package hardcaml_fixed_point

"Fixed point arithmetic types"

hardcaml_fixed_point

A simple fixed point library for Hardcaml.

Numbers are represented in a [I.F] format which provides [I] integer bits and [F] fractional bits. We require [I>0] and [F>=0].

Seperate modules for Signed and Unsigned fixed point numbers are implemented.

The basic arithmetic operators are provided which ensure no overflow is possible and track precision.

Comparisions do not require the same precision for each argument.

The most important function is [resize]. This can increase or decrease [I] and [F]. When increasing appropriate padding is performed. When decreasing [I] the result can either [saturate] or [wrap]. When decreasing [F] a rounding function can be specified:

• neg_infinity
• pos_infinity
• to_zero
• away_from_zero
• tie_to_neg_infinity
• tie_to_pos_infinity
• tie_to_zero
• tie_away_from_zero
• tie_to_nearest_even
• tie_to_nearest_odd

Note that [wrap] along with rounding to [neg_infinity] is effectively just truncating the top and bottom bits and uses no hardware.