Software for testing exact summation of floating-point values

The latest version of this software is available in a gitlab repository.

The original software allows for replication of the performance tests of exact summation methods described in the following paper:

Neal, R. M. (2015) ``Fast exact summation using small and large superaccumulators'', Technical Report, 22 pages: abstract, pdf.

The programs are written in C. Note that to compare with the methods of Zhu and Hayes you will need to get their programs, which are available as supplementary information for their paper.

First release of 20 May 2015: tar archive.

Back to Radford Neal's home page