Extending instruction trace data flow analysis of functional test programs for multicore systems

Functional test strategies have become very popular for efficiently testing devices online and as a substitute for redundancy-based techniques. Unfortunately, they typically require a significant percentage of the overall testing and validation costs. Consequently, a preliminary and quick evaluation of functional test procedures has become essential to guide functional program development. In this work, we develop a technique for multi-core applications based on analyzing the execution traces generated by the functional program running within a debugger. Given the trace, we first create a graph representing the program data flow, i.e., all reading and writing operations on registers and memory addresses. Then, we manipulate the graph to evaluate whether each data is correctly forwarded to the program termination or a signature point. We finally label each operation as “good” (i.e., green) or “bad” (i.e., black) and compute a metric to represent the overall data flow of the program. Unlike previous approaches, our application runs in parallel with the debugger and delivers dynamic results showing the evolution of our metric over time. Moreover, it considers the synchronization points between different cores to evaluate the data flow between different logic cones correctly. We perform experiments on an automotive device manufactured by STMicroelectronics, and we demonstrate the approach’s effectiveness in terms of computation time and beneficial effects on fault coverage.