0002368: (master) sound stuttering/choppy in current spring master

(0006434)
jK (developer)
2011-02-25 22:22

https://github.com/spring/spring/commit/7b22a0edf78d41dd3558defa5aa3b1796b8cc53f#comments [^] ?

(0006439)
zerver (developer)
2011-02-26 18:03

Could you try with /dynamicsun

(0006440)
abma (administrator)
2011-02-26 18:29
edited on: 2011-02-26 18:45

sorry, this was bit to unspecific.

with /dynamicsun 0 drawworld is again at 7percent but the stuttering is still there. in mt i get no sound-stutters.

if debugmode (alt-b) is on or off makes no difference, i alwayws get sound-stutters in singlethread-spring.

tested with 0.82.3-1446-g4adf86d

i tried also with EFX=1 & EFX=0

maybe the inflog.txt helps?

(real) cpu is at 100percent, but i have to play with VSync=0 (ATI-Bug)
cpu in springs debug mode shown is at average ~5percent or so...

even in pause mode sound stutters...

i'll bisect

(0006441)
abma (administrator)
2011-02-26 18:57
edited on: 2011-02-26 19:01

66de8f60e863befa61e5addaea539877dd78d064 is the last working version (the others don't compile)
so, it looks like the efx-commits creates the problem :-/

$ git bisect log
git bisect start
# bad: [4adf86d0e239a3d6196a8f4bc69fc35b93e5df11] fix inverted logic for track-deletion
git bisect bad 4adf86d0e239a3d6196a8f4bc69fc35b93e5df11
# good: [a8709bf8e7032f0a67cfeb68f1014f5fb968380e] Merge branch 'master' of git@github.com:spring/spring
git bisect good a8709bf8e7032f0a67cfeb68f1014f5fb968380e
# good: [8c869ab098ed0fb54304d4902230200a4946d8c7] fix (rare) teleportation of aircraft built on raised platforms
git bisect good 8c869ab098ed0fb54304d4902230200a4946d8c7
# bad: [39d0eb4a68824cb35052d797e9d679c2105abf85] 3rd try
git bisect bad 39d0eb4a68824cb35052d797e9d679c2105abf85
# good: [66de8f60e863befa61e5addaea539877dd78d064] modularized Watchdog & added Linux implementation * it can now handle any number of threads * it gives threads names - currently "main" (:= render-thread in spring-mt) & "sim" (only in spring-mt) * syntax of the Lua function changed too: Spring.ClearWatchDogTimer() --clears current thread's timer Spring.ClearWatchDogTimer("sim") --clears the timer of the "sim" thread
git bisect good 66de8f60e863befa61e5addaea539877dd78d064

(0006482)
abma (administrator)
2011-03-10 21:20

hm, maybe sounds are only played twice with a tiny time difference?

(0006483)
zerver (developer)
2011-03-11 18:30

could be, that has happened in the past too

(0006563)
abma (administrator)
2011-04-24 04:10

@jk:

output of pulseaudio -v -v:

http://pastebin.com/FiLThj1q [^]

the stuttering seems to come from buffer-underruns:
D: protocol-native.c: Underrun on 'Playback Stream', 0 bytes in queue.
D: protocol-native.c: Requesting rewind due to end of underrun.

(0006564)
abma (administrator)
2011-04-24 04:15

when changing
/etc/openal/alsoft.conf:
drivers = pulse (ubuntu default)
to
drivers = alsa,pulse

stuttering seems to be gone

(0006565)
jK (developer)
2011-04-24 04:15
edited on: 2011-04-24 04:16

don't use pulseaudio, use alsa directly?

EDIT: this explains why I didn't encountered this issue.

(0006566)
hoijui (administrator)
2011-04-24 18:32

hmm.. ok... should we consider this a sufficient fix?
it would be nice if not 95% of our linux playerbase would have to do that on next release. can we prevent the buffer-under-runs from within spring? i would guess yes, as it does not happen in release version... so the question is, should we try to do it:
* how much time would that require and is the time worth it?
* would the "fix" likely be ok, or rather an ugly hack?

(0006567)
abma (administrator)
2011-04-25 13:16
edited on: 2011-04-25 13:34

when adding this line to /etc/openal/alsoft.conf (with drivers = pulse)

excludefx = eaxreverb

stuttering is also gone + log spam of [Sound] CSound::UpdateListener: Invalid Name

in my .springrc is a "UseEFX=0"

(0006633)
hoijui (administrator)
2011-05-09 23:22

as discussed in meeting of 9. May 2011:

problem:
when EFX is enabled, pulseaudio is too slow in passing on packets to ALSA, and thus creates sound-lag

solution:
when pulseaudio is detected (linux only :D ), print a FAT warning like for low tech GFX cards, and disable EFX; do so only for runtime, not in the config, so it gets auto-enabled for the user when he un-uses pulseaudio.

(0006642)
hoijui (administrator)
2011-05-11 10:16

i can not see anything about pulseaudio in my infolog, only ALSA, and i think it never was there... so the detection mechanism will be more complicated to implement.
..then again.. i have UseEFX=0, and i still get the stuttering. :/
in other words.. this solution is none :/

(0006643)
abma (administrator)
2011-05-11 11:04

is LogSubsystems=Sound enabled?

you should see something like this:
[f=0000000] [Sound] Device: ALSA Software
[f=0000000] [Sound] Available Devices:
[f=0000000] [Sound] ALSA Software
[f=0000000] [Sound] PulseAudio Software

(0006644)
abma (administrator)
2011-05-11 11:19
edited on: 2011-05-11 11:20

imo UseEFX=0 doesn't work... or why does "excludefx = eaxreverb" change behaviour with UseEFX=0?

(0006650)
hoijui (administrator)
2011-05-12 00:03
edited on: 2011-06-10 16:21

yes i see that, and this subsystem is enabled by default (Aus made it so when he reworked sound, and that never got changed).
yeah.. sorry, read you wrong there.
we'll have to investigate more here...

(0007594)
abma (administrator)
2011-11-13 14:56

some more people have the stuttering problems:
http://springrts.com/phpbb/viewtopic.php?uid=8601&f=11&t=27079 [^]

(0007595)
spliff (reporter)
2011-11-14 08:47

Had a discussion with an OpenAL dev on openal@freenode. Got some information that may be useful but no conclusions except disabling either eaxreverb or pulseaudio. Chat log is here: http://echelog.matzon.dk/logs/browse/openal/1321225200 [^]

(0008687)
abma (administrator)
2012-05-13 11:51

seems to be fixed in spring 88 / newer linux distributions...

Issue History
Date Modified	Username	Field	Change
2011-02-25 21:13	abma	New Issue
2011-02-25 22:22	jK	Note Added: 0006434
2011-02-26 18:03	zerver	Note Added: 0006439
2011-02-26 18:29	abma	Note Added: 0006440
2011-02-26 18:30	abma	Summary	(master) big spikes in drawworld (debug-mode/alt-b) => (master) sound stuttering in current spring master
2011-02-26 18:30	abma	File Added: infolog.txt
2011-02-26 18:32	abma	Note Edited: 0006440
2011-02-26 18:32	abma	Note Edited: 0006440
2011-02-26 18:38	abma	Note Edited: 0006440
2011-02-26 18:41	abma	Note Edited: 0006440
2011-02-26 18:45	abma	Note Edited: 0006440
2011-02-26 18:57	abma	Note Added: 0006441
2011-02-26 19:00	abma	Note Edited: 0006441
2011-02-26 19:00	abma	Note Edited: 0006441
2011-02-26 19:01	abma	Note Edited: 0006441
2011-02-26 19:02	abma	Summary	(master) sound stuttering in current spring master => (master) sound stuttering in current spring master (singlethread only)
2011-02-26 20:06	abma	Relationship added	related to 0002184
2011-03-10 21:20	abma	Note Added: 0006482
2011-03-11 18:30	zerver	Note Added: 0006483
2011-04-24 04:10	abma	Note Added: 0006563
2011-04-24 04:15	abma	Note Added: 0006564
2011-04-24 04:15	jK	Note Added: 0006565
2011-04-24 04:16	jK	Note Edited: 0006565
2011-04-24 18:32	hoijui	Note Added: 0006566
2011-04-25 13:16	abma	Note Added: 0006567
2011-04-25 13:17	abma	Note Edited: 0006567
2011-04-25 13:19	abma	Note Edited: 0006567
2011-04-25 13:19	abma	Note Edited: 0006567
2011-04-25 13:34	abma	Note Edited: 0006567
2011-05-09 23:22	hoijui	Note Added: 0006633
2011-05-11 10:16	hoijui	Note Added: 0006642
2011-05-11 11:04	abma	Note Added: 0006643
2011-05-11 11:19	abma	Note Added: 0006644
2011-05-11 11:20	abma	Note Edited: 0006644
2011-05-12 00:03	hoijui	Note Added: 0006650
2011-05-12 04:04	abma	Summary	(master) sound stuttering in current spring master (singlethread only) => (master) sound stuttering in current spring master
2011-05-12 04:04	abma	Summary	(master) sound stuttering in current spring master => (master) sound stuttering/choppy in current spring master
2011-06-10 16:21	hoijui	Note Edited: 0006650
2011-06-10 16:21	hoijui	Target Version	=> 0.83.0
2011-06-10 16:21	hoijui	Description Updated
2011-11-13 14:56	abma	Note Added: 0007594
2011-11-14 08:47	spliff	Note Added: 0007595
2011-11-23 10:45	hoijui	Product Version	=> 0.82.7+git
2011-11-23 10:45	hoijui	Build	=> 0.82.3-1440-g97d08b9
2011-11-23 10:45	hoijui	Additional Information Updated
2012-05-13 11:49	abma	Relationship added	duplicate of 0002928
2012-05-13 11:51	abma	Note Added: 0008687
2012-05-13 11:51	abma	Status	new => resolved
2012-05-13 11:51	abma	Resolution	open => fixed
2012-05-13 11:51	abma	Assigned To	=> abma

Notes
(0006434) jK (developer) 2011-02-25 22:22	https://github.com/spring/spring/commit/7b22a0edf78d41dd3558defa5aa3b1796b8cc53f#comments [^] ?

(0006439) zerver (developer) 2011-02-26 18:03	Could you try with /dynamicsun

(0006440) abma (administrator) 2011-02-26 18:29 edited on: 2011-02-26 18:45	sorry, this was bit to unspecific. with /dynamicsun 0 drawworld is again at 7percent but the stuttering is still there. in mt i get no sound-stutters. if debugmode (alt-b) is on or off makes no difference, i alwayws get sound-stutters in singlethread-spring. tested with 0.82.3-1446-g4adf86d i tried also with EFX=1 & EFX=0 maybe the inflog.txt helps? (real) cpu is at 100percent, but i have to play with VSync=0 (ATI-Bug) cpu in springs debug mode shown is at average ~5percent or so... even in pause mode sound stutters... i'll bisect

(0006441) abma (administrator) 2011-02-26 18:57 edited on: 2011-02-26 19:01	66de8f60e863befa61e5addaea539877dd78d064 is the last working version (the others don't compile) so, it looks like the efx-commits creates the problem :-/ $ git bisect log git bisect start # bad: [4adf86d0e239a3d6196a8f4bc69fc35b93e5df11] fix inverted logic for track-deletion git bisect bad 4adf86d0e239a3d6196a8f4bc69fc35b93e5df11 # good: [a8709bf8e7032f0a67cfeb68f1014f5fb968380e] Merge branch 'master' of git@github.com:spring/spring git bisect good a8709bf8e7032f0a67cfeb68f1014f5fb968380e # good: [8c869ab098ed0fb54304d4902230200a4946d8c7] fix (rare) teleportation of aircraft built on raised platforms git bisect good 8c869ab098ed0fb54304d4902230200a4946d8c7 # bad: [39d0eb4a68824cb35052d797e9d679c2105abf85] 3rd try git bisect bad 39d0eb4a68824cb35052d797e9d679c2105abf85 # good: [66de8f60e863befa61e5addaea539877dd78d064] modularized Watchdog & added Linux implementation * it can now handle any number of threads * it gives threads names - currently "main" (:= render-thread in spring-mt) & "sim" (only in spring-mt) * syntax of the Lua function changed too: Spring.ClearWatchDogTimer() --clears current thread's timer Spring.ClearWatchDogTimer("sim") --clears the timer of the "sim" thread git bisect good 66de8f60e863befa61e5addaea539877dd78d064

(0006482) abma (administrator) 2011-03-10 21:20	hm, maybe sounds are only played twice with a tiny time difference?

(0006483) zerver (developer) 2011-03-11 18:30	could be, that has happened in the past too

(0006563) abma (administrator) 2011-04-24 04:10	@jk: output of pulseaudio -v -v: http://pastebin.com/FiLThj1q [^] the stuttering seems to come from buffer-underruns: D: protocol-native.c: Underrun on 'Playback Stream', 0 bytes in queue. D: protocol-native.c: Requesting rewind due to end of underrun.

(0006564) abma (administrator) 2011-04-24 04:15	when changing /etc/openal/alsoft.conf: drivers = pulse (ubuntu default) to drivers = alsa,pulse stuttering seems to be gone

(0006565) jK (developer) 2011-04-24 04:15 edited on: 2011-04-24 04:16	don't use pulseaudio, use alsa directly? EDIT: this explains why I didn't encountered this issue.

(0006566) hoijui (administrator) 2011-04-24 18:32	hmm.. ok... should we consider this a sufficient fix? it would be nice if not 95% of our linux playerbase would have to do that on next release. can we prevent the buffer-under-runs from within spring? i would guess yes, as it does not happen in release version... so the question is, should we try to do it: * how much time would that require and is the time worth it? * would the "fix" likely be ok, or rather an ugly hack?

(0006567) abma (administrator) 2011-04-25 13:16 edited on: 2011-04-25 13:34	when adding this line to /etc/openal/alsoft.conf (with drivers = pulse) excludefx = eaxreverb stuttering is also gone + log spam of [Sound] CSound::UpdateListener: Invalid Name in my .springrc is a "UseEFX=0"

(0006633) hoijui (administrator) 2011-05-09 23:22	as discussed in meeting of 9. May 2011: problem: when EFX is enabled, pulseaudio is too slow in passing on packets to ALSA, and thus creates sound-lag solution: when pulseaudio is detected (linux only :D ), print a FAT warning like for low tech GFX cards, and disable EFX; do so only for runtime, not in the config, so it gets auto-enabled for the user when he un-uses pulseaudio.

(0006642) hoijui (administrator) 2011-05-11 10:16	i can not see anything about pulseaudio in my infolog, only ALSA, and i think it never was there... so the detection mechanism will be more complicated to implement. ..then again.. i have UseEFX=0, and i still get the stuttering. :/ in other words.. this solution is none :/

(0006643) abma (administrator) 2011-05-11 11:04	is LogSubsystems=Sound enabled? you should see something like this: [f=0000000] [Sound] Device: ALSA Software [f=0000000] [Sound] Available Devices: [f=0000000] [Sound] ALSA Software [f=0000000] [Sound] PulseAudio Software

(0006644) abma (administrator) 2011-05-11 11:19 edited on: 2011-05-11 11:20	imo UseEFX=0 doesn't work... or why does "excludefx = eaxreverb" change behaviour with UseEFX=0?

(0006650) hoijui (administrator) 2011-05-12 00:03 edited on: 2011-06-10 16:21	yes i see that, and this subsystem is enabled by default (Aus made it so when he reworked sound, and that never got changed). yeah.. sorry, read you wrong there. we'll have to investigate more here...

(0007594) abma (administrator) 2011-11-13 14:56	some more people have the stuttering problems: http://springrts.com/phpbb/viewtopic.php?uid=8601&f=11&t=27079 [^]

(0007595) spliff (reporter) 2011-11-14 08:47	Had a discussion with an OpenAL dev on openal@freenode. Got some information that may be useful but no conclusions except disabling either eaxreverb or pulseaudio. Chat log is here: http://echelog.matzon.dk/logs/browse/openal/1321225200 [^]

(0008687) abma (administrator) 2012-05-13 11:51	seems to be fixed in spring 88 / newer linux distributions...