Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Precision Estimation and Control in Scientific and Engineering Computing

IP.com Disclosure Number: IPCOM000036540D
Original Publication Date: 1989-Oct-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 2 page(s) / 35K

Publishing Venue

IBM

Related People

Chatelin, F: AUTHOR [+3]

Abstract

This article describes a device for evaluating precision in scientific computations [*] through a random perturbation of floating point numbers exchanged between the central processing unit (CPU) of a personal computer and an arithmetic coprocessor. This perturbation is totally transparent to the micro-computer.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 59% of the total text.

Page 1 of 2

Precision Estimation and Control in Scientific and Engineering Computing

This article describes a device for evaluating precision in scientific computations [*] through a random perturbation of floating point numbers exchanged between the central processing unit (CPU) of a personal computer and an arithmetic coprocessor. This perturbation is totally transparent to the micro-computer.

Using a hardware generator of random numbers, the device automatically collects any number that must be perturbed on the bus from the arithmetic coprocessor, perturbs it randomly, and places it back on the bus to the CPU, without introducing any delay in the computing process.

Referring to Fig. 1, the scanning module monitors instructions sent over the data bus, and recognizes data to be perturbed (e.g., when instructed to move the data from the arithmetic coprocessor to the memory via the CPU).

The arithmetic and logic unit (ALU) systematically perturbs data on the bus, as required, by adding a "1" bit to the least significant data bit or by subtracting a "1" bit therefrom. In response to the control module, the perturbed data is put back on the bus and the carry is stored pending receipt of the next part of the floating point number to be perturbed. The perturbation process meets the requirements of the IEEE 754 (1985) standard relating to the representation of floating point numbers. The process only perturbs those numbers that have either normal or long format, and resolves probl...