A Revised Presentation of the CENA Method
Abstract
The CENA method is a new automatic method to correct the first-order effect of floating point rounding errors on the result of a numerical algorithm. A first presentation of this method is proposed in the previous report RR-3828 (december 1999). The current report completes and replaces the previous one. We hark back to the presentation of the CENA method exhibiting new important aspects. We illustrate the pros-and-cons of the method solving triangular linear systems. These systems are designed to generate a significantly inaccurate computed solution. We propose a new analysis of the analogy and discrepancy between the linear correction and computing with extented precision.