Support forum for DG Tools

Third slipstream. Getting close to a stable 2041.

* Fixed MKV parser so it accepts files with DisplayUnit=3 (recent change to the spec).

* First phase of reporting field order when doing Save Project. Currently this works for MPEG2
and VC1. When I add AVC support I will document it properly.

* The DGI file format version number is bumped to NV14.

Re-download and update both DGIndexNV and DGDecodeNV.

Another slipstream for 2041. The crackers will be mad. Just in case anyone is wondering I bump the version number on major functionality change only. If you want to be sure to have the correct latest version, just re-download it at any time.

* Fixed a problem with honoring frame repeats for MPEG2.

* Fixed a problem with the CLI -at option.

Slipstreamed a fix for LPCM audio demuxing from program streams.

Build 2042
----------

* Fix broken audio demuxing from interlaced VC1 streams.

* Some streams were encountered that had all I frames but for which the
GOPs were marked as open. This was fixed so that the GOPs are considered closed.

* Added support for GTX 680 (Kepler). This required moving to CUDA 4.2,
so some notebook/OEM drivers may be impacted (especially on WinXP).

Slipstreamed a fix into 2042. Some cases where start codes were split between transport packets were not handled correctly. Redownload 2042 and update DGIndexNV. You'll have to remake your project only if you were affected by this bug.

Slipstreamed another fix into 2042. The save project function could abort prematurely for some transport streams.

Here is a version 2043.

http://rationalqm.us/dgdecnv/dgdecnv.html

Build 2043
----------

* Support DXVA mode. Note that DXVA mode requires D3D mode to be enabled.

* Add support for AVC field order reporting.

* Implement field order reporting on preview (used to be done only on save project).

* Make field/frame repeats display in display order in the info dialog (previously was incorrectly
shown in decode order).

* Correct inaccuracies in film percentage reporting.

* Report more information in the 'Options/List GPU Devices' dialog, including the VP engine version.

* Initial window size of DGIndexNV is increased to 488 x 320.

* Added a fun anime public-domain splash screen.

* Revised the INI file version handling so that an existing INI file isn't defaulted by a new
DGIndexNV version unless the INI file format actually changed.

* Progress percentage is now reported to standard output during save project as in DGIndex.

* Fixed a bug that caused the wrong frame size to be written to the DGI file if cropping was
first enabled and then disabled.

* Fixed a bug such that an Avisynth script was created even though script creation was disabled.

* For CLI input, forward slashes are converted to back slashes.

* The timestamps dump now included GOP timecodes for MPEG2 streams.

* Demuxing VORBIS audio to Ogg from MKV files is now supported.

* Previously, if you opened an MKV file with video other than AVC/MPEG2/VC1,
you received an error popup and then closing DGIndexNV caused a crash. The
crash no longer occurs and the file is left open to allow audio demuxing.

* Previously, some VC1 streams with spurious, but also illegal, sequence headers
not followed immediately by an I picture were not properly handled. These streams
were seen on bluray disks and so a workaround has been implemented such that spurious
sequence headers are ignored instead of generating an error.

* Update documents for latest functionality.

admin it's still 2042 on the download page.
cheers!

hejhopp1 wrote:admin it's still 2042 on the download page.

What download page? It all looks fine to me.

same here; still 2042 on
http://rationalqm.us/dgdecnv/dgdecnv.html
weird.

It's only in your cache, hit Ctrl+F5 to re-download the page.

In what kind of situations is DXVA mode usefull/needed? Is it faster than "default" mode to decode frames? or is it giving support for more formats than MPEG2, h264 and VC1?

The only thing I know about it definitely is that it solved the decode glitches that Nick007 has reported in another thread. I am currently working with nVidia on this and I will seek clarification on the functionality of these different modes. They exist in the CUVID API so I make them available.

I have slipstreamed some stuff into 2043. Redownload to get these things.

* For CLI input, forward slashes are converted to back slashes.

* Fixed a regression, no audio was shown in the audio listbox of the Info dialog
during play/preview after the first play or preview showed it.

* The timestamps dump now includes GOP timecodes for MPEG2 streams.

* Demuxing VORBIS audio to Ogg from MKV files is now supported.

* Previously, if you opened an MKV file with video other than AVC/MPEG2/VC1,
you received an error popup and then closing DGIndexNV caused a crash. The
crash no longer occurs and the file is left open to allow audio demuxing.

Made another slipstream:

* Previously, some VC1 streams with spurious, but also illegal, sequence headers
not followed immediately by an I picture were not properly handled. These streams
were seen on bluray disks and so a workaround has been implemented such that spurious
sequence headers are ignored instead of generating an error.

If you re-downloaded to get that last slipstream, please do so again as it had a bug.

Build 2044
------------

* Added MP4 container support.

* Fixed wrong filename being used for LPCM demuxed from program streams.

http://rationalqm.us/dgdecnv/dgdecnv2044.zip

Build 2045
----------

* Include "demuxed." in the filename of a demuxed file only if the
source stream is an elementary stream (ES).

* Provide options for what the program response should be when an
audio mismatch is seen.

* Show the available template files in the AVS Template drop down
menu and make these clickable to quickly change the active template file.

* Add checks for encrypted audio and offer user the option of how to act
when they are detected.

* Add check on hrd->cpb_cnt_minus1 so that erroneous values do not crash
DGIndexNV.

http://rationalqm.us/dgdecnv/dgdecnv2045.zip

I slipstreamed two fixes into 2045:

* For AVC video, B pictures that immediately follow an IDR in coding order are decodable and are now properly decoded and displayed. Previously,
they were treated as leading B pictures and were replaced with a copy of the first decodable reference picture, i.e., the IDR.

* An error check was added when opening MKV files to prevent a hard crash when a stream type is not known to DGIndexNV.

Redownload 2045 to get these fixes.

Build 2046
----------

* Added an INI file option 'Bare_Demux' for compatibility of demuxed video files
with some authoring applications.

* Fixed bugs in the multiple AVS template file support: 1) do not list .avsi files in the
template list, 2) fix overrun of the template file list leading to crashes, 3) increase the
maximum number of template files to 20 (from 10).

http://rationalqm.us/dgdecnv/dgdecnv2046.zip

I have slipstreamed a fix into 2046. Some streams that should have their audio demuxed with 0ms delay were getting demuxed with a delay. Also made a very minor fix to the AVC info dialog layout. Re-download and update DGIndexNV.exe to get these fixes.

I slipstreamed VP6 (Maxwell) support into 2046, so re-download if you need it. All you need is to take the new NV12ToRGB24_sm_50.cubin and put it in your existing installation. Be sure to use the correct 32-bit or 64-bit version.

VP6 doubles the decode frame rate. The GTX 750 Ti has a VP6 and is inexpensive.

Double compared to...? VP5 ?

Yes.

neuron2 wrote:I slipstreamed VP6 (Maxwell) support into 2046, so re-download if you need it. All you need is to take the new NV12ToRGB24_sm_50.cubin and put it in your existing installation. Be sure to use the correct 32-bit or 64-bit version.

VP6 doubles the decode frame rate. The GTX 750 Ti has a VP6 and is inexpensive.

My 9600GT is a VP2 I think and I bought-but-uninstalled the GT250 so your post here caught my attention!

It's not obvious to me what the "Ti" designation means--is this better than the non-Ti version(s) somehow? I used Newegg to compare two cards and the one diff I could find was 512 CUDA cores vs. 640 CUDA cores for the Ti version of it. Izzat it?