DGDecNV (linux)
Moderator: Sherman
DGDecNV (linux)
OK, thank you DJ. It's working! I'll sort out the library path stuff first, then address your other matters and further ones as they arise.
DGDecNV (linux)
Hey DJ, where did you put libcuda.so and libnvcuvid.so? I'm thinking of having them in /usr/local/lib using -L to specify that path. libmp4v2.so will go there too. libdgdecodenv.so can be anywhere that Vapoursynth can find. I usually load plugins explicitly in the script.
DGDecNV (linux)
I already had cuda stuff installed in /usr/lib, so didn't confirm for overwrite.
I think it's ok to put into /usr/local/lib, alternatively you can build static libmp4v2 and link against that static lib .
I think it's ok to put into /usr/local/lib, alternatively you can build static libmp4v2 and link against that static lib .
DGDecNV (linux)
Good idea, I'll do that. Thank you.
DGDecNV (linux)
Have another snafu that is causing relative paths on linux to be problematic.
On Windows, if I open a vpy script and do GetCurrentDirectory(), it shows the path of the DGI file in DGSource(). However, if I do getcwd() on linux, it shows the path of the libdgdecodenv executable. I'm confused but still investigating.
Would it be horrible to require full paths in the DGI file and script, i.e., ditch Full_Path_In_Files = 0?
The mp4v2 library is now static.
On Windows, if I open a vpy script and do GetCurrentDirectory(), it shows the path of the DGI file in DGSource(). However, if I do getcwd() on linux, it shows the path of the libdgdecodenv executable. I'm confused but still investigating.
Would it be horrible to require full paths in the DGI file and script, i.e., ditch Full_Path_In_Files = 0?
The mp4v2 library is now static.
DGDecNV (linux)
Since I don't rely onto relative path thing, it's fine to ditch by me. I'm storing files on different drives and making simlinks for easier access at the project directory, so relative paths usually not required.
DGDecNV (linux)
Thank you for the feedback, DJ. I'll probably just ditch it. The original intent was to simplify moving things to a different directory but how often does that really happen after an index is made? And it's not so hard to edit a DGI file if needed.
I'll make a new test release today with the library situation sorted out. Then I will address the documented limitations.
Did your encode go OK and what did you use for encoding?
I'll make a new test release today with the library situation sorted out. Then I will address the documented limitations.
Did your encode go OK and what did you use for encoding?
DGDecNV (linux)
I have a deadline this friday, not even tried to encode yet. Will try this night...
DGDecNV (linux)
No problem, no hurry.
DGDecNV (linux)
Uploaded a new release:
http://rationalqm.us/dgdecnv/binaries/dgdecnv_linux.zip
The MP4V2 library is now static and the cuda/cuvid libraries are removed as they will be used from the installed nVidia driver. The user manuals are updated.
http://rationalqm.us/dgdecnv/binaries/dgdecnv_linux.zip
The MP4V2 library is now static and the cuda/cuvid libraries are removed as they will be used from the installed nVidia driver. The user manuals are updated.
DGDecNV (linux)
Then I will test now with new version, since I worked to 1 a.m. yesterday and had no stamina to arrange some encode job before sleep.
DGDecNV (linux)
Few tests made. It looks ok for me, just a small issue with dirty last frame which I encountered in windows version. It's non-deterministic, so I don't know if it can be fixed at all, I simply trim dark frames in the end to deal with those glitches.
DGDecNV (linux)
Thank you, DJ. If it matches the Windows version then the design goal was met. Nevertheless, if you want to cut the last 50MB of the stream for me I'd be willing to investigate the last frame issue. There are several possible explanations.
DGDecNV (linux)
I have uploaded entire m2ts.
DGDecNV (linux)
Thank you!
DGDecNV (linux)
The 23 black frames at the end appear to be encoded in the stream. They show up with MPC-HC as well.
DGDecNV (linux)
Sometimes I get last frame black (expected behavior) and sometimes last frame contains glitches, frame with empty luma and repeated UV / full repeat from last non-black frame (that's wrong)
DGDecNV (linux)
Apparently it only happens when frames requested asynchronously.
Linear scroll "frame-by-frame" is fine for some reason.
Linear scroll "frame-by-frame" is fine for some reason.
DGDecNV (linux)
Gotcha, thanks. I'll look into that.
DGDecNV (linux)
Can't see any issue with it. I put Reverse() in the script so that every frame is accessed randomly. Never saw a glitch.
DGDecNV (linux)
Then just leave it as is. It might be a cuvid thing, so nothing to do with plugin itself.
DGDecNV (linux)
Got all the FPS detection fixed up. Well, at least I think so. Gonna make a new release but want to sort out the INI file situation first. We may not even need one for linux. But putting dgindexnv in a root directory, such as /usr/local/bin is currently an issue, as the INI file is not writable there.
Then need to get back to Windows to catch up on some trouble reports and enhancements.
Then need to get back to Windows to catch up on some trouble reports and enhancements.
DGDecNV (linux)
As for settings, you might want to check ~/.config/dgtools/dgindexnv.ini (per user config) or try second time for /etc/dgtools/dgindexnv.ini (global config directory). If nothing found, fallback to write new config into user dir.
DGDecNV (linux)
There is no need for an INI file so I removed it. All the settings affected only things that were stripped out. I'll keep your suggestion in mind if the need for settings re-arises.
Gonna release a new test version shortly.
Gonna release a new test version shortly.