A new data type
posit for direct drop-in replacement for IEEE754. - Simpler hardware implementation from 2 to thousands of bits.
- Larger dynamic range, higher accuracy, better closure under arithmetic operations, and simpler exception-handling
- Unlike unum arithmetic, posits do not require interval-type mathematics or variable size operands, and they round if an answer is inexact, much the way floats do.
Talk: https://www.youtube.com/watch?v=aP0Y1uAA-2Y
Slides: http://web.stanford.edu/class/ee380/Abstracts/170201-slides.pdf