Browse Prior Art Database

Method for Computing Reference Results for Floating Point Operations

IP.com Disclosure Number: IPCOM000031133D
Original Publication Date: 2004-Sep-14
Included in the Prior Art Database: 2004-Sep-14

Publishing Venue

IBM

Abstract

The disclosed method describes an implementation of a Floating-Point reference model for multiply-and-add instructions, ensuring correctness by effectively computing the infinitely precise reference result as defined in the IEEE 754 standard. From this intermediate result the final result can be easily obtained.