L-SMASH Source

[stax76]

70% CPU usage according to avsmeter using a 4k hevc sample.

[raffriff42]

Originally Posted by stax76

70% CPU usage according to avsmeter using a 4k hevc sample.

This might be why:https://ffmpeg.org/ffmpeg-all.html

Note that most acceleration methods are intended for playback and will not be faster than software decoding on modern CPUs.
Additionally, ffmpeg will usually need to copy the decoded frames from the GPU memory into the system memory, resulting in further performance loss.
This option is thus mainly useful for testing.

[dipje]

although you must not look at the cpu usage for stuff like that (because a lot of it might just be 'waiting for response from gpu' in which time other filters might happily run) it's a bit hit or miss here as well.

'h264' vs 'h264_cuvid' on my gtx1060, pure decoding of 1920x1080 h264 4:2:0 8bit, vapoursynth benchmark -> +/- 125fps vs 190fps. If I resample the file to 4:4:4: 16bit, then run f3kdb on it with a 'dither down back to 10bit' I get 35.2fps vs 39.8 fps. So a boost, but a small one.

Then, a 1920x1080 HEVC 4:2:2 10bit file, 'hevc' vs 'h265_cuvid'. Software decoding reaches 19.7 fps, then if I switch to 'hevc_cuvid' I get around 4.5 fps... which keeps on dropping bit by bit till it goes less than 1 fps then Vapoursynth crashes. So something is not right there :P. (LAVFilters refuses to use cuvid for the same file btw, so who knows they know it shouldn't work at all?)

[sneaker_ger]

NVDECODE 10 bit needs the not yet released Video Codec SDK 8.0.

(Or has it been?)

[stax76]

Originally Posted by Selur

for me it's dxva2 and cuvid

dxva2 and qsv for me

OS          : Windows 10 Pro
CPU         : Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
GPU         : NVIDIA GeForce GTX 1060 6GB, Intel(R) HD Graphics 530

[poisondeathray]

"h264_cuvid" works for me, much lower CPU usage, and it translates to faster actual encoding (tested with no filters, just decoding/encoding).

But issues with frames out of order, stuck frames (duplicates), etc.. ie. unreliable. I have similar experiences with full GPU encode/decode with ffmpeg or rigaya's nvencc also - the issue is the decoding part for all of them - because replacing with sw decoding works. lsmash "h264_cuvid" seems completely unreliable in terms of 1:1 frame consistency. It doesn't matter if threads=1 or indexed.

Even DirectShowSource() with a configured GPU decoder through lav or .grf is more reliable. That is saying a lot (!)

I couldn't get "h264_qsv" to work, it looks like software decoding speed and cpu usage (it even shows correct frames ). This was on r929 from the_weirdo

[B.F.]

Is there any way to get fpsnum/fpsden commands working?
No matter what I tryed, this settings just ignored.

[sneaker_ger]

It should "just work". Post a sample of your source and your complete script.

[B.F.]

https://www.dropbox.com/s/pigfbxmkaegc0ew/321.mkv?dl=0
LWLibavVideoSource("321.mkv", fpsnum = 24000, fpsden = 1001)
It works on other source files, but not on this one.

[poisondeathray]

Originally Posted by B.F.

https://www.dropbox.com/s/pigfbxmkaegc0ew/321.mkv?dl=0
LWLibavVideoSource("321.mkv", fpsnum = 24000, fpsden = 1001)
It works on other source files, but not on this one.

interesting that fpsnum , fpsden work with ffms2 but not lsmash on that one

[sneaker_ger]

Mkvextract says first video timecode at 10003ms, no audio timecodes. I'd say the file can be considered broken. If you created it yourself I'd open a ticket on the respective bug tracker (ffmpeg?).

[B.F.]

No, thats not my file, thats small piece of one of the files I need to encode.
And there is a lot of files like that in the internets.
And if ffms2 works on that one, L-SMASH should work too.

[poisondeathray]

Several lsmash versions (I didn't check very old versions), both avisynth and vpy, crash when opening ULRG (UT Video Codec RGB24), but ffms2 is able to open no problems . I know there are differences between original VFW and ffmpeg/libav versions in terms of decoders, but not sure about encoder. Eitherway, it doesn't matter if it was written by the ffmpeg/libav implementation, or with the VFW version they all induce crash

EDIT: probably don't need a sample but just in case

colorbars()
Trim(0,1)
showframenumber()
converttorgb24()

Both ffmpeg and vfw encoder versions included
https://www.mediafire.com/?914801rwjq07k22

[Groucho2004]

Originally Posted by poisondeathray

Several lsmash versions (I didn't check very old versions), both avisynth and vpy, crash when opening ULRG (UT Video Codec RGB24), but ffms2 is able to open no problems .

AVISource() also has no problems with this. I always use it for UTVideo.

[poisondeathray]

Originally Posted by Groucho2004

AVISource() also has no problems with this. I always use it for UTVideo.

Same here, the VFW version has the fastest decoder > libavcodec version . The later is about 2/3 speed and apparently doesn't have all the optimizations yet . But that doesn't help , say linux or mac users using vpy . But if ffms2 can do it, lsmash should be able to I woudl think

[speedyrazor]

I have been using LibavSMASHSource in VapourSynth 64 for a while, very successfully. I am now trying to use this on the below XDCAM HD 422 file, but I get an error, any suggestions please:

General
Complete name : M:\797.mov
Format : QuickTime
Format/Info : Original Apple specifications
Commercial name : XDCAM HD422
File size : 22.8 GiB
Duration : 59 min 0 s
Overall bit rate : 55.4 Mb/s
Encoded date : UTC 2014-02-15 02:48:42
Tagged date : UTC 2014-02-15 02:48:42
Writing application : Omneon OmMedia.dll 6.4.0.5 02-27-2012 18:28:44,ex={0,-1},rng={0,-1,0},trimAu,exPre

Video
ID : 1
Format : MPEG Video
Commercial name : XDCAM HD422
Format version : Version 2
Format profile : 4:2:2@High
Format settings, BVOP : Yes
Format settings, Matrix : Custom
Format settings, GOP : M=3, N=12
Format settings, picture structure : Frame
Codec ID : xd5c
Duration : 59 min 0 s
Bit rate mode : Constant
Bit rate : 50.0 Mb/s
Width : 1 920 pixels
Clean aperture width : 1 888 pixels
Height : 1 080 pixels
Clean aperture height : 1 062 pixels
Display aspect ratio : 16:9
Clean aperture display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 25.000 FPS
Color space : YUV
Chroma subsampling : 4:2:2
Bit depth : 8 bits
Scan type : Interlaced
Scan order : Top Field First
Compression mode : Lossy
Bits/(Pixel*Frame) : 0.965
Time code of first frame : 00:00:00:00
GOP, Open/Closed : Closed
Stream size : 20.6 GiB (90%)
Language : English
Encoded date : UTC 2014-02-15 03:30:44
Tagged date : UTC 2014-02-15 03:30:44
Source :

Audio #1
ID : 2
Format : PCM
Format settings, Endianness : Little
Format settings, Sign : Signed
Muxing mode : Wave
Codec ID : sowt / 1
Duration : 59 min 0 s
Bit rate mode : Constant
Bit rate : 1 536 kb/s
Channel(s) : 2 channels
Sampling rate : 48.0 kHz
Bit depth : 16 bits
Stream size : 648 MiB (3%)
Language : English
Encoded date : UTC 2014-02-15 03:30:44
Tagged date : UTC 2014-02-15 03:30:44
Source :

Audio #2
ID : 3
Format : PCM
Format settings, Endianness : Little
Format settings, Sign : Signed
Muxing mode : Wave
Codec ID : sowt / 1
Duration : 59 min 0 s
Bit rate mode : Constant
Bit rate : 1 536 kb/s
Channel(s) : 2 channels
Sampling rate : 48.0 kHz
Bit depth : 16 bits
Stream size : 648 MiB (3%)
Language : English
Encoded date : UTC 2014-02-15 03:30:44
Tagged date : UTC 2014-02-15 03:30:44
Source :

Audio #3
ID : 4
Format : PCM
Format settings, Endianness : Little
Format settings, Sign : Signed
Muxing mode : Wave
Codec ID : in24 / 1
Duration : 59 min 0 s
Bit rate mode : Constant
Bit rate : 2 304 kb/s
Channel(s) : 2 channels
Sampling rate : 48.0 kHz
Bit depth : 24 bits
Stream size : 972 MiB (4%)
Language : English
Encoded date : UTC 2014-02-15 03:30:44
Tagged date : UTC 2014-02-15 03:30:44
Source :

Other
ID : 5
Type : Time code
Format : QuickTime TC
Duration : 59 min 0 s
Time code of first frame : 00:00:00:00
Time code, striped : Yes
Language : English
Encoded date : UTC 2014-02-15 03:30:44
Tagged date : UTC 2014-02-15 03:30:44

[poisondeathray]

Originally Posted by speedyrazor

I have been using LibavSMASHSource in VapourSynth 64 for a while, very successfully. I am now trying to use this on the below XDCAM HD 422 file, but I get an error, any suggestions please:

What error specifically ? Was there an message ?

I noticed the MOV container; XDCAM HD422 in MXF container works ok in 64bit vpy with lsmash - could it be container related issue ? Maybe try re-wrapping it

[StainlessS]

SpeedyRazor,

I have been using LibavSMASHSource

Would you like to clarify, LSMASHVideoSource or LWLibavVideoSource.

[poisondeathray]

Originally Posted by StainlessS

SpeedyRazor,

Would you like to clarify, LSMASHVideoSource or LWLibavVideoSource.

vapoursynth version is slightly different than avisynth version

LibavSMASHSource, and LWLibavSource

[StainlessS]

A rose by any other name would smell as sweat.

Thanx PDR, was not expecting that.