IPB

Welcome Guest ( Log In | Register )

70 Pages V  « < 47 48 49 50 51 > »   
Reply to this topicStart new topic
foo_wave_seekbar, pretty accelerated seekbar
marc2003
post Mar 24 2013, 01:01
Post #1201





Group: Members
Posts: 4453
Joined: 27-January 05
From: England
Member No.: 19379



if that question is for me about the youtube clip, my original file is FLAC which i encoded myself from CD. i just searched youtube for the same song. its creation has nothing to do with me.

edit: 20 second sample here: https://dl.dropbox.com/u/22801321/2013/marc...ing%20More.flac (1.9MB)

@ gundelgauk, it seems i was a bit premature with my theory. i'm still getting the odd "normal" track which won't scan but then does on a restart. it's all a bit random really.

This post has been edited by marc2003: Mar 24 2013, 01:11
Go to the top of the page
+Quote Post
gundelgauk
post Mar 24 2013, 01:15
Post #1202





Group: Members
Posts: 8
Joined: 23-March 13
Member No.: 107351



QUOTE (marc2003 @ Mar 24 2013, 01:01) *
i'm still getting the odd "normal" track which won't scan but then does on a restart. it's all a bit random really.

I agree. While I was trying other files, I noticed that even files which scan and generate waveforms properly can cause one CPU core to spike and stay at 100% for me. This happens with MP3 and OGG as well as FLAC, at least for me. I also tried some "exotic" files like NSF and SID. Those cause more CPU cores to stay at 100% even though waveforms are generated, possibly because of multiple embedded tracks which are scanned in multiple threads (only speculation on my part)?

[EDIT]I tried your sample file. It does not generate a waveform with 0.2.30.1 and sends three cores through the roof for me.[/EDIT]

This post has been edited by gundelgauk: Mar 24 2013, 01:24
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 01:31
Post #1203





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



That sample reproduces the problem here indeed, it's related to the format and length of the file. It gets stuck in a twilight zone where the sample generator thinks it has emitted enough samples, while the consumer still lacks a few thousand samples.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
derty2
post Mar 24 2013, 06:14
Post #1204





Group: Members
Posts: 581
Joined: 5-June 11
Member No.: 91257



I tested the sample file from marc2003 . . .

Yes it breaks my foobar000 too. The waveform doesn't render and the CPU shoots to 25% and stays there permanently. The rest of foobar2000 is still usable and does not crash this time, however if I stop playing the marc2003 sample track and change playlists and start a new track, the CPU usage continues at 25% . . . . . . . in fact, I tried restarting foobar2000 and played a track, but CPU usage spiked to 25% again.

The offending sample needs to be permanently removed from any playlist to bring foobar2000 back to normal. . . . . Nope that doesn't work either! The offending track has caused foobar2000 CPU usage to permanently spike up to 25% no matter how many times I restart, no matter if I delete the track from all playlists.

Eventually, all went back to normal only when I did this: exit foobar2000, delete the track from my computer, start foobar2000.

This post has been edited by derty2: Mar 24 2013, 06:30
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 07:42
Post #1205





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



derty2: In order to let you restart foobar2000 freely when there are background scans in progress, the component helpfully stores the pending scans in the database, and restores the queue on startup. Normally this is desired behavior, and if you really don't want a job, you can close foobar2000, delete the offending job in the database with some surgery, and restart. There is way too much work to make a scan queue management interface, when the normal usage pattern has nothing that needs one.

Anyway, I've just uploaded 0.2.31, which handles all known problem tracks properly, by taking some morning hours to rewrite the offending snake's nest of code.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
gundelgauk
post Mar 24 2013, 08:56
Post #1206





Group: Members
Posts: 8
Joined: 23-March 13
Member No.: 107351



That's great work, Zao. I can confirm that for me the problem does not occur with marc's and my FLAC files. That also includes MP3 and OGG files which would generate a waveform but still cause permanent CPU usage. Thanks!

Some quick testing revealed that it still occurs with a SID (C64) file, however a NSF (NES) file works. But normal audio files seem to work perfectly. Thanks again!

[EDIT]Somehow I got foobar to crash by playing a midi file. Then after restarting foobar, it would crash on playing any file. Clearing the playlist in safe mode and deleting the waveform DB did not help. After uninstalling the component, deleting the DB and reinstalling the component it works normally again. I guess I just shouldn't play such strange files. biggrin.gif[/EDIT]

This post has been edited by gundelgauk: Mar 24 2013, 09:09
Go to the top of the page
+Quote Post
EpicForever
post Mar 24 2013, 08:58
Post #1207





Group: Members
Posts: 706
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



I got exactly the same situation like derty2 with marc2003 sample... When I restart foobar it opens the handle for this file, despite it isn't on any playlist nor in Library, and it continuously uses 17% of my CPU. That's exactly 1/6 of six core Phenom... In a minute I will try 0.2.31 to check it.

EDIT:
At least with marc's FLAC iv. 0.2.31 works 100% OK. I haven't noticed any file from my own Library that exposed such behaviour, as I forced foobar some time ago to calculate all seekbar signatures at one time smile.gif . I went to a job, and foobar did the thing when I was out tongue.gif

This post has been edited by EpicForever: Mar 24 2013, 09:10
Go to the top of the page
+Quote Post
marc2003
post Mar 24 2013, 10:36
Post #1208





Group: Members
Posts: 4453
Joined: 27-January 05
From: England
Member No.: 19379



rather bizarrely, the 20s sample still doesn't work for me. <<oops, i just remembered i have it configured not to scan waveforms outside my library. laugh.gif

but the full length track does and that's all that matters. i've also tested the quietest track i can think of and that is also fine.

thanks for all your work Zao.

This post has been edited by marc2003: Mar 24 2013, 11:09
Go to the top of the page
+Quote Post
gundelgauk
post Mar 24 2013, 10:51
Post #1209





Group: Members
Posts: 8
Joined: 23-March 13
Member No.: 107351



I actually got very curious so I tried to compile the component myself to be able to debug it. But I'm inexperienced with VS and after installing boost I found out that VS Express does not have ATL so I'm kind of stuck. Too bad, it seemed like a very interesting problem.

It works well enough with the latest version, so thanks again for your work!
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 18:36
Post #1210





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



gundelgauk: Good luck with that, it was never planned for independent building, so it's kind of tied to my SDK directory and system-installed libraries.
I don't think I have modified the SDK or the Boost too much, but I can't be sure, my system install is shared between a bunch of projects.

The dependencies are pretty much:
  • Visual Studio 2010 compiler,
  • DirectX SDK February 2010,
  • ATL,
  • WTL,
  • Boost 1.53-ish,
  • Scintilla (for D3D9 frontend),
  • reasonably vanilla foobar2000 SDK,
  • possibly Intel TBB, can't remember if I still use it.

In particular, June 2010 won't do, as there's bugs in the shader compiler.
The VS 2010 dependency is because all those filthy XP users won't stop crying if I migrate to 2012 and drop XP support tongue.gif


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
lvqcl
post Mar 24 2013, 19:56
Post #1211





Group: Developer
Posts: 3362
Joined: 2-December 07
Member No.: 49183



AFAIK MSVS 2012 Update 1 does support Win XP
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 20:00
Post #1212





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



The v110_xp toolset is not usable in my world. For starters, Boost.Build will not ever target it, and it has lots of hidden caveats and is not as tested as a real compiler.
Not to mention that it doesn't _gain_ you anything but a slightly newer and untested mutilation of the toolchain. It's still using a bastardised WinSDK7.

To be really honest, I cannot trust a toolset that they put out in a hurry because people got their panties in a twist over not being able to target an EoL:d OS.

This post has been edited by Zao: Mar 24 2013, 20:03


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
gundelgauk
post Mar 24 2013, 20:15
Post #1213





Group: Members
Posts: 8
Joined: 23-March 13
Member No.: 107351



Zao, thanks for elaborating. It looks like I won't be able to set up a build environment in a reasonable time then, especially with not much VS experience. As an alternative, would it be possible for you to include debug information in a build or something, so that we can get you more helpful info with a standalone debugger or something to that account? It's just that I used to be able to play anything in foobar with your waveform seekbar up until recently and now that I seem to have problems with some file types, I really miss it.
Go to the top of the page
+Quote Post
EpicForever
post Mar 24 2013, 20:56
Post #1214





Group: Members
Posts: 706
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



Zao - as I said before I think that there should be "normalization" applied (division every point of seekbar waveform by maximal value calculated in whole file) to waveform when it's downmixed to stereo. If my file was looking really like this:



it would be absolutely unlistenable (especially at the end).

The same file, after conversion to 2ch FLAC with Converter and "Downmix 5.1 to stereo" DSP enabled gives completely different waveform:



And this is listenable... And This is how I expect to Waveform Seekar look in "Downmix to stereo" mode, when I play regular 5.1 version. Of course I understand that accuracy may be limited, but definitely it isn't diplayed properly at this time.

This post has been edited by EpicForever: Mar 24 2013, 20:57
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 21:06
Post #1215





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



I will never normalize, as that destroys information as to the actual signal level of the waveform. If you want that, the best I can provide is a min/max parameter to the effects so that they may normalize.

The fundamental problem is two-fold.
  • I downmix the final waveform, not the source samples as they are not available.
  • The waveform I extract is optimistic. For each bucket, it contains the numerically minimum/maximum value for the whole bucket.

The optimistic extraction means that I probably introduce some additional energy into the waveform.
I could try to use the standard DSP if present, but I doubt it'd do much good at this stage.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
EpicForever
post Mar 24 2013, 21:31
Post #1216





Group: Members
Posts: 706
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



Hmm... You are right about normalization performed on every waveform. This will destroy component functionality. Maybe it would be possible to trigger normalization only when there are 2 conditions met: source signal has more than 2 channels and user has chosen "downmix to stereo/mono" display mode? Or maybe it is possible to check how "Downmix 5.1 to stereo" DSP performs calculations of final track level, and use it to downmix final waveforms to resulting "downmixed" one ?

When I complain about functionalities of certain components I recall someone's forum signature... "This job would be great if it wasn't for the users." Haha... Peter is right. Good for me I am not programmer - I don't have to listen others complaints. Bad for me I am not programmer - I have to rely on others good will, free time and skills...
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 22:14
Post #1217





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



I cannot reproduce any failure to scan with the test files I have for MID, MOD, NSF.
Whoever fails to analyse these, please provide files.

Edit: I got a repro on SID, yay.

This post has been edited by Zao: Mar 24 2013, 22:50


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 22:36
Post #1218





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



QUOTE (gundelgauk @ Mar 24 2013, 21:15) *
... include debug information in a build or something, so that we can get you more helpful info with a standalone debugger or something to that account ...

For a given release foo_wave_seekbar-X.fb2k-component, there's an archive of the debug symbols at http://www.acc.umu.se/~zao/foo_wave_seekbar-X-archive.fb2k-component, as an off-site backup. They should match the corresponding tag in the Github project.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
Zao
post Mar 24 2013, 22:46
Post #1219





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



For all you LFE lads, there's 0.2.32 now, which drops the LFE channel on the floor. If you want to see it, don't use downmixing.
This release also fixes an accident in RMS computation which I didn't spot as I don't visualise RMS myself.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
gundelgauk
post Mar 24 2013, 22:59
Post #1220





Group: Members
Posts: 8
Joined: 23-March 13
Member No.: 107351



Wow, Zao, I don't know what you did but this is awesome. I just finished uploading some midi, mod and other files which caused crashes when I saw the new version. Now with 0.2.32, these crashes are gone and waveforms are generated normally. So far I could only find one file which still seems to cause high CPU load. It's a SID file, I'll send you a message with a link.

Thank you very much! It seems I can listen to all my weird stuff with waveform seekbars again! biggrin.gif

Also, thanks for the hint about debug symbols.
Go to the top of the page
+Quote Post
Zao
post Mar 26 2013, 14:15
Post #1221





Group: Members (Donating)
Posts: 902
Joined: 25-September 03
From: Umeň, Sweden
Member No.: 9001



This should probably resolve your SID issues, turns out that my component brings out the worst in others laugh.gif

foo_sid version 1.30 highlights: Fixed a race condition which affected starting a batch of conversions or wave seekbar scans without starting a single track for playback first.


--------------------
Zao shang yong zao nong zao rang zao ren zao.
To, early in the morning, use a chisel to build a bathtub makes impatient people hot-tempered.
Go to the top of the page
+Quote Post
tiptoetan
post Mar 26 2013, 22:06
Post #1222





Group: Members
Posts: 1
Joined: 26-March 13
Member No.: 107399



hello and sorry for asking probably not something very very related - personally I found wave_seekbar really really useful - I use foobar + wave_seekbar to somewhat come closer to programs like vdj for basic dj-ing. there are various reasons for this and they are off-topic.. what I miss most and can't find any plugin to do is to insert custom mix points on the run - wave_seekbar helps to identify visually the place to start, and if you are fast enough, with long crossfade enabled you can skip lengths by clicking - but wouldn't it be great to be able to set track start point visually before activating it for play? I know this is quite beyond the reasons why wave_seekbar exists, and would probably call for another plugin entirely, possibly basing on wave_seekbar... but - was just wondering.. it would be interesting to hear your thoughts on the matter... thank you.
Go to the top of the page
+Quote Post
Dario
post Mar 27 2013, 23:49
Post #1223





Group: Members
Posts: 158
Joined: 20-September 11
Member No.: 93842



There is something terribly wrong with this component. It was undoubtedly caused by the recent updates. Waveforms disappear randomly, tracks display wrong waveforms (for example, the second track of an album, when player after the first, displays the waveform of the first track)...

This post has been edited by Dario: Mar 27 2013, 23:50
Go to the top of the page
+Quote Post
Propheticus
post Mar 28 2013, 00:08
Post #1224





Group: Members
Posts: 219
Joined: 10-September 11
Member No.: 93615



Having no such problems here with the latest version. Maybe another component or settings causes this?
Go to the top of the page
+Quote Post
Manchesterer
post Mar 28 2013, 07:48
Post #1225





Group: Members
Posts: 25
Joined: 2-August 11
Member No.: 92721



+1 for problems with latest update (0.2.33) - files (mp3) added to the library after component update have no waveform, trying to extract seekbar signature manually also does not help. Some waveforms appear after few plays, only to disappear on the next play.
Go to the top of the page
+Quote Post

70 Pages V  « < 47 48 49 50 51 > » 
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 30th August 2014 - 11:36