This code implements the algorithms described in the following paper:

Daniel Lemire, Streaming Maximum-Minimum Filter Using No More than Three Comparisons per Element. Nordic Journal of Computing, 13 (4), pages 328-339, 2006.

A preprint is available there:

http://arxiv.org/abs/cs.DS/0610046

To reproduce the numbers from the paper, do the following:

make

./runningmaxmin --sine 1000000 10000 --windowrange 4 100 --times 1

./runningmaxmin --white 1000000 --windowrange 4 100 --times 1

The new algorithm introduced in the manuscript is most suitable for piecewise monotonic data or when low-latency is required. Otherwise, Gil-Kimmel and van Herk are good choices.

For an application of this idea to rolling statistics in JavaScript, see

https://github.com/shimondoodkin/efficient-rolling-stats

For an application in Go, please see

https://github.com/notnot/movingminmax