Detail publikace

Analysis of Distortion Caused by Methods of Floating-Point Number Denormalization

Originální název

Analysis of Distortion Caused by Methods of Floating-Point Number Denormalization

Anglický název

Analysis of Distortion Caused by Methods of Floating-Point Number Denormalization

Jazyk

en

Originální abstrakt

This paper deals with an analysis of spectrum changes caused by methods for eliminating denormal floating-point numbers used in multimedia systems with floating-point arithmetic. The behavior of these methods can be simulated using nonlinear discrete systems without memory whose transfer characteristics can be approximated by a piecewise-linear function. An analytical solution of spectrum changes that appear in such systems is presented as well as the simulation results for three most widely used methods for elimination of denormal numbers.

Anglický abstrakt

This paper deals with an analysis of spectrum changes caused by methods for eliminating denormal floating-point numbers used in multimedia systems with floating-point arithmetic. The behavior of these methods can be simulated using nonlinear discrete systems without memory whose transfer characteristics can be approximated by a piecewise-linear function. An analytical solution of spectrum changes that appear in such systems is presented as well as the simulation results for three most widely used methods for elimination of denormal numbers.

BibTex


@inproceedings{BUT29017,
  author="Jiří {Schimmel}",
  title="Analysis of Distortion Caused by Methods of Floating-Point Number Denormalization",
  annote="This paper deals with an analysis of spectrum changes caused by methods for eliminating denormal floating-point numbers used in multimedia systems with floating-point arithmetic. The behavior of these methods can be simulated using nonlinear discrete systems without memory whose transfer characteristics can be approximated by a piecewise-linear function. An analytical solution of spectrum changes that appear in such systems is presented as well as the simulation results for three most widely used methods for elimination of denormal numbers.",
  address="BME - TMIT",
  booktitle="Proceedings of 32nd International Conference Telecommunications and Signal Processing TSP - 2009",
  chapter="29017",
  howpublished="print",
  institution="BME - TMIT",
  year="2009",
  month="august",
  pages="1--4",
  publisher="BME - TMIT",
  type="conference paper"
}