numeric Accurate Floating Point Summation

[James Demmel]

You  may be interested in the final version of a paper by Yozo Hida and myself
on how to exploit the arithmetic formats in the IEEE 754 floating point standard
to get guaranteed accurate sums and dot products in an efficient manner.

You can find the paper on my web page www.cs.berkeley.edu/~demmel under

   Research Projects and Books/BeBOP/Selected Papers and Reports

or directly at

   www.cs.berkeley.edu/~demmel/AccurateSummation.{ps,pdf}

Regards,
Jim Demmel