Getting weird artifacting at the same place when encoding

[nomolu]

I've downloaded the new, slipstreamed dgdecnv2041.zip file. And I've updated my DGI file to use "DECODE_MODES 0,1,0". When I go to use the new DGDecodeNV.dll I get this error message from VirtualDub:

"Avisynth open failure:
DGISource: DGIndexNV/DGDecodeNV mismatch. You are picking up a version of DGDecodeNV, possibly from your plugins directory, that does not match the version of DGIndexNV used to make the index file. Search your hard disk for all copies of DGDecodeNV.dll and delete or rename all of them except for the one that has the same version number as the DGIndexNV.exe that was used to make the index file."

I figured, what the heck, let's recreate the DGI file in case the internal version number has changed. And when I try to launch DGIndexNV.exe, I get this error: "DGIndexNV.exe is not a valid Win32 application."

I turned off my antivirus software and cleared my web browser's cache and downloaded the dgdecnv2041.zip file again, but I'm still getting these errors.

[admin]

Please redownload and it should be OK. I accidentally slipstreamed the 64-bit version onto the 32-bit one. Sorry for the inconvenience.

As I said, nomolu's issue does not appear to be the same as the mystery issue of this thread. DECODE_MODES=0,1,0 affects only MPEG2 streams. Still I'd like all testing going forward to use this newest fixed version.

I'm thinking the D3D angle may be a red herring.

And BTW, any testing of CUDA versus CUVID that we did with 2041 now needs to be repeated, as this bug makes the results invalid for MPG and VC1.

[sandy_geser]

As I said, nomolu's issue does not appear to be the same as the mystery issue of this thread. DECODE_MODES=0,1,0 affects only MPEG2 streams. Still I'd like all testing going forward to use this newest fixed version.
I'm thinking the D3D angle may be a red herring.

I did think his issue might be different but thought I'd try anyway. h264 test reported earlier. now tried the vc1 movie. Same glitch as I had before with DECODE_MODES=0,1,0 + D3D=0. Using DECODE_MODES=0,1,0 + D3D=1 resulted in a glitch in the same spot but different looking one, more colorful for sure.

see comparison here (4 frames out of a 23-24 frames corruption):

http://i.imgur.com/Er00z.jpg

[nomolu]

Thanks for the update, admin. I tested it with a couple of other elementary MPG2 streams that were having the same problem and it fixed it for them as well. Too bad this didn't fix the original problem.

sandy_geser: For h264 files, I think that the first parameter for Decode_Modes needs to be a 1, as in Decode_Modes=1,1,0. And for VC1 files the last parameter needs to be a one, i.e., Decode_Modes=0,1,1.

[sandy_geser]

sandy_geser: For h264 files, I think that the first parameter for Decode_Modes needs to be a 1, as in Decode_Modes=1,1,0. And for VC1 files the last parameter needs to be a one, i.e., Decode_Modes=0,1,1.

I actually didn't touch the Decode_Modes settings as the fixed version of 2041 is already set to 0,1,0. Just mentioned the setting so there won't be any doubt I used the pre-fixed version. So really the only setting I manually changed was D3D.
Looking at the dgi files I created both the vc1 and h264 dgis use DECODE_MODES 0,1,0. I assume if it was supposed to be different DGIndexNV would have made the change itself when creating the project files. We don't normally need to manually
change settings.

[admin]

@sandy_geser

Can you please try with 1,1,1? Thank you.

[sandy_geser]

@sandy_geser

Can you please try with 1,1,1? Thank you.

same results as with 0,1,0.
What card did you use when you did your try with Count of Monte Carlo and couldn't replicate my corruptions?

[admin]

My 520.

[sandy_geser]

My 520.

the plot thickens

[sandy_geser]

admin,

I sent you a pm but my sent items is empty so not quite sure it got through. please let me know if you got it. thanks.

[admin]

I don't see a PM from you.

[sandy_geser]

I don't see a PM from you.

tried again now. think it went through this time.