Releases: xiph/rav1e
Releases · xiph/rav1e
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
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 ofdwarf
andstabs
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 ofdwarf
andstabs
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 andv{semver}
for releases.
Fixes
- Relax the y4m memory limit, now 8k content can be encoded again.
Known issues
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"
Weekly pre-release
Improvements
- Faster (21-26% on 1080p content with 8 tiles).
- Tracing support via hawktracer
- Basic in-bitstream timing info support
- ASM building is parallel now.
Changes
- CDEF damping is adjusted based on quantizer
Fixes
- Builds cleanly on rust
1.40
Docs.rs
generation is working
Known issues
- #1858 `Tune PSNR may cause desyncs on 4k content"
v0.1.0: Made in Tokyo
First official release, published during the Video Dev Days 2019 in Tokyo.
Features
- Intra and inter frames
- 64x64 superblocks
- 4x4 to 64x64 RDO-selected square and 2:1/1:2 rectangular blocks
- DC, H, V, Paeth, smooth, and a subset of directional prediction modes
- DCT, (FLIP-)ADST and identity transforms (up to 64x64, 16x16 and 32x32 respectively)
- 8-, 10- and 12-bit depth color
- 4:2:0 (full support), 4:2:2 and 4:4:4 (limited) chroma sampling
- 11 speed settings (0-10)
- Near real-time encoding at high speed levels
- Rate control (single-pass and two-pass)
- Temporal RDO
- Scene cut detection
- CLI tool and C API
Weekly pre-release
Improvements
- More x86 assembly from dav1d
- Validate the frame size to reject impossibly small frames.
- Initial SSE4.1 support
Changes
- serde is now optional
- the fastrdo training feature is optional as well.
Fixes
- CDEF RDO could infiniloop in certain occasions.