Browse Prior Art Database

Context-Based Error Correction

IP.com Disclosure Number: IPCOM000114874D
Original Publication Date: 1995-Feb-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 52K

Publishing Venue

IBM

Related People

Christensen, NT: AUTHOR [+2]

Abstract

A method for correcting data errors without the need for additional 'check' bits being added to the data, although an error detection method (such as parity checking) is required. The correct value of the data is determined by its context.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 54% of the total text.

Context-Based Error Correction

      A method for correcting data errors without the need for
additional 'check' bits being added to the data, although an error
detection method (such as parity checking) is required.  The correct
value of the data is determined by its context.

      A field of data in computer hardware is generally a collection
of binary bits, with parity checking, all of whose codepoints are
valid.  However, it frequently happens that such a field does not
exist in isolation, but is used in conjunction with other
'neighboring' fields of similar type.  In addition, these fields may
be codependent, in that values in one field may make certain values
in other fields invalid.  This information, then, becomes contextual
and it is possible to provide error correction on failing fields (as
detected by simple parity checking or other redundancy) by an
interpretation of the contents of its relevant neighbors; no
additional check bits need be added.

      In the current embodiment there are N fields, each of which
contains only one bit and its parity (this is required by the
computing system so that the bits may be individually modified
without examining all bits on each cycle to create the necessary
parities.)  N-1 of these fields are assigned to CPs in the system,
one per CP and represent ownership by that CP of some system
resource.  We will call this the 'Read-Only Vector' (ROV).  The Nth
bit is called the 'Exclusive' bit; when it is a 'one'...