Weekly pre-release

Improvements

• Add a quiet (-q) option to the CLI (@lu-zero)
• Optimize SATD computation at high bit depth for around 1% performance improvement (@shssoichiro)

Changes

• Move the Frame data structure to a separate crate: v_frame (@lu-zero)
• Remove the unused speed-test CLI option (@shssoichiro)

Fixes

• Fix encoding configuration serialization (@lu-zero)

v0.3.1

This is a patch release correcting some balancing issues in speed/complexity tradeoffs.

Improvements

• 25-40% faster speed levels 2 to 5
  • This is accomplished by disabling costly coding tools
    • Fine directional prediction
    • Intra block transform splitting in inter frames
  • Encoding quality is slightly inferior (1-2%), but more in line with the target speed levels

v0.3.0: FOSDEM Edition

Improvements

• Faster and better Speed 10
• Smaller binaries (Around 3MB stripped on x86_64/Linux)
• Faster build times (About 14% faster build time)
• Multi-threaded deblocking filter
• Additional x86_64 SIMD code
• More auto-vectorizable codepaths and bounds check elisions
• ⅙ less memory allocations
• Improvements on the intra-modes pruning logic in the RDO
• More float to fixed-point conversions
• Make an early-exit condition in RDO faster
• Simplify logic in Counter/Recorder store method
• Support wasm32-wasi as build target
• 1-2% overall quality improvement (at speed level 2)

New tools

• Intra edge filter
• Switch frame support (use -S, --switch-frame-interval <SWITCH_FRAME_INTERVAL> to enable it from the CLI)
• Fine directional intra prediction
• Still Picture support with reduced still picture header (AVIF)

Changes

• Upstream nasm-rs 0.1.7 can be used to build rav1e.
• The C header produced is now C++-compatible

Fixes

#1930
#2055

Known issues

#2104

Weekly pre-release

Improvements

• speed 10 is about a 15% faster compared to the last pre-release
• Improvements on the intra-modes pruning logic in the RDO
• More float to fixed-point conversions
• Make an early-exit condition in RDO faster
• Simplify logic in Counter/Recorder store method
• Still picture mode now encodes valid bitstreams (note: the reduced still picture header is not yet enabled)

New tools

• Intra edge filter (native mode only, about 4% slower at speed level 2 for 0.5% enhanced compression)

Changes

• Upstream nasm-rs 0.1.7 can be used to build rav1e.

Weekly pre-release

Improvements

• speed 10 is overall 30% faster and the quality is slightly improved

• The quality is improved for all the other speed levels, but with larger encoding times e.g.:
  speed 4 is 5% slower

• About 50% smaller binaries

• About 14% faster build time

New feature

• Switch frame support (use -S, --switch-frame-interval <SWITCH_FRAME_INTERVAL> to enable it from the CLI)

Changes

• The C header produced is now C++-compatible

v0.2.1

Changes

• About 30% smaller binaries (from 4.6MB to 3.2MB for stripped release Linux x86_64)
• About 14% faster build time
• About 0.5-1.5% slower on x86_64
• Better quality

Fixes

• #1930
• #1986
• #1991

v0.2.0: Winter Solstice

Improvements

• Faster, overall 40%-70% faster than v0.1.0 depending on the encoding settings.
• Optional serialization/deserialization of the encoding parameters through the feature serialize
  • Optional cli advanced commands to use it.
• The builds are now using the dwarf debug format for the targets that support it, before it was a mixture of dwarf and stabs due to the nasm defaults.
• Added a --benchmark hidden flag for the cli for MacOS and Linux.
• documentation is now available on docs.rs.

Changes

• Segmentation support is now a tunable SpeedSetting and currently it is default off since it can produce desyncs, this does cause a 3% decrease in quality.

Fixes

• #1903 - edge-of-frame miscomputation
• #1858 - desync on speed 0 and 1 when certain quantizers are selected

Known issues

• #1930 - segmentation encoding may cause desync

Weekly pre-release

Improvements

• Faster, around 20% more compared to the last pre-release
  • More assembly optimizations from dav1d, mainly impacting AArch64
  • Forward transform SIMD, to be accounted for the largest part of this week speedup
  • Simplifications and refactoring in the Motion Estimation and Scene Change
• Optional serialization/deserialization of the encoding parameters through the feature serialize
  • Optional cli advanced commands to use it.
• The builds are now using the dwarf debug format for the targets that support it, before it was a mixture of dwarf and stabs due to the nasm defaults.

Changes

• Segmentation support is now a tunable SpeedSetting and currently it is default off since it can produce desyncs.

Fixes

• #1903 - edge-of-frame miscomputation
• #1858 - desync on speed 0 and 1 when certain quantizers are selected

Known issues

• #1930 - segmentation encoding may cause desync

Weekly pre-release

Improvements

• Slightly faster.
• 60%-80% reduction in the resident set depending on the number of tiles in use. Now the resident set is nearly independent from the number of tiles in use.

Changes

• The CI now uses Github Actions
• The tags will be p${isodate} for prereleases and v{semver} for releases.

Fixes

• Relax the y4m memory limit, now 8k content can be encoded again.

Known issues

• #1903
• #1858

Weekly pre-release

Improvements

• Faster (9.5% on 1080p content with 8 tiles).
• More Aarch64 assembly support
• rdo-lookahead-frames CLI option

Changes

• mp4 output removed to avoid circular dependencies.

Fixes

• Unbreak the desync_finder feature

Known issues

• #1858 `Tune PSNR may cause desyncs on 4k content"