Skip to content

Latest commit

 

History

History
112 lines (93 loc) · 4.41 KB

CHANGELOG.md

File metadata and controls

112 lines (93 loc) · 4.41 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning, specifically the variant used by Rust.

Unreleased - TBD

0.1.8 - 2023-04-27

Fixed

  • Revert usage of slice::copy_within() as this causes a 20-35% slowdown due to the compiler not being able to optimize the code as much as before.
  • Use correct pkg-config filename.

Changed

  • Sync C header version with the C implementation.
  • Fix various new clippy warnings.
  • Use const generics to clean up some code.

0.1.7 - 2022-10-28

Changed

  • Various internal refactors / cleanups.
  • Updated dependencies.

0.1.6 - 2021-07-13

Fixed

  • Flush-to-zero implementation was refactored to prevent the possibility to pass it to mem::forget() and thus leading to UB. This was not a problem in this codebase.

Added

  • Specialized, optimized implementation for common interpolations. This speeds up the resampling/interpolation used by the true peak detection for mono, stereo and four channels by a factor of 2-5x.
  • Buffer-less true peak scanning instead of the previous implementation that goes via a temporary buffer. This speeds up true peak detection even more.
  • Add seeding API for supporting chunked analysis, i.e. analyzing a bigger input split into separate chunks in parallel.

Changed

  • Various refactoring.
  • unsafe code was removed from the histogram handling without a measurable impact on the performance.
  • Use the dasp crates instead of the homegrown sample/frame abstraction.

0.1.5 - 2020-09-07

Fixed

  • Allow only a single channel when setting DualMono also in EbuR128::set_channel_map(). The same constraint was already checked in set_channel().
  • Chunk size in some of the reference tests was changed as it was too big. This fixed the remaining two reference tests that were failing before.

Added

  • EbuR128::reset() to reset all state without reallocation.
  • Support for planar/non-interleaved inputs.

Changed

  • Sample peak measurement was changed slightly and is now faster than the C implementation while still giving the same results.
  • Use SmallVec for temporary vecs in EbuR128::loudness_global_multiple() and loudness_range_multiple() to prevent heap allocations, and also use it in the filter of the interpolator, which slightly speeds it up.
  • CI was switched from Travis to GitHub actions and greatly improved.

0.1.4 - 2020-08-31

Fixed

  • Fix compiler warning about unused use statement in benchmarks
  • Fix various spelling errors in code comments
  • Protect various calculations from overflows when creating a new instance or changing settings so that instead of panicking we return a proper error
  • Use std::f64::INFINITY and similar instead of f64::INFINITY to fix compilation with older Rust versions. Version up to Rust 1.32.0 are supported now.

Changed

  • Replaced various Vec<T> with Box<[T]> to make it clearer that these are never changing their size after the initial allocation, and to also get rid of the additional unused storage of the capacity.

Added

  • More documentation for internal types

0.1.3 - 2020-08-28

Fixed

  • Return -f64::INFINITY instead of f64::MIN for values below the thresholds from various functions. This matches the behaviour of the C library and generally makes more sense.

0.1.2 - 2020-08-28

Changed

  • Port libebur128 C code to Rust. This is a pure Rust implementation of it now and the only remaining C code is in the tests for comparing the two implementations.

0.1.1 - 2020-04-14

Fixed

  • Fix build with the MSVC toolchain.

0.1.0 - 2020-01-06

  • Initial release of ebur128.