HopliteRT Source Queuing Bound Correction

Abstract

We present a correction to the analytical source queuing bound for HopliteRT [1], [2], which addresses the counter-example put forward in Section IV-D of [3] by taking the effect of the in-flight jitter suffered by data flits into account. We reproduce the evaluation experiments from [1], [2] related to source queuing with this corrected approach, observing bounds that are 1.2X to1.7X larger than those originally reported.