Arithmetic coding for floats

This repository contains all the code for the runtime measurements in the thesis Softwarebasierte Fehlertoleranz numerischer Berechnungen für die funktionale Sicherheit am Beispiel der Automatisierungstechnik (Software-Based Fault Tolerance of Numerical Calculations for Functional Safety Using the Example of Automation Technology)

Link between the arithmetic codes and measurements in the thesis and the source code in this repository

• 4 Codierung von softwarebasierten Gleitkommazahlen
  • The measurements for chapter 4.1 Optimierungsmöglichkeiten von SoftFloat-Bibliotheken can be found in sfmaila and in the folder softfloat_simd
  • The arithmetic codes can be found in the folder sfmaila.
  • The measurments for chapter 4.3 can be found in the folder coded_softfloats, cable_bt/softfloat, cable_bt/softfloat_encoded, and cable_bt/softfloatpermutation
• 5 Optimierung von Ganzzahlcodierungen
  • The arithmetic codes and measurements for chapter 5 can be found in the folder coded_float. The code in this folder is an bug-fixed version of arithmetic-coding-int.
• 6 Hardwareunterstützte Codierung von Gleitkommazahlen
  • The arithmetic codes can be found in coded_fpu. The measurements are in folder coded_fpu, cable_bt/fpu_mapping, and cable_bt/fpu_roundend.
• 7 Codierung von Festkommazahlen
  • The transformed and encoded algorithm, as well as the measurements, can be found in the folder coded_fixpt and cable_bt/fixpt.

Build and run

./build.sh
./execute-perf.sh

License

Most of the code is licensed under the MIT license. The sine and cosine implementations are licensed under the LGPL. The coded sfmaila is licensed under the CECILL license.