View Full Version : DGAVCDec 1.0.9
Guest
30th August 2008, 13:51
Remains to be determined. My initial efforts are on Nvidia.
lucassp
30th August 2008, 15:41
any major progress on it?
Guest
30th August 2008, 17:33
When I have news I will post about it.
Why am I getting this deja vu feeling? :)
lchiu7
30th August 2008, 22:59
When I have news I will post about it.
I guess I will have to learn to be patient :)
Guest
1st September 2008, 01:29
* Removed the irritating warning about decoding not starting with a reference frame when starting a
play/preview for open GOPs. Examine the DGA file if you need to know if your GOPs are open.
* Added a new option in the Video menu: Copy Frame to Clipboard.
* Fixed missing info log data for CLI invocation with -e.
* Implemented proper relative/absolute path handling for the Load Project function.
* Implemented mouse scrollwheel support for GOP and frame stepping.
* Fixed missing SPS/PPS's in the DGA file when reloading a file without closing DGAVCIndex.
This can also happen when reloading a different file if it has identical SPS/PPS's to the previous file.
* Fixed crashing when a source file with no extension is loaded.
http://neuron2.net/dgavcdec/dgavcdec.html
rebkell
1st September 2008, 02:05
* Removed the irritating warning about decoding not starting with a reference frame when starting a
play/preview for open GOPs. Examine the DGA file if you need to know if your GOPS are open.
Love the playing without irritating warning fix, before it got difficult to determine how far between I(DR) frames certain jumps were, this works great.
I don't guess there is any chance you could display the current frame or time in the preview window?
Guest
1st September 2008, 07:38
I don't guess there is any chance you could display the current frame or time in the preview window? Can't do that because DGAVCIndex is not frame accurate (but DGAVCDecode is).
BTW, re-download alpha 35 as my fix for crashing when a source file with no extension is loaded was bugged.
canTsTop
1st September 2008, 09:33
Hello, Thank You for new version
With alpha 35 i get these warnings when open my any TS files (drag and drop file, and before i see video image):
http://www.ipix.lt/images/51501837.png
http://www.ipix.lt/images/22678374.png
There was no such errors with previous versions. For sample i re-uploaded with_errors_LTV.ts http://www.mediafire.com/file/azxjyatwffn/with_errors_LTV.ts
One more thing, then i select range in DGAVCIndex an chose to Save Project and Demux Video, isn't *.dga file should use demuxed video instead TS?
lchiu7
1st September 2008, 12:10
I am getting the first error message now, but not the second. Not sure it's related but I also am occasionally having dgavcindex when demuxing a file with AAC-HE LATM audio, stop demuxing part way through the file (the slider doesn't reach the end and there is still time to go on the counter). No other error occurs. But the output AAC file is shorter than the video length and when converted to AC3 and muxed back, the new file has no audio at the end of the file.
Can create and upload a sample if required
Thanks
Guest
1st September 2008, 16:51
Can create and upload a sample if required Yes, please upload a sample so that I can duplicate your issue.
Guest
1st September 2008, 17:40
With alpha 35 i get these warnings when open my any TS files (drag and drop file, and before i see video image Please re-download alpha 35. I fixed the regression. It happens only for files that have slices before the first SPS/PPS's, usually due to faulty cutting. Thanks for pointing it out.
One more thing, when i select range in DGAVCIndex and chose to Save Project and Demux Video, isn't *.dga file should use demuxed video instead TS? It just can't work that way because the video is demuxed as the project is saved, and the offsets in the TS and the ES files are different. You have to load the ES and save a project for that.
lchiu7
1st September 2008, 20:41
Yes, please upload a sample so that I can duplicate your issue.
OK - have you downloaded the other file I uploaded since I will need the space to upload this sample.
Thanks
Guest
1st September 2008, 20:59
You mean the one that crashes? Yes, you can delete it. And regarding that one, I am having trouble catching the error with an exception handler and I don't know why. But I haven't given up on it. :)
canTsTop
1st September 2008, 23:47
Please re-download alpha 35. I fixed the regression. It happens only for files that have slices before the first SPS/PPS's, usually due to faulty cutting. Thanks for pointing it out.
Some times i have problems with audio delay, maybe its related to this 'faulty cutting' (actualy its how DVBViewer starts recording), here is sample http://www.mediafire.com/file/aznb3yn3mwh/aud_del_prob_0.ts it has a/v synch problem (look at dialog at ~3:35min). But if i cut TS file with TS Packet Editor at keyframe (IF button) then audio synch is correct.
Thank You
meatwad
2nd September 2008, 05:39
I just recently began using DGAVCIndex with StaxRip and noticed my encodes would come out with blocky artifacts (probably due to the interlaced content). I'd seen this before in some other encoding programs and the cause was Haali Media Splitter. Assuming this was the case, I uninstalled Haali and made sure the codecs were unregistered with GSpot. I was hoping that this would be all I'd have to do, but it's rarely that easy for me... The transport streams are H264 1440X1080 and they are interlaced. If anyone else has come across this problem before, could you please pass the solution on to me. I've been pulling my hair out trying to figure out the cause and I'm contemplating a clean install. I've tried different versions of libavcodec as well as 5 different versions of DGAVCIndex. Google is usually my friend but I most have pissed him off because he's hardly saying anything. Anyway, I'll hold off on a clean install for a bit in case someone can mentor me out of this situation. Thanks in advance!
xopowo
2nd September 2008, 07:17
libavcodec = no workie for your interlaced content.
patience is a virtue
meatwad
2nd September 2008, 07:26
Thanks for the reply! That's sad, but a relief. I could have swore I read somewhere that Don had PAFF interlaced content worked into DGAVCIndex back in 2007 when everyone kept asking for it. Oh well, at least now I know why I'm getting the blocks. I will wait until he has it implemented or someone comes up with a work around.
lchiu7
2nd September 2008, 09:59
Yes, please upload a sample so that I can duplicate your issue.
OK - here it is. Once again sorry I can't make it smaller but when I try to make it smaller (cutting to include the section that fails) it either then works(!) or the program crashes as it did before. This is about 180Mb
www.datamix.info/chuck_ends_early.ts
This file will cause the program to stop demuxing about 7s into the file with about 8s to go and at frame 3444
Thanks
dansus
3rd September 2008, 17:17
libavcodec = no workie for your interlaced content.
patience is a virtue
Ah, thats why its keeps crashing at 0 frame.
Thats a shame, just started doing hd stuff today and im guessing all the BBC stuff will be interlaced, how annoying.
Guest
3rd September 2008, 17:26
Ah, thats why its keeps crashing at 0 frame. Nope, it just causes macroblocking errors. If it is crashing, then you should post a stream for me to use to duplicate the issue.
dansus
3rd September 2008, 18:35
Nope, it just causes macroblocking errors. If it is crashing, then you should post a stream for me to use to duplicate the issue.
http://www.mediafire.com/download.php?f4yo9xxyjp2
Basic script so far using Megui;
SetMTMode(2,0)
AVCSource("E:\Enc\hd.dga")
Lanczos4Resize(1280,720)
Thanks.
Edit: removing setmtmode fixes it.
yes, its got bad macroblocking. Deinterlacing seems to fix that.
Ranguvar
3rd September 2008, 18:37
SetMTMode is likely the problem... try SetMTMode(3, 0) SetMTMode(4,0) and SetMTMode(5, 0).
Sharktooth
3rd September 2008, 18:38
... try avisynth 2.57... the non MT version...
dansus
3rd September 2008, 19:18
SetMTMode is likely the problem... try SetMTMode(3, 0) SetMTMode(4,0) and SetMTMode(5, 0).
SetMTMode(3, 0) Works 22fps
SetMTMode(4,0) Works 7fps
SetMTMode(5, 0) Works 19fps
Many Thanks.
Guest
5th September 2008, 02:24
Update on CUDA developments...
All clips that fail with libavcodec are correctly decoded with CUDA. This is the big deal for me, because I'm highly doubtful that libavcodec will ever be completed. Using CUDA also finesses the difficult-to-debug crashes that libavcodec is prone too.
Performance without any optimization and several known inefficiencies: 1080P decoded at 52 fps with ~3% CPU. Nvidia says I can expect 65 fps with my video card, which is a low-end card. Also, planned changes to decouple CUDA from DX and D3D will bring further gains.
The API is highly conducive to incorporation into DGAVCDec. I plan to implement a decoder "HAL" layer to easily allow different decoders to be slotted in, including CoreAVC if they provide an API and SDK as they have promised.
BTW, Nvidia plans to support 5.1 AVC in the near future. I have been blown away by their excellent support as well. Started browsing their employment pages. :)
Sharktooth
5th September 2008, 02:29
uhm... i smell no more wine...
rebkell
5th September 2008, 02:38
Are you still contemplating a h264 editor type application, maybe using the cuda to re-encode to get frame accurate cuts? I'll buy an Nvidia card in a heartbeat, if you plan on going that route.
Guest
5th September 2008, 02:39
Don't you need it for Avisynth?
Sharktooth
5th September 2008, 02:42
yep, but if cuda highly depends on drivers and on DX/D3D, dgindex wont run on wine...
Guest
5th September 2008, 02:59
yep, but if cuda highly depends on drivers and on DX/D3D, dgindex wont run on wine... You mean DGAVCIndex, no doubt. But no, not at all, for two reasons:
1. I will continue to support libavcodec, such as it is. And CoreAVC if they release an API/SDK to me.
2. Nvidia is currently removing DX and D3D dependence, and CUDA video decoding will be cross-platform. It's underway, not a glimmer in someone's eye.
So the linux situation for correct native AVC decoding will be improved, not degraded.
Guest
5th September 2008, 03:01
Are you still contemplating a h264 editor type application, maybe using the cuda to re-encode to get frame accurate cuts? I'll buy an Nvidia card in a heartbeat, if you plan on going that route. Contemplation is cheap. :)
I work on things that 1) I need, and 2) I enjoy doing. Such a thing satisfies both. So, it's a matter of time and competing projects.
Guest
5th September 2008, 03:08
Here's another thought for you...
The CUDA video decoder engine decodes VC1! Would you like a frame accurate VC1 Avisynth frame server? I thought you might. :)
And oh, BTW, it also does MPEG2. Want better decoding performance for DGIndex? I thought you might. :)
Sharktooth
5th September 2008, 03:23
what about OpenCL?
rebkell
5th September 2008, 03:24
Here's another thought for you...
The CUDA video decoder engine decodes VC1! Would you like a frame accurate VC1 Avisynth frame server? I thought you might. :)
And oh, BTW, it also does MPEG2. Want better decoding performance for DGIndex? I thought you might. :)
:) I think, I could live with any and/or all of the above without a problem.
Guest
5th September 2008, 03:36
what about OpenCL? http://www.electronista.com/articles/08/06/17/opencl.standard/
It's a ways off but Nvidia is on-board.
Sharktooth
5th September 2008, 03:48
thats quite an old arcticle. things keep moving... and now there is official support from AMD/ATI (amd just started to integrate OpenCL in their Stream SDK...) and the OpenGL group (Khronos consortium) is working on standardizing OpenCL... things are moving fast.
canTsTop
5th September 2008, 10:41
Wow! It seems that CUDA is answer for DGACVDec decoding :) Does anyone knows if there are video card supporting CUDA with AGP interface?
Ranguvar
5th September 2008, 11:23
@canTsTop: Time to upgrade :p
lucassp
5th September 2008, 12:09
Here's another thought for you...
The CUDA video decoder engine decodes VC1! Would you like a frame accurate VC1 Avisynth frame server? I thought you might. :)
And oh, BTW, it also does MPEG2. Want better decoding performance for DGIndex? I thought you might. :)
I guess it doesn't do that on the VP2 Processor. The VP2 can do only H.264 and MPEG2 full hardware decoding.
There is also a CUDA Linux version. Does the VideoDecode Sample work on Linux?
EDIT: @Sharktooth: It doesn't matter which API is used (CUDA, OpenCL or FireStream, etc.) as long as it gives you direct access to the VP2/UVD. At this moment only CUDA gives us access to VP2.
Guest
5th September 2008, 13:20
I guess it doesn't do that on the VP2 Processor. The VP2 can do only H.264 and MPEG2 full hardware decoding. It still offers VC1 decoding. It's just that some of the work is still done on the CPU.
Sharktooth
5th September 2008, 13:28
@lucassp: im not so sure. being openCL an open computing language (it's not yet fully standardized, so i cant call it an open standard) hardware producers can implement what features they like including VP2 or UVD access... in OSX it is an OS API, so it's even easier...
the difference between opencl and cuda is cuda is a proprietary solution and works only on nvidia cards while opencl is ... ehrr... open.
Sulik
5th September 2008, 14:56
I don't think OpenCL includes any video-decode-specific APIs, pretty much everything is focused on general purpose parallel computing (just like DX11 Compute Shaders, and generic CUDA functionality), not dedicated hardware engines (though it would be nice to have a cross-platform equivalent to DXVA).
crypto
5th September 2008, 16:27
Update on CUDA developments...
All clips that fail with libavcodec are correctly decoded with CUDA. This is the big deal for me, because I'm highly doubtful that libavcodec will ever be completed. Using CUDA also finesses the difficult-to-debug crashes that libavcodec is prone too..)
Wow, this is great news. I am also suffering from libavcodecs PAFF macroblocking errors. If you need testers, let me know.
Guest
5th September 2008, 17:46
Wow, this is great news. I am also suffering from libavcodecs PAFF macroblocking errors. If you need testers, let me know. What video card do you own?
Manao
5th September 2008, 17:59
There seems to be a either huge memory leak or a robustness issue with version 1.0.0.35. I tried to index a 35+ GB m2ts file, and it ended up, 30 minutes or so later, with a popup saying starting with "AllocPPS" (i don't remember the precise error message :( ) When checking the memory used by DGAvcIndex at that moment, it reported 1.9+ GB of virtual memory.
I don't know whether the stream was corrupted (I doubt it since I know somebody else at work managed to index it with a previous version of DGAvcIndex - dunno which one) or if it's a regression.
I'll try to get more information (especially which previous version did work), but that won't be before I get back to work on monday. Where can I find previous versions, in order to search for a possible regression (if that would help) ?
rebkell
5th September 2008, 18:11
I'll try to get more information (especially which previous version did work), but that won't be before I get back to work on monday. Where can I find previous versions, in order to search for a possible regression (if that would help) ?
http://www.neuron2.net/dgavcdec/
Guest
5th September 2008, 18:12
They're all still online. Replace the XX as required:
http://neuron2.net/dgavcdec/dgavcdec100aXX.zip
I'll have a look at my memory management for SPS/PPS's.
Guest
6th September 2008, 03:09
Just for fun, I benchmarked Nvidia 8500GT versus CoreAVC on E8500 @ 3.8GHz using a 1080P25 AVC video. I forced the frame rate to 120fps using an Avisynth script.
Nvidia GPU: 66 fps (CPU 4%)
CoreAVC: 60 fps (CPU 60%)
Which would you prefer?
And consider that the PureVideo deinterlacing and 3:2 processing are available as well with the GPU. That gives you excellent deinterlacing in hardware!
The 8500GT is < $100.
Sharktooth
6th September 2008, 03:22
well the answer is obvious, isnt it?
Guest
6th September 2008, 03:36
I think so.
Waiting for the naysayers to chime in...
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.