mixing, no way to use it without getting into synchronization hell because it uses threads,

Oh, never forget that the mplayer folks hate threads.

Quoted 9 lines

hundreds of function calls to initialize it, and those memory leaks that to my best knowledge seem to come from the alsa lib (my best guess is that this is because like in conf.c, in _snd_config_make pointers are overwritten without being freed first). What makes it a bit difficult to use it that it's quite difficult to find the function that is supposed to free things again, they e.g. often aren't located directly below the allocation function in the headers, and that there is a snd_mixer_load function but no snd_mixer_unload as I would expect it.

OK, the documentation could be better.

Quoted 1 line

All these things together made me give up on ALSA.

I'll skip the insult this time.

Signature · 10 lines
Måns Rullgård
[email protected]



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
author
#2 lore raw
Quoted 3 lines · Måns Rullgård
Quoted 3 lines
Quoted 2 lines

From the users perspective things like this make ALSA look bad, while the problem is really with gmplayer's shoddy ALSA support.

Me

Quoted 3 lines
Quoted 2 lines
Quoted 1 line

From what I saw of ALSA as a developer it is bad, but that's a bit of

personal opinion of course. What I refer to is very slow software

Reimar

Quoted 2 lines

Who said what here?

Anyway, sorry for my previous tone, I would really like to improve the situation, by helping the mplayer developers get good alsa support.

Lee

-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now.
http://productguide.itmanagersjournal.com/

#3 lore raw
In reply to Måns Rullgård

Hi,

Quoted 9 lines · Måns Rullgård <mru <at> inprovide.com> writes:
Quoted 3 lines · Lee Revell <rlrevell <at> joe-job.com> writes:

OK, I trolled the mplayer dev list, and got this response.

Are Reimar's statements about memory leaks valid?

No. I regularly run my player under valgrind, and it reports no ALSA related leaks. I also run the same player for weeks on end, and the memory usage stays below 4MB, after opening, playing a bit, and closing the sound device thousands of times.

And how about opening and closing the mixer thousands of times? I posted a stripped-down sample code that ahould show the problem a few days ago...

Quoted 1 line
Quoted 1 line · Lee Revell

[...]

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

What's the deal? Why is it so hard for the mplayer developers to get ALSA support right? I hack on ALSA a lot, so if there is some problem with the ALSA API, we would like to hear about it on the mailing list.

Yes, please fix the huge memory leak with ALSA output,

Never seen that one. Could be a leak in mplayer's ALSA support, of course.

Well, I didn't say that it was in the ALSA lib. I only said I think so ;-)

Quoted 10 lines
Quoted 8 lines

especially when changing the volume (it is cause by the call to snd_mixer_load, but I couldn't find an ALSA function that frees that memory again snd_mixer_free has absolutely no effect ). Also the problem is more that the Gui is not well supported.

Quoted 1 line

Or do you all just use OSS still?

The OSS interface is actually a lot nicer, so yes it is supported better.

LOL

Well, maybe it's just because whoever wrote ao_alsa in MPlayer did it the wrong way, but compare the size of ao_alsa.c and ao_oss.c and you'll probably see what I meant...

[...]

Quoted 4 lines
Quoted 2 lines

mixing, no way to use it without getting into synchronization hell because it uses threads,

Oh, never forget that the mplayer folks hate threads.

Yep, and it's not entirely unjustified...

Quoted 11 lines
Quoted 9 lines

hundreds of function calls to initialize it, and those memory leaks that to my best knowledge seem to come from the alsa lib (my best guess is that this is because like in conf.c, in _snd_config_make pointers are overwritten without being freed first). What makes it a bit difficult to use it that it's quite difficult to find the function that is supposed to free things again, they e.g. often aren't located directly below the allocation function in the headers, and that there is a snd_mixer_load function but no snd_mixer_unload as I would expect it.

OK, the documentation could be better.

Well, that's true for most projects, certainly including MPlayer ;-) I have to admit I only looked at the headers, I didn't look for/download the documentation, but I'm not the maintainer of that code anyway...

Quoted 3 lines
Quoted 1 line

All these things together made me give up on ALSA.

I'll skip the insult this time.

Thanks :-)

Greetings,
Reimar Döffinger

-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt

author
#4 lore raw
Quoted 10 lines · Reimar Döffinger
Quoted 10 lines
Quoted 9 lines

hundreds of function calls to initialize it, and those memory leaks that to my best knowledge seem to come from the alsa lib (my best guess is that this is because like in conf.c, in _snd_config_make pointers are overwritten without being freed first). What makes it a bit difficult to use it that it's quite difficult to find the function that is supposed to free things again, they e.g. often aren't located directly below the allocation function in the headers, and that there is a snd_mixer_load function but no snd_mixer_unload as I would expect it.

Quoted 6 lines
Quoted 1 line

OK, the documentation could be better.

Well, that's true for most projects, certainly including MPlayer ;-) I have to admit I only looked at the headers, I didn't look for/download the documentation, but I'm not the maintainer of that code anyway...

Are there any plans to fix the ALSA output so that it doesn't reset the stream with every xrun? Since you don't use a realtime thread to handle the audio there are certain to be some xruns, for example whenever using the GUI controls in any way or moving the mouse.

http://sourceforge.net/mailarchive/message.php?msg_id=8437146

I am still forced to use the OSS emulation for this reason.

Lee

-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt

#5 lore raw

Hi,

Quoted 4 lines · Lee Revell

Are there any plans to fix the ALSA output so that it doesn't reset the stream with every xrun? Since you don't use a realtime thread to handle the audio there are certain to be some xruns, for example whenever using the GUI controls in any way or moving the mouse.

xruns just should not happen. Ignoring them seems to be the very wrong way. Also a real-time thread should not be necessary just to avoid xruns when only moving the mouse IMHO.
Maybe you can just increase the buffer size somehow? At least for MPlayer a big buffer should cause no problems.

Greetings,
Reimar Döffinger

P.S.: I wont cc further mails about this to the list, because I don't think this belongs on alsa-devel.

-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt

#6 lore raw

At Tue, 18 Jan 2005 21:23:09 +0100,

Quoted 11 lines · Reimar Döffinger

Hi,

Quoted 4 lines · Lee Revell

Are there any plans to fix the ALSA output so that it doesn't reset the stream with every xrun? Since you don't use a realtime thread to handle the audio there are certain to be some xruns, for example whenever using the GUI controls in any way or moving the mouse.

xruns just should not happen. Ignoring them seems to be the very wrong way. Also a real-time thread should not be necessary just to avoid xruns when only moving the mouse IMHO.

No, it's not guaranteed at all.

Quoted 2 lines

Maybe you can just increase the buffer size somehow? At least for MPlayer a big buffer should cause no problems.

The hardware buffer size is limited. Suppose the hardware which can have the max buffer size 64kB. When you play a DVD with 5.1 output, it corresponds to 0.11 second.

The dedicated audio thread works better than the single thread model because of the following reasons (maybe more):

- It has a large intermediate buffer. The buffer size can be as large as you like.
- It runs only in short time and sleeps.
This is the preferred behavior for the kernel scheduler. So, this thread may receive higher priority bonus.
- You can change the priority of only audio thread but keep others.

Takashi

-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt