DGAVCDecDI build 2001

[MOS-Marauder]

Hi Donald...a few small questions.

If i already generated a license file for the DGToolsNV can i simply copy it to the DGAVCDecDI dir ? (INdexer and decoder)

Is the Decoding it it faster then with the NV Tools? (Sometimes the NV Tools crash on 1080i Files .. i guess due Out of Memory on the Card (8xx MB)- Crashing the NV display driver then)

Chris

[Guest]

Yes, you can re-use the license on the same machine.

NV is usually marginally faster for transcoding HD. But they are on a par. A lot will depend on your CPU.

I'd appreciate hearing more about the crash. Is there an explicit pop-op error message or is it a hard crash?

[MOS-Marauder]

Yes, you can re-use the license on the same machine.

NV is usually marginally faster for transcoding HD. But they are on a par. A lot will depend on your CPU.

I'd appreciate hearing more about the crash. Is there an explicit pop-op error message or is it a hard crash?

Well at first the NVIdia Driver crashes and is being restarted. Then i get the Cuvid error decoding frame...(popup).

Then i have to kill the x264 Task because it continues to popup the errors. I never had this on 720p Files so i guess its related to the GFX Ram. (Zotac nVidia GeForce GTX 260 Synergy Grafikkarte (PCI-e, 896MB GDDR3 Speicher, Dual DVI, HDTV-Out, 1 GPU).

Functions used: Decode, deinterlace (mode1) and cropping directly in the DGI. (Source provided crop...message)

Chris

[Guest]

You have more than enough memory, so I doubt that is the problem unless you open multiple instances.

More likely is the Nvidia driver. The latest version has shown issues like this. You can try 196.21 and see if it makes a difference.

Followup in the NV thread please.

[MOS-Marauder]

You have more than enough memory, so I doubt that is the problem unless you open multiple instances.

More likely is the Nvidia driver. The latest version has shown issues like this. You can try 196.21 and see if it makes a difference.

Followup in the NV thread please.

Ok Trying... Heading to NV Thread...


Chris

[Alex-Kid]

Hi neuron2!

At first, I have to say that DGAVCDecDi works very well with some captures I have made from local ISDB-Tb (H)DTV channels, so I cannot do anything but to say congratulations. The same to schweinsz for being part of this.

I want to report a possible bug with streams that have errors (maybe due to lost packets or what): when trying to trim from 0 to k, and then from k+1 to n, (1080i source that I want to apply inverse telecine using Avisynth) it seems that fields (maybe frames) are shifted during playback, but it "plays" fine when stopping it and going frame by frame (with VirtualDub) everytime I go through frames k to k+1. I thought that MeGUI will handle it the right way, but it does wrong, and I finally got a file with video partially combed from frame k+1 to n. I'll put a sample when I get home and find a tool to cut it :)

I hope it can be understood (my native language is spanish). Saludos

By ALEX-KID

[Guest]

I'll need the stream and instructions for recreating the issue.

[Emulgator]

Many thanks for incorporating a frame-accurate software AVC decoder !

Just bought a DiAVC license.

The following applies for playback applications only !
DGAVCDecDI needs the diavc.ax additionally in its folder of course.

Just in case if somebody happens to have the diavc.ax registered from a folder that is not appropriate anymore
(as i did) and wants to move diavc.ax or switch diavc.ax versions:

The path to diavc.ax is registered in:

HKEY_CLASSES_ROOT\CLSID\{0BDBF960-0DE2-488D-8DFE-322908990B93}\InprocServer32
HKEY_CLASSES_ROOT\CLSID\{5DFA20A4-C343-45DF-A46E-EC1711E4EE32}\InprocServer32
HKEY_LOCAL_MACHINE\SOFTWARE\Classes\CLSID\{0BDBF960-0DE2-488D-8DFE-322908990B93}\InprocServer32
HKEY_LOCAL_MACHINE\SOFTWARE\Classes\CLSID\{5DFA20A4-C343-45DF-A46E-EC1711E4EE32}\InprocServer32

So if you need to move diavc.ax to a new folder for playback applications (MPC etc.)
you need to change locations in these regkeys using regedit.
I created such a folder in C:\Programs\DiAVC

It worked nicely for me and so I can control access easily by renaming the .ax.

(BTW, for playback purposes libavcodec performed with a bit less CPU load here
(1920x1090x24p BBB DivX7 public sample encode) using -5..-10..-15% less on a T7600G CPU
compared to DiAVC, but for frameserving DiAVC is of course the tool of choice, and can digest more flavours of AVC on my system)

[Alex-Kid]

Now I could cut it! Here's the file http://www.mediafire.com/?0th3ba16538923c (67.6 MB)

I've used the following script in Avisynth to apply inverse telecine:
dgsource("TVN HD1_07_25_2010_22_04_31_edit.dgi")
assumetff()

trim(0,631).doubleweave().pulldown(1,4)
\ +trim(632,664).doubleweave().pulldown(0,3)
\ +trim(665,713).doubleweave().pulldown(0,2)
\ +trim(714,749).doubleweave().pulldown(1,3)
\ +trim(750,1255).doubleweave().pulldown(1,3)
When going frame by frame in VirtualDub, inverse telecine is well done. When playing it (in VirtualDub too), the video looks combed in some intervals. When it's encoded to MKV with MeGUI (using x264), the video looks combed too, so frame-accuracy is lost.

That's it. Saludos

By ALEX-KID

[Guest]

Your stream has multiple corruptions. If I was in a bad mood I would ask why you are wasting my time with this.

[Alex-Kid]

Maybe I wasn't clear...

I know the stream has errors, as I said in post 56. And I know that corruptions couldn't be repaired if I apply inverse telecine the way I am doing it. I'll try to show you my point with a diagram:

I load the script into VirtualDub. Then:

1. I advance the video frame by frame:


error trim end| |trim start
| | |
frame 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 ...
status ok ok ok ok corr corr corr corr corr corr ok ok ok ok ok

2. If I hit play:

error trim end| |trim start
| | |
frame 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 ...
status ok ok ok ok corr corr corr corr corr corr comb comb comb comb comb

This occurs everytime I pass from a trim to another one: good frames show different status when going frame by frame and during playback. Which one is correct? Playback frames, so it wouldn't allow me to remove pulldown.

I hope it's clearer now. Saludos.

By ALEX-KID

[Guest]

As far as I know VirtualDub does not distinguish between play and stepping, i.e., my filter will see the same linear sequences of GetFrame() requests. The only difference will be the rate that the requests are issued. So I'm skeptical about this. Nevertheless, I will try to duplicate it.

BTW, why can't you use an adaptive IVTC?

[Alex-Kid]

BTW, why can't you use an adaptive IVTC? I just wanted to apply a "clean" IVTC. But you're right, TFM and TDecimate saved the day :)

Maybe it's not important anymore, but could you duplicate the issue? Could it be a possible bug?

Thanks for your advice. Saludos

By ALEX-KID

[Guest]

Random access will not be reliable when the preceding GOP is not decodable due to corruption.

[sumawo13]

Will this eventually support MPEG2 and VC1? I already have a license for DGDecNV and I'm not sure I want to spend $25 on something that only supports AVC.

[Underground78]

Will this eventually support MPEG2 and VC1? I already have a license for DGDecNV and I'm not sure I want to spend $25 on something that only supports AVC.

Then you will just have to buy a license for DiAVC and it's only $9.90. :)

[Guest]

Will this eventually support MPEG2 and VC1? I already have a license for DGDecNV and I'm not sure I want to spend $25 on something that only supports AVC. I have no plans for a SW version that supports all video types.

Anyway, if you have an Nvidia card and a DG license, there really is no point for you to get the DI version. I made it to try to bring ATI people in from the cold, at least for AVC video.

[Emulgator]

Here is a 720x576x25i 4:3 raw AVC test stream coming from PAL broadcast source, encoded by x264 r1703 through MeGUI 0.3.5.8,
to test compliancy for BD-R authoring using Sony DVD-A, using the following commandline:

--level 4.1 --bframes 3 --ref 4 --slices 4 --nal-hrd vbr --aud --b-pyramid strict --keyint 25 --min-keyint 2 --vbv-bufsize 24000 --vbv-maxrate 27500 --sar 12:11 --weightp 0 --colorprim "bt470bg" --transfer "bt470bg" --colormatrix "bt470bg"

http://dvd-manufactur.de/files/Werb.avc

Via Virtual Dub 1.9.9 DGAVCDecode.dll Version 1.0.9.0 can deliver frames that represent the source.
http://dvd-manufactur.de/files/Werb 0000.jpg

Via VirtualDub 1.9.9 DGAVCDecDI 2001 using DGAVCDecodeDI.dll Version 2024 and diavc.ax Version 1.1.1 delivers offset lines from the same source.
I guess the 12:11 SAR might blow it, because it looks like lines are displayed shorter than they should appear
and at the end only empty green lines are left to fill the frame.
With a 1920x1080x24p SAR 1:1 source (BBB) the very same chain had already worked nicely.
http://dvd-manufactur.de/files/WerbDI 0000.jpg

P.S. Playback of this source through MPC using Cyberlink Splitter + DiAVC 32bit 1.1.1 or libavcodec from ffdshow 3406 delivers a picture
but has some MBs still combed, maybe these can not do MBAFF correctly yet.
Preview in DVD-A looks ok, but shows "green" field markers only anyway.

BTW, DVD-A started muxing without transcoding, but aborted with the well known "This program has a bug. - m_ptsOfNextGOP is empty."

P.P.S. The index files:
http://dvd-manufactur.de/files/Werb.dga
http://dvd-manufactur.de/files/Werb.dgi

[Guest]

OK, I wasn't honoring the Avisynth frame pitch when copying the decoded frame. :stupid:

I have slipstreamed the 2001 build with the fix. You need the new DGAVCDecodeDI.dll.

After a reasonable stability level is achieved, I'll start bumping build numbers properly.

EDIT: Something is still not quite right, because is goes bad again at frame 1108 when playing straight through in VirtualDub. Investigating...

[Emulgator]

Many thanks for the fast response !
BTW, the preview in DGAVCIndexDI is ok.
I will continue testing.

Results with the new 2024.dll from 20:21 MESZ

Here frames from 0 to 1118 are displayed correctly,
from 1119 on i get skewed lines, first without green frames,
1133 being the first frame with green bottom lines through 1139.

Still many access violations in VirtualDub when hitting F2 to reload script,
but a step ahead...

[Guest]

It's great clip to shake things down with. Thanks for it. Working on fixes...

[Guest]

I slipstreamed the fix for the pitch going wonky at the end. Re-download and get DGAVCDecodeDI.dll.

I'll look at the VirtualDub F2 issue now.

[Emulgator]

Many thanks for your work, Donald !

[Guest]

No need to thank me, after all, you paid for it!

But in any case, thanks for your thanks. :thanks:

[Emulgator]

And the 1140 frame clip passes here beautifully.

While playing in VD using Enter and/or stepping backwards sometimes VirtualDub 1.9.9 crashes with:

"An out-of-bounds memory access (access violation) occurred in module 'diavc'...
...reading address 0802B288." (varying addresses)

VD 1.9.8 behaves the same. 5 of 5 positioning crashes now point to diavc.

F2 reload: these crashes seem to point rather to DGAVCDecodeDI.

[Dal]

I thank you for this!
And yes, this IS a day I've been waiting for :)

I actually licensed DGIndexNV and bought myself a cheap NVIDIA card. Just to realize I didn't have two PCI-Express slots in my motherboard :)
So I had to put it into an older PC, but it was so slow I ended up doing things "the old way" on my main computer.

$25 dollars more for this is pocket change, so I will donate another $15 for DGAVCDecDI alone. I know I don't need to, but fine software (and it's developers) like this are worth every penny.

Sorry for the happy-ranting :)

Update: Got a real cold shower now, when I realized this program only supports AVC, not MPEG2, which is my main usage. Of course, the name of the program should have tipped me off, but I just assumed it had the same functionality as DGIndexNV has. Also, when I read this forum thread again, this is actually mention before.

Oh well, here's hoping it will support MPEG2 in future versions.

[RedDwarf1]

I thank you for this!
And yes, this IS a day I've been waiting for :)

I actually licensed DGIndexNV and bought myself a cheap NVIDIA card. Just to realize I didn't have two PCI-Express slots in my motherboard :)
So I had to put it into an older PC, but it was so slow I ended up doing things "the old way" on my main computer.

$25 dollars more for this is pocket change, so I will donate another $15 for DGAVCDecDI alone. I know I don't need to, but fine software (and it's developers) like this are worth every penny.

Sorry for the happy-ranting :)

Update: Got a real cold shower now, when I realized this program only supports AVC, not MPEG2, which is my main usage. Of course, the name of the program should have tipped me off, but I just assumed it had the same functionality as DGIndexNV has. Also, when I read this forum thread again, this is actually mention before.

Oh well, here's hoping it will support MPEG2 in future versions.

Why don't you just use dgmpgdec if you are not wanting to use GPU assistance? It's been available for a very long time and you can have the satisfaction that your donation partly paid for that as well, even though it's free.

[Dal]

Why don't you just use dgmpgdec if you are not wanting to use GPU assistance? It's been available for a very long time and you can have the satisfaction that your donation partly paid for that as well, even though it's free.
That is what I'm using today, but it has some limitations the other tools don't have.
For example frame accurate cutting and demuxing several audio tracks in one go.

[LigH]

Reposted for convenience:

As described in the DGDecNV thread (http://forum.doom9.org/showthread.php?p=1433633#post1433633), I have some strange kinds of application crashes with different choices of crash report dialogs - sometimes the simple one (Application failure - unknown software exception 0xc000001d at address...), sometimes the extended one with "Message to Microsoft".

Because I have only a 6800 GS at home, I tried a batch encode with ivfenc on an AviSynth script which uses DGAVCDecDI instead of DGDecNV. Around the beginning of each pass, during the initialization phase (before the frame counter starts), dialogs pop up alarming that an application has crashed ... but ivfenc keeps encoding nicely, without issues, and the result looks quite correct. There are just two crash dialogs left when the encode finishes.

ED_0360_700_ivf.batE:\Programme\ivfenc\ivfenc.exe --codec=vp8 --passes=2 --pass=1 --fpf=ED_0360.stats --best --threads=1 --token-parts=1 --end-usage=0 --target-bitrate=700 ED_0360_DI.avs ED_0360_700.ivf
E:\Programme\ivfenc\ivfenc.exe --codec=vp8 --passes=2 --pass=2 --fpf=ED_0360.stats --best --threads=1 --token-parts=1 --end-usage=0 --target-bitrate=700 ED_0360_DI.avs ED_0360_700.ivf
E:\Programme\MKVtoolnix\mkvmerge.exe -o ED_0360_700.webm -w ED_0360_700.ivf ED.ogg

ED_0360_DI.avsLoadPlugin("E:\Programme\DGAVCDecDI\DGAVCDecodeDI.dll")
DGSource("ED_1080_CRF06_DI.dgi")
Spline36Resize(640, 360)

I wonder if DGAVCDecDI has similar "racing conditions" like the one which was recently fixed in DGDecNV Build 2026.

[Ulf]

neuron2 and schweinsz:
Will you sync the development of DGAVCDecDI and DiAVC so that the latest version of DGAVCDecDI works with the latest official version of DiAVC? This is not the case right now. DGAVCDecDI only works with the a version of DiAVC that neuron2 emailed me.

neuron2:
I have a short 1080i AVCHD clip that I did some tests with. DGAVCIndexDI reports the clip correctly as having TFF field order. In an AviSynth script, the same stream will be flagged as BFF so I have to add ComplementParity() to get the correct field order. An issue with DGAVCDecodeDI.dll?

neuron2 and schweinsz:
VirtualDub (build 32817) repetedly crashes if I play a script and then jump back a few frames. VirtualDub reports:
An out-of-bounds memory access (access violation) occurred in module 'diavc'...
...reading address 08AF01D0.
I can email VirtualDubs crash report if that would be of any help.

[Guest]

The source filter needs to return the correct parity to Avisynth.

I'm aware of the crashes on seeking and am working on it. Remember, this is still early alpha SW.

[Ulf]

Remember, this is still early alpha SW.

It's still very useful for me. Well worth the cost!

[Rob100]

My GTX260 (VP2) is holding back my encodes using DGDecNV as it will only decode 1080p AVC at about 50fps, so I was going to give this a try.

When encoding a Blu-ray to DVD my CPU is only running at 16-20%, which is great when I'm doing other things, but not when I'm in a hurry!

However, when I run DiAVC Registration.exe I get a blank machine ID :rolleyes: So I'm stuck. Have tried running as admin, I have UAC disabled - not sure what else to try.

Any idea's?

[rohn]

Is mkv header compression gonna be supported?

[onlym3]

When i open an avs (created by dgavcdecdi) in megui and then open another avs everything works fine. However when i open the same avs, then use the bitrate calculator, close that and then open another avs, megui crashes.

This is what win 7 tells me:

Problemereignisname: APPCRASH
Anwendungsname: MeGUI.exe
Anwendungsversion: 0.3.5.10
Anwendungszeitstempel: 4c8a8033
Fehlermodulname: DGAVCDecodeDI.dll
Fehlermodulversion: 0.0.0.2024
Fehlermodulzeitstempel: 4c7935ed
Ausnahmecode: c0000005
Ausnahmeoffset: 00012942
Betriebsystemversion: 6.1.7600.2.0.0.256.1
Gebietsschema-ID: 1031
Zusatzinformation 1: 0a9e
Zusatzinformation 2: 0a9e372d3b4ad19135b953a78882e789
Zusatzinformation 3: 0a9e
Zusatzinformation 4: 0a9e372d3b4ad19135b953a78882e789


The same thing happens on my other pc with win xp. The reason i am posting it here is because the error report mentions DGAVCDecodeDI.dll as the error module.

[LigH]

Donald - did you already check if DGAVCDecDI needs the same fix (loading at least one frame) as DGDecNV?

[Groucho2004]

@Ligh

You might want to post this at his support forum (http://neuron2.net/board/index.php), the response time there is much better. :o

[LigH]

Doing so then...

[dtv_user]

@neuron2

Hi Don;
I just bought a license for your lastest version of DGAVCDecDI and quite pleased with it.
Your dedication to doom9 and hard work is very much appreciated.
(Hope this is the right place to post this.)

Bob

[kypec]

(Hope this is the right place to post this.)

Err, not really...neuron2 is focused on his own forum (http://neuron2.net/board/index.php) since few months already.:cool: