Skip to content

Latest commit

 

History

History
98 lines (61 loc) · 3 KB

CHANGELOG.md

File metadata and controls

98 lines (61 loc) · 3 KB

Version 0.4.1

  • Improve speed of FP math on platforms without FMA at the cost of precision
  • Update v_frame dependency to 0.3.8 in order to remove unsound code
  • Increase MSRV to 1.64.0
  • Factor out math abstractions into new subcrate yuvxyb-math
  • Refactor code to avoid some proc-macro dependencies

Version 0.4.0

  • Optimize HLG transfer function
  • Improve error types (breaking change)
  • Improve struct documentation
  • Remove unnecessary dependencies

Version 0.3.1

  • Fix some clippy lints, including some which may improve performace a bit
  • Bump nalgebra dependency version

Version 0.3.0

  • More math optimizations
  • Increase speed and reduce memory usage by converting in place
    • [Breaking] This required that some of the implementations of From<&T> be removed. The recommendation is that crates which were using these and require the input to not be moved should use from(input.clone()).
  • Add "fastmath" cargo feature, which is on by default, and can be turned off to disable fastmath optimizations
    • This is intended primarily for development and should not be disabled in the real world, unless you just like making your crate 5x slower for no good reason.

Version 0.2.3

  • More optimizations to linear/gamma conversion functions
  • More optimizations to everything basically

Version 0.2.2

  • Optimizations to linear/gamma conversion functions

Version 0.2.1

  • Add more exports from the crate root

Version 0.2.0

  • Add HSL color space, which can be converted to/from Linear RGB

Version 0.1.5

  • Speed up cube roots by allowing autovectorization

Version 0.1.4

  • Improve performance of conversions between linear RGB and XYB

Version 0.1.3

  • Fix conversions to and from XYB. All previous versions should not be used, as this conversion was very wrong prior to this.

Version 0.1.2

  • Add intermediate structs for Rgb and LinearRgb to make this crate more versatile

Version 0.1.1

  • Improvements to XYB conversion accuracy

Version 0.1.0

  • Make things faster
  • Unofficially declare that the interface is not going to change much after this

Version 0.0.6

  • Fix more conversion bugs

Version 0.0.5

  • Bump v_frame crate

Version 0.0.4

  • Use the v_frame crate instead of rolling our own
  • Add data_mut method for Xyb, useful for some metrics and does not violate invariants of frame size
  • Minor optimizations

Version 0.0.3

  • Add tests
  • Fix as many bugs as possible so that this crate is usable
    • The current state is that this crate is usable but there are slight discrepancies between the expected test values and some of the results. The discrepancies are small enough that this crate should be safe to use for things that do not critically require accuracy (i.e. if you just want some butteraugli metrics), but if you're using this for things that must be color exact, it's better to wait until these discrepancies are resolved.

Version 0.0.2

  • Initial release
  • There may be bugs, but the API is setup