Skip to content

Latest commit

 

History

History
39 lines (29 loc) · 1.46 KB

CHANGELOG.md

File metadata and controls

39 lines (29 loc) · 1.46 KB

Planned improvements

  • Better rounding (with compiler option)
  • Additional functions (sin,cos,1/x,1/sqrt,...)
  • Comparison operators (<, <=, == ...)
  • Right and left shifts + <<= and >>=
  • Use of Posit16 routines for Posit8 (if ES8==ES16) to reduce library size
  • Avoid 32-bit arithmetic for addition
  • Avoid float casting for construction from integer
  • Create custom byte storage format for 0..1 interval (probabilities) using Posit10,0

Posit 0.1.2 - 2024.12.13

  • Additional functions (sin, cos, tan, atan, sign, abs, negate)
  • Use of Posit16 routine for Posit8_2float

Posit 0.1.1 - 2024.12.07

  • Posit16 <-> Posit8 translations
  • Next and prior functions
  • Sqrt for Posit8 and Posit 16 (different strategies, not always converging for Posit8)
  • Operator overloading for +=, -=, *=, /=
  • Now using 2's complement to store negative numbers

Posit 0.1.0 - 2024.09.15

Posit 0.1.0rc1 - 2024.09.12

  • Initial release
  • test sketch example, README, CHANGELOG, ...
  • Supports Posit8,0, Posit8,2 and Posit16,2
  • Constructor functions from (raw) uint8_t or uint16_t
  • Constructor from float32 (and int or double via float32 casting)
  • Four ops : addition, substraction, multiplication, division with overloading
  • Storage of negative numbers similar to IEEE754, no 2's complement