Wilkinson-type error analysis applied to double-precision accumulation under contamination, II

Abstract

The floating point error analysis of Wilkinson of summation and inner products, modified in the preceding paper to account for contaminated inputs, is modified in a different way to be in terms of known quantities, to consider partial contamination, so as to give computable error bounds and, unlike the preceding paper, to mainly consider slight contamination. As before, the effects of extra (double) precision accumulation is compared.

Keywords:

• Wilkinson floating point error analysis
• Contamination
• Summation
• Inner product
• Double precision accumulation

C.R. Category::

• G.1.0