GPU implementation of multiline TRL calibration for efficient Monte-Carlo uncertainty analysis

We present a graphics-processing-unit (GPU) OpenCL implementation of the multiline through-reflect-line (TRL) algorithm of vector-network-analyzer (VNA) calibration. We obtained about 40× of speed improvement on a middle-end GPU as compared to a single-threaded C/C++ implementation. This speed improvement leads to a significant acceleration of VNA-measurement Monte-Carlo uncertainty analysis, making it possible to evaluate VNA measurement uncertainties in a timeframe of a few minutes.