An Innovative Strategy to Quickly Grade Functional Test Programs

Starting from the basic theory of dependencies between instructions inside an assembly code, this project aims to exploit them to evaluate firmware reliability. When a piece is produced, it has to be tested to ensure its reliability. Testing, in the case that is considered, means letting execute a "gold" piece (a piece that has been already tested and it can be considered working), extracting from it a register sign depending on every register value at the moment in which the sign is computed and, finally, using the "gold" sign as comparison with all other signs belonging to the pieces to be tested. If the sign of a normal piece does not match the "gold" sign, something surely went wrong; but what if the two signs match? It can not be said that the piece is working only considering this result. Whether the piece is working properly or not, depends also on the program that has been used to perform testing. Here is where instruction dependencies become useful. They are exploited to understand how the data are propagated though a particular execution of the "gold" piece: if the data propagation is very high, we can say that, in a probabilistic way, if an error occurs it should be propagated until the end (or until the sign is computed); by the contrary, if the data propagation is low, in case of something goes wrong the source of the problem could be lost. Graph theory is used as tool to perform analysis.