According to Radford Neal, it is possible to sum a sequence of n-bit floating po...

rhaps0dy on June 11, 2021 | parent | context | favorite | on: How to average floating point numbers (2015)

According to Radford Neal, it is possible to sum a sequence of n-bit floating point numbers exactly, i.e. rounding to the n-bit floating point value closest to the exact result.

The algorithm uses 2n bits for the sum. https://radfordneal.wordpress.com/2015/05/21/exact-computati...