A fixed point library with functions that manipulate 8.24 fixed point numbers. Other bases to be added.
C++ C
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
app_example_fixed_point
app_example_uint64
app_test_against_double
doc
module_lib_fixed_8_24
module_lib_uint_32
module_lib_uint_64
.cproject
.gitignore
.project
LICENSE.txt
Makefile
README.rst

README.rst

Fixed point library

Version: unreleased
Status: 32-bit 8.24 feature complete.
Description:Fixed point library
Maintainer:https://github.com/henkmuller

This is a library of functions that implement standard mathematical functions in fixed point. One group of functions use a "8.24" fixed point format with 24 bits behind the binary point, and 8 bits before the binary point. Numbers are represented in 2's complement. Another group works on unsigned integers

More details to follow in the doc directory.

Key Features

  • 8.24 signed fixed point:
    • basic functions: mul, div, sqrt
    • emulation of manipulation of exponent and mantissa: ldexp, frexp
    • trigonometric functions: sin, cos
    • exponential functions: exp, log, sinh, cosh
    • test harness that checks on precision against double precision C.
  • 32.0 unsigned fiex point:
    • sqrt

To Do

  • Get size and speed into test harness.
  • Make 64 bit version, probably 16.48 (blocked by lack of long long in XC)
  • Run minimiser over errors in order to improve coefficients for our domain.
  • Add other mathematical functions (atan2, errf etc)
  • Add optional check on overflows (with a compile time flag for code that has size constraints)
  • Add optional check on domain errors, eg. sqrt(-1)

Firmware Overview

8.24

A collection of functions, all stored in a single module. The function names are suffixed with f8_24 to denote that these are single precision fixed point functions. There are some obvious dependencies, mostly on ldexp and frexp.

Two extra functions, mulf8_24 and divf8_24 perform multiplication and division on fixed point numbers.

32.0

A single function to compute sqrt

Known Issues

none

Required Repositories

  • xcommon git@github.com:xmos/xcommon.git

Support

Issues may be submitted via the Issues tab in this github repo. Response to any issues submitted as at the discretion of the maintainer for this line.