Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > VP9 and AV1

Reply
 
Thread Tools Search this Thread Display Modes
Old 27th August 2018, 13:06   #861  |  Link
clsid
*****
 
Join Date: Feb 2005
Posts: 5,643
I previously used this for mingw build of decoder lib:
Code:
rm -rf CmakeCache.txt CMakeFiles
cmake.exe ..\aom -G "MSYS Makefiles" -DCMAKE_AR="C:\msys64\mingw64\bin\x86_64-w64-mingw32-gcc-ar.exe" -DCMAKE_TOOLCHAIN_FILE="..\aom\build\cmake\toolchains\x86_64-mingw-gcc.cmake" -DCONFIG_LOWBITDEPTH=1 -DCONFIG_AV1_ENCODER=0 -DCONFIG_LIBYUV=0 -DCONFIG_WEBM_IO=0 -DENABLE_DOCS=0 -DENABLE_EXAMPLES=0 -DENABLE_TOOLS=0 -DENABLE_TESTS=0 -DENABLE_TESTDATA=0
make
But imho it is too early to include in the player. There is not even (demo) content available yet.
clsid is offline   Reply With Quote
Old 27th August 2018, 13:43   #862  |  Link
Aleksoid1978
Registered User
 
Aleksoid1978's Avatar
 
Join Date: Apr 2008
Location: Russia, Vladivostok
Posts: 2,783
I try - but cmake show me error about avx2 and exit. Maybe it's because my CPU don't support AVX2.
__________________
AMD Ryzen 5 3600 /GIGABYTE B450 Gaming X /Patriot 32Gb@3200 /Kingston 500Gb M.2 /RTX 4060 /Samsung U28R550UQI /OLED Philips 55OLED707 /Yamaha RX-V471 + NS-555 + NS-C444 + NS-333 + YST-SW215
Aleksoid1978 is offline   Reply With Quote
Old 27th August 2018, 13:52   #863  |  Link
clsid
*****
 
Join Date: Feb 2005
Posts: 5,643
Mine doesn't either. Works here with CMake 3.10.2. Haven't tried newer versions.
clsid is offline   Reply With Quote
Old 27th August 2018, 15:55   #864  |  Link
MoSal
Registered User
 
Join Date: Jun 2013
Posts: 95
Quote:
Originally Posted by v0lt View Post
I have a problem with the rav1e encoder. The resulting IVF files play poorly in ffplay and mpv.
rav1e still uses a libaom snapshot from February. That is, the output is not yet compatible with the final AV1 specification.

Edit: this is outdated info.
__________________
https://github.com/MoSal

Last edited by MoSal; 27th August 2018 at 18:12.
MoSal is offline   Reply With Quote
Old 27th August 2018, 16:30   #865  |  Link
v0lt
Registered User
 
Join Date: Dec 2008
Posts: 1,959
Quote:
Originally Posted by MoSal View Post
rav1e still uses a libaom snapshot from February. That is, the output is not yet compatible with the final AV1 specification.
This is unexpected for me. I spent a lot of time using it.
v0lt is offline   Reply With Quote
Old 27th August 2018, 16:40   #866  |  Link
magistral
Registered User
 
Join Date: Aug 2014
Posts: 4
Quote:
Originally Posted by MoSal View Post
rav1e still uses a libaom snapshot from February. That is, the output is not yet compatible with the final AV1 specification.
rav1e has been compliant for almost a month now. See https://github.com/xiph/rav1e/pull/404
magistral is offline   Reply With Quote
Old 27th August 2018, 18:13   #867  |  Link
MoSal
Registered User
 
Join Date: Jun 2013
Posts: 95
Quote:
Originally Posted by magistral View Post
rav1e has been compliant for almost a month now. See https://github.com/xiph/rav1e/pull/404
Ouch. I was looking at the wrong branch.
__________________
https://github.com/MoSal
MoSal is offline   Reply With Quote
Old 27th August 2018, 20:08   #868  |  Link
IgorC
Registered User
 
Join Date: Apr 2004
Posts: 1,315
Quote:
Originally Posted by mzso View Post
I suppose you don't happen to have a graph on speed improvement?
No, I don't have it.

See here https://forum.doom9.org/showpost.php...&postcount=855
IgorC is offline   Reply With Quote
Old 28th August 2018, 02:31   #869  |  Link
TD-Linux
Registered User
 
Join Date: Aug 2015
Posts: 34
We have a very experimental option called "--tune psychovisual" that uses a different distortion metric. I'd be interested if anyone wants to compare. It's still relatively untested at the moment.
TD-Linux is offline   Reply With Quote
Old 28th August 2018, 18:57   #870  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,751
Quote:
Originally Posted by IgorC View Post
It's interesting the comparison is just versus x264/4 using --preset slower. Even at --preset placebo they'd still be way faster than current AV1 implementations. A fairer comparison versus x265 would look more like:

--preset placebo --subme 7 --cu-lossless --tskip --bframes 16 --no-wpp

Which would still be faster than libaom, and would make more exhaustive use of HEVC's features. At very low bitrates, maybe another 10-15% reduction in bits @ quality versus slower.

Quality @ Speed is the name of the game here, and comparisons at orders of magnitude different speed aren't that applicable to estimating real-world advantages of different bitstream formats.

(not a diss at the OP; that data might have been useful for an internal comparison for just posting to Twitter. I just want to warn against premature optimism based on Excel RD plots).
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is online now   Reply With Quote
Old 1st September 2018, 12:40   #871  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
Quote:
Originally Posted by TD-Linux View Post
We have a very experimental option called "--tune psychovisual" that uses a different distortion metric. I'd be interested if anyone wants to compare. It's still relatively untested at the moment.
Most users probably aren't routinely compiling stuff, are there (Windows) builds of Rav1e available from somebody to test with ?
mandarinka is offline   Reply With Quote
Old 1st September 2018, 15:17   #872  |  Link
Tommy Carrot
Registered User
 
Tommy Carrot's Avatar
 
Join Date: Mar 2002
Posts: 863
Quote:
Originally Posted by mandarinka View Post
Most users probably aren't routinely compiling stuff, are there (Windows) builds of Rav1e available from somebody to test with ?
Here:
Quote:
Originally Posted by TD-Linux View Post
rav1e now has automatic Windows builds via Appveyor. To get the latest build, go to this page:

https://ci.appveyor.com/project/tdaede/rav1e/history

Click the latest build labeled with "master", then click Artifacts to download the executable. I'll wire up a better interface for some at this point.
Tommy Carrot is offline   Reply With Quote
Old 5th September 2018, 02:26   #873  |  Link
hajj_3
Registered User
 
Join Date: Mar 2004
Posts: 1,120
chrome 69 was supposed to add av1 decoding support (which could be enabled using about:flags) however they removed it a week or so ago. If you were wondering why the newly released v69 didn't have it now you know why. I'm pretty sure it will be in v70, though probably disabled by default.

Last edited by hajj_3; 5th September 2018 at 23:19.
hajj_3 is offline   Reply With Quote
Old 5th September 2018, 05:01   #874  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
Quote:
Originally Posted by hajj_3 View Post
chrome 63 was supposed to add av1 decoding support (which could be enabled using about:flags) however they removed it a week or so ago. If you were wondering why the newly released v63 didn't have it now you know why. I'm pretty sure it will be in v64, though probably disabled by default.
Google release today Chrome 69.

Chrome 69 adds an AV1 decoder to Chrome Desktop stable (Windows, Mac, Linux, ChromeOS) based on the official bitstream specification. At this time, support is limited to "Main" profile 0 and does not include encoding capabilities. The supported container is ISO-BMFF (MP4). To enable this feature use the chrome://flags/#enable-av1-decoder flag.
source

At this moment it's planing be enabled by default in Chrome v70.

Last edited by marcomsousa; 5th September 2018 at 16:51.
marcomsousa is offline   Reply With Quote
Old 5th September 2018, 10:40   #875  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,565
I don't have that setting in Chrome 69.0.3497.81 (stable).
sneaker_ger is offline   Reply With Quote
Old 5th September 2018, 10:53   #876  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by hajj_3 View Post
chrome 63 was supposed to add av1 decoding support (which could be enabled using about:flags) however they removed it a week or so ago. If you were wondering why the newly released v63 didn't have it now you know why. I'm pretty sure it will be in v64, though probably disabled by default.
Chrome 63 was released last year. And 64 in January.
Quote:
Originally Posted by marcomsousa View Post
Google release today Chrome 69.

Chrome 69 adds an AV1 decoder to Chrome Desktop stable (Windows, Mac, Linux, ChromeOS) based on the official bitstream specification. At this time, support is limited to "Main" profile 0 and does not include encoding capabilities. The supported container is ISO-BMFF (MP4). To enable this feature use the chrome://flags/#enable-av1-decoder flag.

At this moment it's planing be enabled by default in Chrome v70. https://www.chromestatus.com/features/5729898442260480
So I guess the encoder will be usable enough in six weeks for youtube beta testing at least.
mzso is offline   Reply With Quote
Old 5th September 2018, 14:49   #877  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
[Rav1e tune Psychovisual vs PSNR]

Quote:
Originally Posted by TD-Linux View Post
We have a very experimental option called "--tune psychovisual" that uses a different distortion metric. I'd be interested if anyone wants to compare. It's still relatively untested at the moment.
Hmm, only a short test because I ran into time issues but it seems psychovisual tune doesn't on its own prevent the blocking/banding (and texture/noise smoothing/blurring) in areas like sky or other flat regions. But it does seem to help on some noisier/dirtier textures though I can't say it would make the whole picture strictly better. Sometimes it keeps some grain where psnr drops it all, but only sometimes, elsewhere it removes everything just as "well" as psnr. Since it does do something though, I guess it might be useful down the road?

I did not get those "super weird lines" shown in the github issue thread, but I think some similar stripe artifacts do appear in smaller degree too (some transform basis function showing?).

Here are some example images: http://imgbox.com/g/l1gKciGNMk I tested it on a bluray sample I had for other purposes, 1985 animation (remaster with some noise reduction. Chroma noise has been temporally filtered by me earlier). I don't normally deal with live action footage so I have to leave that to others. The sample is a mashup of dunno, 15 scenes, 910 frames total. Lossless encode is about 500 megabytes in case anybody wants to take a look.

I noticed that rav1e at the default speed setting had about the same performance/thread as x265 settings I used for testing before (though that was at much higher bitrate than what I received now), so I decided to use the exact same CLI except for 2-pass to make a comparison clip. So these are not x265 default settings, note.

x265_2.6+2.exe - --input-depth 8 --input-res 1440x1080 --fps 24000/1001 --preset slower --output-depth 10 --ctu 32 --max-tu-size 16 --bitrate 4200 --pass 1 --tu-intra-depth 2 --tu-inter-depth 2 --rdpenalty 2 --me 3 --subme 5 --merange 92 --amp --rect --ref 6 --weightb --weightp --keyint 300 --min-keyint 1 --bframes 8 --aq-mode 1 --aq-strength 1.0 --rd 5 --psy-rd 1.6 --psy-rdoq 8.0 --rdoq-level 1 --no-sao --no-open-gop --rc-lookahead 80 --max-merge 5 --qcomp 0.70 --no-strong-intra-smoothing --no-limit-modes --limit-refs 0 --limit-tu 0 --frame-threads 1 --no-wpp --deblock -2:-2 --qg-size 8 --pbratio 1.2 --no-cutree --cu-lossless --lookahead-slices 1 --sar 1:1 --range limited --chromaloc 0 --colormatrix bt709 --no-rskip --rd-refine --cbqpoffs -2 --crqpoffs -2 -o hevc-test.hevc
(this got 4146 kbps on second pass)

Rav1e commandlines:
rav1e.exe --speed 3 --quantizer 70 --tune Psnr dump.y4m -o psnr.ivf (got ~4200 kbps)
rav1e.exe --speed 3 --quantizer 75 --tune Psychovisual dump.y4m -o psy.ivf (got ~4270 kbps)

binary used: https://ci.appveyor.com/project/tdae....195/artifacts
mandarinka is offline   Reply With Quote
Old 5th September 2018, 14:50   #878  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
Also, may I have some suggestions? I realize that Rav1e is very early in development (I guess no more can be expected from encoders under the age of 2 years and/or supporting adaptive quantization and such important features).
However, there are some changes that I think could be useful to do even at this early stage.

1) there is a general lack of configurable settings and I heard some rumor that this might actually be somewhat a policy (like the nonexistent user-facing configurabiltiy in Theora?) and not just effect of being early in development.
This is an unfortunate position IMHO. x264/x265 have shown that configurability actually adds to both performance and quality because people can tweak settings. No defaults can be good for everybody.
In addition, the it's the configurability that allows testers to actually test the encoder. You can't even get feedback about your default internal tuning if random people can't test it against different set of parameters.
For example, with things like the psychovisual tuning you requested testing off, it would be useful to be able to test what changing the strength does (whether more is better or less is better, where's an apparent sweetspot, whether it is different depending on content etc...)

Things like aq-strenth, psychovisual bias settings, ratecontrol parameters like qcompress, these things absolutely have to be user-configurable in a serious encoder. Anything that has some quality-compression balance effect too, or just about any compression tool that is a bonus for compression but not always - or when there are cases where it hurts and people will want to disable it. All encoders/formats have tools like that (hello there SAO) and I don't think there are reasons for AV1 (rav1e) to be completely different. If you bar access to these parameters, you handicap the encoder by leaving compression performance potential locked away. You should strive to sell what you got, like x264/x265 does, not keep it in the vault.

Even if you disagree with all of the above, you should think about it form the PR point of view. I'm not the only person feeling like this, so exposing the parameters to users will make your encoder more attractive to users even if you think it's useless...

2) a smaller thing: there is very little feedback given by the commandline encoder. A FPS counter probably isn't completely important, but there is one thing that really would help with testing, and that is a achieved bitrate being reported at the end. I was rather suprised Rav1e didn't report this, because I wanted to encode a comparison clip with 2pass with another encoder and this is rather complicating. I have to figure it out based on filesize but the output's already in a container that has some unknown overhead... I used bitrate calculator like in 2006 but note that those generally only have 1 second precision, which is not very exact for short testing samples.
mandarinka is offline   Reply With Quote
Old 5th September 2018, 16:11   #879  |  Link
MoSal
Registered User
 
Join Date: Jun 2013
Posts: 95
Quote:
Originally Posted by sneaker_ger View Post
I don't have that setting in Chrome 69.0.3497.81 (stable).
Is it Chrome? or a Chromium distribution package?
__________________
https://github.com/MoSal
MoSal is offline   Reply With Quote
Old 5th September 2018, 16:44   #880  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
It seems that AV1 has disable in M69 by Google in last minute

But in official docs it's still there Source
marcomsousa is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 00:02.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.