high performance computing on graphics processing units: hgpu.org

Histograms are commonly used in instrumentation to produce a visual representation of the probability density of a random signal from repeated measurements. However, histograms have a number of shortcomings as a method of data visualization. We propose using kernel density estimation as a replacement for histograms in instrumentation. Kernel density estimation has a number of advantages as a means of visualizing the probability density of a waveform or derived measurement. However, kernel density estimates have been considered too computationally burdensome for inclusion in instruments and virtual instruments. In this paper, we demonstrate that a graphics processing unit (GPU) can be used to compute and display kernel density estimates of actual measured data at a full video rate.