Intel SVT-AV1

[nakTT]

Version 2.2.1

Changelog

[2.2.1] - 2024-08-23

Cleanup and bug fixes and documentation

- Fixed debug arm macOS builds by removing invalid assert

[ShortKatz]

Version 2.3.0

Versionshinweise
[2.3.0] - 2024-10-28

API updates

Preset shift: M12/M13 mapped to M11, M7-M11 shifted one position down. API does not change, all presets from MR-M13 will still be accepted
svt_av1_enc_get_packet API is now a blocking call for low-delay enforcing a picture in, picture out model
--fast-decode range changed from 0-1 to 0-2 available on all presets
Introducing a new definition of --lp being levels of parallelism with a new field in the config structure level_of_parallelism
logical_processors will be deprecated in the 3.0 release
Encoder

NEW FAST DECODE MODE - (!2280)
New fast-decode (2) to allow for an average AV1 software cycle reduction of 25-50% vs fast-decode 0 with a 1-3% BD-Rate loss across the presets
Improved fast-decode (1) option to increase its AV1 software cycle reduction by ~10% while maintaining the same quality levels
Improved --lp settings for high resolutions, with CRF gaining a ~4% improvement in speed and VBR gaining ~15% (!2323)
Further Arm-based optimizations improving the efficiency of previously written Arm-neon implementations by an average of 30%. See below for more information on specific presets
Address speed regressions for high resolutions first pass encode by tuning the threading parameters, with 1080p showing the biggest gains
Enabled AVX512 by default in cmake allowing for ~2-4% speedup
Enabled LTO by default if using a new enough compiler (!2288, !2305)
If LTO is a problem or causes one, it can be disabled by adding -DSVT_AV1_LTO=OFF to cmake to force it off.
Please report any issues that occur when using it.
Cleanup Build and bug fixes and documentation

third_party: update safestringlib with applicable upstream changes
Improved the unit test coverage for Arm-neon code
Updated documentation
Arm Improvements

Speed comparison was done against v2.2 on AWS Graviton4 instances with Clang 19.1.1
--lp 1 was used for all tests

https://gitlab.com/AOMediaCodec/SVT-AV1/-/releases

[ShortKatz]

SVT-AV1 now has support for lossless encoding.

[ShortKatz]

This merge request has a comment about a AV1 spec bug.
https://gitlab.com/AOMediaCodec/SVT-...873cb8b_80_101

And a link to this bug https://aomedia.issues.chromium.org/issues/42302342

But I quite don't understand what this bug actually means for AV1 encoding. Will this interfere with encoding quality? Or performance?

[nevcairiel]

Originally Posted by ShortKatz

This merge request has a comment about a AV1 spec bug.
https://gitlab.com/AOMediaCodec/SVT-...873cb8b_80_101

And a link to this bug https://aomedia.issues.chromium.org/issues/42302342

But I quite don't understand what this bug actually means for AV1 encoding. Will this interfere with encoding quality? Or performance?

It doesnt cause anything, other then headaches for developers to obey the wrong-way-around values.

[oibaf]

Originally Posted by ShortKatz

This merge request has a comment about a AV1 spec bug.
https://gitlab.com/AOMediaCodec/SVT-...873cb8b_80_101

And a link to this bug https://aomedia.issues.chromium.org/issues/42302342


But I quite don't understand what this bug actually means for AV1 encoding. Will this interfere with encoding quality? Or performance?

According to the bug link it should impact BDRATE, so the quality, however "change is ~neutral overall":

* Change the encoder logic so that we never select a TRANSLATION
type global motion model. If the global motion search process
produces a TRANSLATION type model, we reset to an IDENTITY
type model.

The impact of this change is ~neutral overall, with no video
changing by more than +/- 0.2% BDRATE.

Overall impact on 150 frames, speed 1, const q mode:

lowres2: -0.002% BDRATE, -0.003% instr count
midres2: +0.005% BDRATE, +0.005% instr count
hdres2: -0.001% BDRATE, -0.035% instr count

[ShortKatz]

OK, thanks. :-)

[ShortKatz]

Interesting, it seems some SVT-AV1-PSY features are getting backported into SVT-AV1 mainline.

[oibaf]

Originally Posted by ShortKatz

Interesting, it seems some SVT-AV1-PSY features are getting backported into SVT-AV1 mainline.

Nice, here is the link to the MRs: https://gitlab.com/AOMediaCodec/SVT-...ch=SVT-AV1-PSY

[Z2697]

Originally Posted by ShortKatz

Interesting, it seems some SVT-AV1-PSY features are getting backported into SVT-AV1 mainline.

Now they shifted their focus to getting more -psy fork features merged into mainline.

[Kurt.noise]

Hi,

I'm experiencing an issue using svt-av1-psy with any different source files : on windows platform (windows 11 24H2), it seems encoding to 8 bits does not ending correctly within preset < 8 :

svt-av1-psy 2.3.0-A 8 bits :

E:\tools\svtav1psy>SvtAv1EncApp.exe -i "E:\Video\Raw.yuv" -w 1920 -h 1080 --input-depth 8 -n 1200 --fps-num 30000 --fps-denom 1001 --progress 3 --crf 30 --preset 4 -b "E:\Video\output.ivf"
Svt[info]: -------------------------------------------
Svt[info]: SVT [version]:       SVT-AV1-PSY Encoder Lib v2.3.0-A-5-gf560b40
Svt[info]: SVT [build]  :       Clang 18.1.8     64 bit
Svt[info]: LIB Build date: Dec 30 2024 07:49:20
Svt[info]: -------------------------------------------
Svt[info]: Level of Parallelism: 5
Svt[info]: Number of PPCS 140
Svt[info]: [asm level on system : up to avx2]
Svt[info]: [asm level selected : up to avx2]
Svt[info]: -------------------------------------------
Svt[info]: SVT [config]: main profile   tier (auto)     level (auto)
Svt[info]: SVT [config]: width / height / fps numerator / fps denominator               : 1920 / 1080 / 30000 / 1001
Svt[info]: SVT [config]: bit-depth / color format                                       : 8 / YUV420
Svt[info]: SVT [config]: preset / tune / pred struct                                    : 4 / SSIM / random access
Svt[info]: SVT [config]: gop size / mini-gop size / key-frame type                      : 321 / 32 / key frame
Svt[info]: SVT [config]: BRC mode / rate factor                                         : CRF / 30.00
Svt[info]: SVT [config]: AQ mode / variance boost strength / octile / curve             : 2 / 2 / 6 / regular
Svt[info]: SVT [config]: Sharpness / QP scale compress strength / Frame low-luma bias   : 1 / 1 / 0
Svt[info]: SVT [config]: Temporal Filtering Strength                                    : 1
Svt[info]: SVT [config]: Keyframe TF Strength                                           : 1
Svt[info]: -------------------------------------------
Encoding:    1/1200 Frames @ 23.60 fpm | 380.05 kb/s | Time: 0:00:03 [-0:00:19] | Size: 0.22 MB [1.90 MB]

svt-av1-psy 2.3.0-A 10 bits:

E:\tools\svtav1psy>SvtAv1EncApp.exe -i "E:\Video\Raw.yuv" -w 1920 -h 1080 --input-depth 10 -n 1200 -
-fps-num 30000 --fps-denom 1001 --progress 3 --crf 30 --preset 4 -b "E:\Video\output.ivf"
Svt[info]: -------------------------------------------
Svt[info]: SVT [version]:       SVT-AV1-PSY Encoder Lib v2.3.0-A-5-gf560b40
Svt[info]: SVT [build]  :       Clang 18.1.8     64 bit
Svt[info]: LIB Build date: Dec 30 2024 07:49:20
Svt[info]: -------------------------------------------
Svt[info]: Level of Parallelism: 5
Svt[info]: Number of PPCS 140
Svt[info]: [asm level on system : up to avx2]
Svt[info]: [asm level selected : up to avx2]
Svt[info]: -------------------------------------------
Svt[info]: SVT [config]: main profile   tier (auto)     level (auto)
Svt[info]: SVT [config]: width / height / fps numerator / fps denominator               : 1920 / 1080 / 30000 / 1001
Svt[info]: SVT [config]: bit-depth / color format                                       : 10 / YUV420
Svt[info]: SVT [config]: preset / tune / pred struct                                    : 4 / SSIM / random access
Svt[info]: SVT [config]: gop size / mini-gop size / key-frame type                      : 321 / 32 / key frame
Svt[info]: SVT [config]: BRC mode / rate factor                                         : CRF / 30.00
Svt[info]: SVT [config]: AQ mode / variance boost strength / octile / curve             : 2 / 2 / 6 / regular
Svt[info]: SVT [config]: Sharpness / QP scale compress strength / Frame low-luma bias   : 1 / 1 / 0
Svt[info]: SVT [config]: Temporal Filtering Strength                                    : 1
Svt[info]: SVT [config]: Keyframe TF Strength                                           : 1
Svt[info]: -------------------------------------------
Encoding: 1200/1200 Frames @ 2.48 fps | 51442.81 kb/s | Time: 0:08:04 [-0:00:00] | Size: 257.47 MB [257.47 MB]
SUMMARY --------------------------------- Channel 1  --------------------------------
Total Frames            Frame Rate              Byte Count              Bitrate
        1200            29.97 fps                257471283              51442.81 kbps


Channel 1
Average Speed:          2.478 fps
Total Encoding Time:    484198 ms
Total Execution Time:   484534 ms
Average Latency:        58361 ms
Max Latency:            96320 ms

For reference, testing also the current svt-av1 2.3.0 8 bits works fine :

E:\tools\svtav1psy>SvtAv1EncApp.exe -i "E:\Video\Raw.yuv" -w 1920 -h 1080 --input-depth 8 -n 1200 --fps-num 30000 --fps-denom 1001 --progress 3 --crf 30 --preset 4 -b "E:\Video\output.ivf"
Svt[info]: -------------------------------------------
Svt[info]: SVT [version]:       SVT-AV1 Encoder Lib v2.3.0
Svt[info]: SVT [build]  :       Visual Studio 2022 64 bit
Svt[info]: LIB Build date: Jan  5 2025 09:41:32
Svt[info]: -------------------------------------------
Svt[info]: Level of Parallelism: 5
Svt[info]: Number of PPCS 140
Svt[info]: [asm level on system : up to avx2]
Svt[info]: [asm level selected : up to avx2]
Svt[info]: -------------------------------------------
Svt[info]: SVT [config]: main profile   tier (auto)     level (auto)
Svt[info]: SVT [config]: width / height / fps numerator / fps denominator               : 1920 / 1080 / 30000 / 1001
Svt[info]: SVT [config]: bit-depth / color format                                       : 8 / YUV420
Svt[info]: SVT [config]: preset / tune / pred struct                                    : 4 / PSNR / random access
Svt[info]: SVT [config]: gop size / mini-gop size / key-frame type                      : 161 / 32 / key frame
Svt[info]: SVT [config]: BRC mode / rate factor                                         : CRF / 30
Svt[info]: SVT [config]: AQ mode / variance boost                                       : 2 / 0
Svt[info]: -------------------------------------------
Encoding      1200
SUMMARY --------------------------------- Channel 1  --------------------------------
Total Frames            Frame Rate              Byte Count              Bitrate
        1200            29.97 fps                 11649088              2327.49 kbps


Channel 1
Average Speed:          20.573 fps
Total Encoding Time:    58328 ms
Total Execution Time:   58618 ms
Average Latency:        6539 ms
Max Latency:            8242 ms

Does anybody also experiment this ?

Same story using also this library through ffmpeg...

[GeoffreyA]

Yes, I had a similar problem with the latest version of SVT-AV1-PSY when using a lower preset. The encoding starts, then ends prematurely. Likely, something got broken recently.

[oibaf]

Check if your issue is already reported at https://github.com/psy-ex/svt-av1-psy/issues (maybe this: https://github.com/psy-ex/svt-av1-psy/issues/115), otherwise open a new issue so that they can fix it.

[Kurt.noise]

Thanks for confirmation. Yeah, it looks like issue #115, as already being reported.

And btw, debugging more deeper : issue occured between 2.3.0 & current releases.

[GeoffreyA]

Same for me. It was the recent A release. Before that, never once crashed.

[oibaf]

Please discuss it at: https://github.com/psy-ex/svt-av1-psy/issues/115 where their developers are working. Posting the details here won't help to fix it...

[GeoffreyA]

Sure.

[Kurt.noise]

A workaround has been found : https://github.com/psy-ex/svt-av1-psy/discussions/118

[MZ/X]

A new AV1 vfw encoder/decoder codec (xAV1vfw) is released on http://mpxplay.sourceforge.net with svt-av1 v2.3.0
(it can be used with VirtualDub2 for example)
You can follow the updates here too:
https://www.facebook.com/mpxplay

[BlueSwordM]

Originally Posted by Kurt.noise

A workaround has been found : https://github.com/psy-ex/svt-av1-psy/discussions/118

It's been fully fixed now.