Using GPU Programming for Inverse Spectroscopy

The Idaho National Laboratory (INL) has developed a detector that relies heavily on computationally expensive inverse spectroscopy algorithms to determine probabilistic three dimensional mappings of the source and intensity of a radioactive material. This inverse spectroscopy algorithm applies to material accountability by determining the location and strength of nuclear materials. Because of the computational expense, the INL has incorporated new hardware from the commercial graphics community. General programming for graphics processing units (GPUs) is not a new concept. However, the application of GPUs to evidence theory-based inverse spectroscopy is both novel and particularly apropos. Improvements while using a (slightly upgraded) standard PC are approximately three orders of magnitude, making a ninety-minute computation in less than six seconds. This significantly changes the concept of prohibitively expensive calculations and makes application to materials accountability possible in near real time, with the limiting condition reduced to sensor signal acquisition time.