Updates on Testing Microprocessors Effectively
Heather Quinn, Krishna Prasad Gnawali, Spyros Tragoudas
Abstract
In recent years, there has been an effort to create a standardized benchmark for radiation testing field-programmable gate arrays and microprocessors. The first version of benchmark was released in 2015. The input vectors and the error checkers have an unknown effect on previous test results gathered by the released benchmark. The checkers in the released benchmarks are efficient, but the variability of the input data that can be used is limited. In this article, a discussion of the changes to both the input vectors and the checkers are presented with test results showing the effect of both on the previous and new algorithms. Results show that the error checking method can increase the software's cross section by ten times. The results also show that the input vectors can mask errors in the system and in extreme situations mask radiation-induced errors completely.