Gabriel Arthur Gerber Andrade

Universidade Federal de Santa Catarina

Automation and Systems Engineering, PhD•  March 2017 - March 2021

Published a paper on which an hybrid coverage-driven random test generation was designed to be reusable across derivative designs, protocol variants, and distinct coverage metrics. The results shown that such generation can be superior to purely data-driven or model-based approaches, not as a result of plain composition, but due to proper decoupling of competences: data-driven (neighborhood) exploration and model-based (constraint) exploitation. By allowing superior coverage evolution (e.g. 2.7 times faster than a model-driven generator to reach maximal coverage) and high error discovery rate (e.g. required 10-15 minutes to find all errors evaluated in the study versus 45 minutes to 7 hours), a hybrid approach can raise the confidence in face of tight verification deadlines.

Universidade Federal de Santa Catarina

Computer Science, MS•  March 2014 - February 2017

Published a paper on which canonical dependence chains and address space constraints were exploited to improve the quality (e.g. improved the median functional coverage by 44\% and 83\% at the two highest hierarchical) of constrained random test generation for shared-memory verification.

Universidade Federal de Santa Catarina

Computer Science, BS•  March 2010 - February 2014

Joined the Embedded Computing Laboratory with the support of a program for scientific initiation. Took part in national qualifiers during four years.