tag	716340c56e2c7e2154598716566d9c3b3f91d435
tagger	Urvang Joshi <urvang@google.com>	Tue May 27 16:10:40 2025 -0700
object	c9e28fd96bc9c8b4ba94537fc664d220f7124dd3

Release candidate #1 for Research-v10.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-F009: No-show key frame and key overlay frame
- CWG-E255: IDIF Intra Directional Interpolation Filter Fix
- CWG-F017: Redesign of the warp mode signaling flow
- CWG-F018: CDEF Enhancements
- CWG-F008: Enabling in-loop filtering with global intra-block-copy (Intra-BC)
- CWG-F019: Simplification of luma downsampling in the CFL mode
- CWG-F052: Remove OBMC
- CWG-F055: Non-separable luma wiener filters with 9x9 support (16-tap)
- CWG-F037: CDF Scaling Improvement
- CWG-F035: Disable IBP and ORIP for 4x4 only
- CWG-F034: Multiple Reference Line Selection (MRLS) Improvement (Test3)
- CWG-F057: On Flexible Control of the Deblocking Parameter Bit Depth
- CWG-F075: TMVP Motion Vector Memory Reduction
- CWG-F007: Improved scanning order for DRL list generation
- CWG-F050: DRL and WRL Line Buffer Reduction
- CWG-F045: Improved frame context initialization
- CWG-F086: Simplify MHCCP convolve
- CWG-F072: Improved Primary Reference Frame Selection
- CWG-F021: MHCCP design for superblock boundary
- CWG-F105: Combination of CWG-F012 and CWG-F056 for IST memory reduction
- CWG-F033: Disable Inter Modes for 4x4 Partition Blocks
- CWG-F089: Delta-q Entropy Coding Optimization
- CWG-F080: Transform Partition Restriction and Encoder Speed-up
- CWG-F064: Improvements in refined MVs and EXT-WARP
- CWG-F065: Modification of DMVR search points
- CWG-F076: Simplifications for context modeling: A data-driven approach (Test 1 and Test 3)
- CWG-F084: Simplifications for intra BAWP
(Aspect 1)
- CWG-E229: Sample reduction for MHCCP model derivation
- CWG-F069: On the Maximum Number of Segments
- CWG-F115: Table size reduction for CCSO
- CWG-F044: On warp mode of AVM (disabled in LD)
- CWG-F120: On Intra Bi-Prediction
- CWG-E230: On Core Transform for AV2
- CWG-F101: On luma downsampling of cross-plane wiener restoration filter
- CWG-F092: TMVP Simplification (Test2)
- CWG-F113: Improvements for MHCCP
- CWG-F061: Symbol size reduction for wedge angle and joint shell class signaling
- CWG-F059: Re-Design of Partition Symbol Context Models
- CWG-F111: On DAMR Memory Bandwidth
- CWG-F068: Bugfix of CCSO filtering unit size for 4:2:2 and 4:4:4
- CWG-D181: On block-adaptive weighted prediction for across scale prediction (BAWP + RPR disabled by default)
- CWG-F130: Bugfix for Local IntraBC Search Range under non Common Test Condition
- CWG-E121: Enhancements of reference frame list construction for temporal scalability
- CWG-F121: Simplification for IDIF (Test 2)
- CWG-F085: Improvements to motion vector trajectory tracking
- CWG-F099: Modified deblocking filter for AVM
- CWG-F128: Context Model Optimization for Optical Flow Refinement Flag
- CWG-F096: Removal of Super Resolution Mode
- CWG-F144: On the Interaction Between Uneven Transform Partitioning and Sub-Block Deblocking
- CWG-F087: Precision Adjustment and Retraining for RESTORE_PC_WIENER
- CWG-F116: Improvements in storing MVs for TMVP list
- CWG-F051: TIP Simplification
- CWG-F146: Transform partition type replacement
- CWG-F088: MHCCP with 3 Parameters
- CWG-F078: Inter coding mode redesign and consolidation
- CWG-F127: Reduction of CDFs of MVD coding
- CWG-E254: TIP Enhancement
- CWG-F093: Improvements to Data-Driven Intra Prediction
- CWG-F043: Improved SMVP for TIP mode candidates
- CWG-F082: Extended DPB mode for AV2 in RTC
- CWG-F119: On skip mode parsing dependency
- CWG-F054: Picture boundary handling for loop filtering (HW SG recommended version)
- CWG-F110: Enhanced Transform Coding for Lossless AVM (Aspects 1 and 2 with Option 1)
- CWG-F112: Amendment to EXT_QUANT:  Add seq-level base uv ac offset and misc flags
- CWG-E206: Guided Detail Filter - r3
- CWG-E127: BRU: Backward Reference Updating for Ultra Low Latency Video Coding

Other Normative Changes
-----------------------
- Enable virtual line buffer for cross-component non-separable wiener filter
- CWG-E066: Interleaved parsing at 64x64 level with SDP in key frames (missing part from previous cycle)

Encoder-only Improvements (Non-normative)
-----------------------------------------
- CWG-F002: Intra-BC decision bug fix for natural content
- CWG-E251: An encoder side improvement for extended SDP
- Disable frame averaging and enable tile averaging for context initialization when the number of tiles is 2.
- Fix bug disabling PC_WIENER in search_pc_wiener_visitor.
- CWG-F136: TCQ Improvements
- CWG-F163: CTC for AVM resize-mode

Encoder / Decoder Speedups (Non-normative)
------------------------------------------
- SIMD for DIP
- Speed-up decoder by optimizing refinemv_highbd_pad_mc_border function
- CWG-F049: AVM v9 encoder and decoder optimizations
- CWG-F073: Improved model RD based transform pruning of inter modes
- CWG-F118: Inter mode evaluation optimization
- CWG-F011: Simplification of Palette Mode (Aspect 1)
- CWG-F046: Encoder speed up using ML-based Partition Pruning, additional speedup

Code Cleanups
-------------
- Suppress TFLite compile time warnings

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/13#tab-issues

commit	c9e28fd96bc9c8b4ba94537fc664d220f7124dd3	[log] [tgz]
author	Urvang Joshi <urvang@google.com>	Mon May 19 19:25:33 2025 -0700
committer	Leo (Liang) Zhao <leolzhao@global.tencent.com>	Tue May 20 21:24:15 2025 +0000
tree	85c58e5a9ca32bbf58e209ad0fc9ddce9a6501b0
parent	a4989671b30140b7301f8ec772be1128deeab8b8 [diff]

ForwardKeyTestLarge: tweak PSNR threshold

Earlier threshold (24.5) was too tight, and slight deviations (e.g. 24.48) were causing it to fail.

So, decrease the threshold and keep some margin.

test/fwd_kf_test.cc[diff]

1 file changed

tree: 85c58e5a9ca32bbf58e209ad0fc9ddce9a6501b0

README.md

AVM

Building

Prerequisites

CMake version 3.16 or higher.
Git.
Perl.
For x86 targets, yasm, which is preferred, or a recent version of nasm. If you download yasm with the intention to work with Visual Studio, please download win32.exe or win64.exe and rename it into yasm.exe. DO NOT download or use vsyasm.exe.
Building the documentation requires doxygen version 1.8.10 or newer.
Building the unit tests requires Python.

Get the code

The AVM project source code is stored in the Alliance of Open Media’s GitLab repository. To get the code,

    $ git clone https://gitlab.com/AOMediaCodec/avm.git
    # By default, the above command stores the source in the avm directory:
    $ cd avm

Basic build

CMake replaces the configure step typical of many projects. Running CMake will produce configuration and build files for the currently selected CMake generator. For most systems the default generator is Unix Makefiles. The basic form of a makefile build is the following:

    $ cmake path/to/avm
    $ make

The above will generate a makefile build that produces the AVM library and applications for the current host system after the make step completes successfully. The compiler chosen varies by host platform, but a general rule applies: On systems where cc and c++ are present in $PATH at the time CMake is run the generated build will use cc and c++ by default.

Configuration options

The AVM codec library has a great many configuration options. These come in two varieties:

Build system configuration options. These have the form ENABLE_FEATURE.
AVM codec configuration options. These have the form CONFIG_FEATURE.

Both types of options are set at the time CMake is run. The following example enables ccache and disables the AVM encoder:

    $ cmake path/to/avm -DENABLE_CCACHE=1 -DCONFIG_MULTITHREAD=0
    $ make

The available configuration options are too numerous to list here. Build system configuration options can be found at the top of the CMakeLists.txt file found in the root of the AVM repository, and AVM codec configuration options can currently be found in the file build/cmake/aom_config_defaults.cmake.

Dylib builds

A dylib (shared object) build of the AVM codec library can be enabled via the CMake built in variable BUILD_SHARED_LIBS:

    $ cmake path/to/avm -DBUILD_SHARED_LIBS=1
    $ make

This is currently only supported on non-Windows targets.

Debugging

Depending on the generator used there are multiple ways of going about debugging AVM components. For single configuration generators like the Unix Makefiles generator, setting CMAKE_BUILD_TYPE to Debug is sufficient:

    $ cmake path/to/avm -DCMAKE_BUILD_TYPE=Debug

For Xcode, mainly because configuration controls for Xcode builds are buried two configuration windows deep and must be set for each subproject within the Xcode IDE individually, CMAKE_CONFIGURATION_TYPES should be set to Debug:

    $ cmake path/to/avm -G Xcode -DCMAKE_CONFIGURATION_TYPES=Debug

For Visual Studio the in-IDE configuration controls should be used. Simply set the IDE project configuration to Debug to allow for stepping through the code.

In addition to the above it can sometimes be useful to debug only C and C++ code. To disable all assembly code and intrinsics set AOM_TARGET_CPU to generic at generation time:

    $ cmake path/to/avm -DAOM_TARGET_CPU=generic

Cross compiling

For the purposes of building the AVM codec and applications, relative to the scope of this guide, all builds for architectures differing from the native host architecture will be considered cross compiles. The AVM CMake build handles cross compiling via the use of toolchain files included in the AVM repository. The available toolchain files can be found at cmake folder in the AVM repository. The following example demonstrates use of the x86-linux.cmake toolchain file on a x86_64 linux host:

    $ cmake path/to/avm \
      -DCMAKE_TOOLCHAIN_FILE=path/to/avm/build/cmake/toolchains/x86-linux.cmake
    $ make

Sanitizers

Sanitizer integration is built-in to the CMake build system. To enable a sanitizer, add -DSANITIZE=<type> to the CMake command line. For example, to enable address sanitizer:

    $ cmake path/to/avm -DSANITIZE=address
    $ make

Sanitizers available vary by platform, target, and compiler. Consult your compiler documentation to determine which, if any, are available.

Microsoft Visual Studio builds

Building the AVM codec library in Microsoft Visual Studio is supported. Visual Studio 2019 (16.7) or later is required. The following example demonstrates generating projects and a solution for the Microsoft IDE:

    # This does not require a bash shell; Command Prompt (cmd.exe) is fine.
    # This assumes the build host is a Windows x64 computer.

    # To build with Visual Studio 2019 for the x64 target:
    $ cmake path/to/avm -G "Visual Studio 16 2019"
    $ cmake --build .

    # To build with Visual Studio 2019 for the 32-bit x86 target:
    $ cmake path/to/avm -G "Visual Studio 16 2019" -A Win32
    $ cmake --build .

NOTE: The build system targets Windows 7 or later by compiling files with -D_WIN32_WINNT=0x0601.

Xcode builds

Building the AVM codec library in Xcode is supported. The following example demonstrates generating an Xcode project:

    $ cmake path/to/avm -G Xcode

Build with VMAF support

After installing libvmaf.a, you can use it with the encoder:

    $ cmake path/to/avm -DCONFIG_TUNE_VMAF=1

Please note that the default VMAF model will be used unless you set the following flag when running the encoder:

    # --vmaf-model-path=path/to/model

Testing

Testing basics

There are several methods of testing the AVM codec. All of these methods require the presence of the AVM source code and a working build of the AVM library and applications.

1. Unit tests:

The unit tests can be run at build time:

    # Before running the make command the LIBAOM_TEST_DATA_PATH environment
    # variable should be set to avoid downloading the test files to the
    # cmake build configuration directory.
    $ cmake path/to/avm
    # Note: The AVM CMake build creates many test targets. Running make
    # with multiple jobs will speed up the test run significantly.
    $ make runtests

2. Example tests:

The example tests require a bash shell and can be run in the following manner:

    # See the note above about LIBAOM_TEST_DATA_PATH above.
    $ cmake path/to/avm
    $ make
    # It's best to build the testdata target using many make jobs.
    # Running it like this will verify and download (if necessary)
    # one at a time, which takes a while.
    $ make testdata
    $ path/to/avm/test/examples.sh --bin-path examples

Downloading the test data

The fastest and easiest way to obtain the test data is to use CMake to generate a build using the Unix Makefiles generator, and then to build only the testdata rule:

    $ cmake path/to/avm -G "Unix Makefiles"
    # 28 is used because there are 28 test files as of this writing.
    $ make -j28 testdata

The above make command will only download and verify the test data.

Additional input data for testing the encoder can be obtained from: AV2 - CTC

Sharded testing

The AVM codec library unit tests are built upon gtest which supports sharding of test jobs. Sharded tests can be achieved as follows for example:

   # Set the environment variable GTEST_TOTAL_SHARDS to control the number of
   # shards.
   $ export GTEST_TOTAL_SHARDS=10
   # (GTEST shard indexing is 0 based).
   $ seq 0 $(( $GTEST_TOTAL_SHARDS - 1 )) \
       | xargs -n 1 -P 0 -I{} env GTEST_SHARD_INDEX={} ./test_libaom

To create a test shard for each CPU core available on the current system set GTEST_TOTAL_SHARDS to the number of CPU cores on your system minus one. The maximum number of test targets that can run concurrently is determined by the number of CPUs on the system where the build is configured as detected by CMake. A system with 24 cores can run 24 test shards using a value of 24 with the -j parameter. When CMake is unable to detect the number of cores 10 shards is the default maximum value.

Coding style

We are using the Google C Coding Style defined by the Google C++ Style Guide.

The coding style used by this project is enforced with clang-format and cmake-format.

clang-format can be installed with your system's package manager, or directly from llvm.org. For best results, your version should match the one used by Gitlab CI, noted as a comment in .clang-format.
cmake-format can be obtained by installing cmakelang. Again, for best results, your version should match the one used by Gitlab CI, noted as a comment in .cmake-format.py

Before pushing changes for review, format your code using the tools above.

We recommend automating the formatting by adding a pre-commit hook at .git/hooks/pre-commit. An example pre-commit hook is provided below:

#!/bin/bash
# Replace 'VERSION' below to match clang-format version in the CI.
CLANG_FORMAT_PATH=/usr/lib/llvm-VERSION/bin
CMAKE_FORMAT_PATH=${HOME}/.local/bin/

echo "Applying clang-format ..."
for file in $(git diff-index --cached --name-only HEAD -- "*.[hc]pp" "*.cc" "*.[ch]") ; do
  ${CLANG_FORMAT_PATH}/clang-format -i --style=file ${file}
  git add ${file}
  echo "Formatted file: $file"
done
echo "Done."

echo "Applying cmake-format ..."
for file in $(git diff-index --cached --name-only HEAD -- '*.cmake' CMakeLists.txt) ; do
  ${CMAKE_FORMAT_PATH}/cmake-format -i ${file}
  git add ${file}
  echo "Formatted file: $file"
done
echo "Done."

Submitting patches

We manage the submission of patches using Gitlab's merge request process. This tool implements a workflow on top of the Git version control system to ensure that all changes get peer reviewed and tested prior to their distribution. If you are not able to submit an MR, please contact SW coordinators to make sure necessary contributor agreements are signed for the AOMedia Project.

Follow the one-time set-up steps as detailed here.
For pushing your code modifications, follow the steps detailed here.
Once the code is pushed into a branch, create a merge request as detailed here.
- The code review, approval and CI/CD process will be initiated after the MR is created.
- Once the MR is approved, the software co-ordinators will merge the MR to the main branch.

Follow the Merge request page to check the status of the changes, review comments etc.

Support

This library is an open source project supported by its community. Please email https://aomedia.org/contact/ for help.

Bug reports

Bug reports can be filed in the Alliance for Open Media Gitlab issue tracker.