Skip to main content
. 2019 Jan 28;150(4):044108. doi: 10.1063/1.5063794

FIG. 2.

FIG. 2.

SIMD reformulation of clustering algorithms allows greater scaling. (a) The runtime of the parallelized k-centers code as a function of data input size. (b) The runtime of the parallelized k-medoids code as a function of data input size. (c) The load time of the parallel code as a function of input data size. Points represent the average and error bars the standard deviation across three trials.