AVS 2.5.6 Beta3

[FarCry]

AVS 2.5.6 Beta3 [050505] [show only this release] 2005-05-05 14:45

has anyone tried it with dgdecode 1.3.0 and DVD-RB?

[Fishman0919]

I have tried it with dgdecode 1.1.0 with no problems but DVD-RB doesn't work with dgdecode 1.3.0.

dgdecode 1.3.0 changes the .DV2 structure (Format)so DVD-RB won't work.

http://forum.doom9.org/showthread.php?s=&threadid=89681

About 9 posts down

[onesoul]

@ jdobbs

Are you planning to support latest dgdecode? It has many bug fixes and includes deblock function too which is considered very good.
Currently I'm using Deblock(quant=25) for deblocking ;) and BlindPP(quant=2,cpu2="ooooxx") for deringing. It's incredible how detail is kept while removing blocks/noise, the quant values are the default ones, the authors chose them well I say.
And would it be possible to add a option to choose idct=4 which uses 64-bit IEEE-1180 decoding Reference, it's supposed to be the most accurate which I am using now and actually can see differences if I do frame comparison.

quoting dgdecode user manual:
Which iDCT you should use depends primarily on what CPU you have and to a lesser degree, on how accurate of an iDCT you desire. Most people will not be able to tell the difference in quality between these algorithms but they can be easily observed by combining the AviSynth filters Subtract and Levels. All of the available options are IEEE-1180 compliant, except for SSE/MMX (Skal).
Qualitywise: IEEE-1180 Reference > 64-bit Floating Point > Simple MMX (XviD) > Remaining iDCTs.
Speedwise: SSE2/MMX and SSE/MMX (Skal) are usually the fastest. The IEEE-1180 Reference is easily the slowest.

[Axlemar]

Is there a dgdecode version that supports idct=4 and works with rebuilder or is the new but modified one the only one?

[onesoul]

Originally posted by Axlemar
Is there a dgdecode version that supports idct=4 and works with rebuilder or is the new but modified one the only one? Any version of dgdecode supports idct=4 and you can use this setting with dvd-rb. The setting represents the way the frames are decoded which don't interfere in how dvd-rb works.

[Axlemar]

oh, ok I understand now.

[Wilbert]

A small warning about AVS 2.5.6 Beta3 [050505], dgdecode 1.3.0 (upto 1.3.1 beta3) and BlindPP:

http://forum.doom9.org/showthread.php?s=&threadid=94184 (post Boulder)

[Axlemar]

I hate asking a question in an old post, but how would you go about using idct=4 with rebuilder?

[Fishman0919]

You can manually open the .avs files with notepad and add/change idct=4 or you can use a mass files editor..... http://www.download-by.net/system-utilities/text-docment-editors/13967,mass-file-editor.html

But you will really be hard pressed to see a diff from idct=7 (maybe only with still pics), and idct=4 will 2x's your encoding time.. I go from about 4.0 with idct=7 to about 2.3 with idct=4 in dvd-rb.

[Axlemar]

So rebuilder does the decoding and converting after the prepare stage?

[Fishman0919]

Originally posted by Axlemar
So rebuilder does the decoding and converting after the prepare stage?

???? not sure what you mean, but dvd-rb in the "Prepare" phase calculates bitrates for each cell, writes all the .avs files for them, rebuilder.ini...., rebuilder.ecl...... then in the "Encode" phase the encoder reads the .avs files to reencodes the cell, reducing it to it's new bitrate. The "Rebuild" phase...well rebuilds the.... you get it.

[Axlemar]

What I am wondering is if I change all of the idct=7 to idct=4 after the prepare stage and then encode, will it be using the idct=4 or do I have to try to set it up to the prepare stage somehow? The idct affects the accuracy done in the encoding stage right?

[Fishman0919]

Originally posted by Axlemar
What I am wondering is if I change all of the idct=7 to idct=4 after the prepare stage and then encode, will it be using the idct=4 or do I have to try to set it up to the prepare stage somehow? The idct affects the accuracy done in the encoding stage right?

idct=4... idct is how accurate the mpeg2 steam is read. If none is present in the .avs file then whatever wae used to create the .d2v file will be used.

A quote from DGDecode.html

idct : 0 to 7 (default: 0)

iDCT Algorithm.
For more infomation on iDCTs please see Appendix C.
Please see Appendix D for supported CPUs.
- 0 : Use value specified by DGIndex (iDCTs 6 & 7 unavailable in DGIndex)
- 1 : 32-bit MMX
- 2 : 32-bit SSE/MMX
- 3 : 64-bit Floating Point
- 4 : 64-bit IEEE-1180 Reference
- 5 : 32-bit SSE2/MMX
- 6 : 32-bit SSE/MMX (Skal)
- 7 : 32-bit Simple MMX (XviD)


Which iDCT you should use depends primarily on what CPU you have and to a lesser degree, on how accurate of an iDCT you desire. Most people will not be able to tell the difference in quality between these algorithms but they can be easily observed by combining the AviSynth filters Subtract and Levels. All of the available options are IEEE-1180 compliant, except for SSE/MMX (Skal).

Qualitywise: IEEE-1180 Reference > 64-bit Floating Point > Simple MMX (XviD) > Remaining iDCTs.

Speedwise: SSE2/MMX and SSE/MMX (Skal) are usually the fastest. The IEEE-1180 Reference is easily the slowest.

idct=7 is 3rd for quality but, on my machine, is the fastest with idct=2 and idct=6... I get for speed with CCE SP 2.50 about 4.0x's with idct=2/6/7 and with idct=4 about 2.2x's... and like DGDecode.html said can't see a difference.... definitely not for the speed diff

[Axlemar]

Ok, well I changed idct=7 to idct=4 in all the avs files after prepare and then encoded and it took the same time as all my other encodes do. That's the reason I was asking if I did something wrong.

[Fishman0919]

You can change the d2v file with notepad

DGIndexProjectFile06
7
92 C:\WHITE NOISE\VTS_01_1.VOB
92 C:\WHITE NOISE\VTS_01_2.VOB
92 C:\WHITE NOISE\VTS_01_3.VOB
92 C:\WHITE NOISE\VTS_01_4.VOB
92 C:\WHITE NOISE\VTS_01_5.VOB
92 C:\WHITE NOISE\VTS_01_6.VOB
92 C:\WHITE NOISE\VTS_01_7.VOB

Stream_Type=1
iDCT_Algorithm=2
YUVRGB_Scale=1 (0:TVScale 1:PCScale)
Luminance_Filter=0,0 (Gamma, Offset)
Clipping=0,0,0,0 (ClipLeft, ClipRight, ClipTop, ClipBottom)
Aspect_Ratio=16:9
Picture_Size=720x480
Field_Operation=0 (0:None 1:ForcedFILM 2:RawFrames)
Frame_Rate=23976
Location=0,0,6,17A8 iDCT_Algorithm=2

Change iDCT_Algorithm=2 to iDCT_Algorithm=4 and turn off idct=7 in dvd-rb setup

[Axlemar]

Ok thanks.

[Carpo]

so if you were to have a p4 would it be best to select options 5/6 or the one dvdrb gives u ?

[Fishman0919]

idct=6 is probably last in quality.. All of the available options are IEEE-1180 compliant, except for SSE/MMX (Skal). idct=6

idct=7 is probably the best to use, 3rd on the list for quality (if you can see a diff) and should be as fast as idct=2 or idct=5

[Carpo]

think i'll give 2 4 and 5 a go see what it comes out like on a test clip :)

seems dgdecode picks option 5 for me

[Fishman0919]

probably the same as idct=7 :D

But remember

Qualitywise: IEEE-1180 Reference > 64-bit Floating Point > Simple MMX (XviD) > Remaining iDCTs.

3,4 and 7 might be a better test

[Carpo]

i'll do the lot :)

[Axlemar]

ok. First I made sure that idct=7 was off in rebuilder. I ran prepare, opened all the d2v files, replaced iDCT_Algorithm=2
with iDCT_Algorithm=4, and ran encode. The encode still didn't take very long nor did the speed suffer. All the d2v files kept their settings after encode, and none of the avs files specified an idct. I am not sure what I could be doing wrong.

[Fishman0919]

@ Axlemar, an easy way to tell if it is work is to run dvd-rb in 3-click mode. First with idct=7 on, run prepare then run just the first one or two segments... stop dvd-rb. Go in to the dvd-rb work dir then edit the avs files of the one you just enocoded and change idct=7 to idct=4 then change the name of the files you encoded to idct7###########.m2v(you get what I mean). Then restart the encode phase all over again, select "No" when asked and dvd-rb will start the encoding part from the begining. Compare the size of the files, if they are diff size then it worked.

[Carpo]

cce sp 2.67 takes a while if you change it to idct=3 and have to use ()deen :(

mind u the results are great :)

might see how fast hc/qenc is too :P