Abstract

Software is increasingly complex and is used in increasingly critical applications. Sophisticated techniques are available for verifying that software systems work correctly, but these techniques can be very difficult and expensive to use. Researchers have developed tools to automatically verify software models, but using these tools can still be very costly, in terms of manual effort and expertise required to build accurate models and to formally specify required properties, and also in terms of the time and memory required to run these tools. Much work has been done to simplify the process of building software models and to improve the performance of verification tools, resulting in a variety of different modeling languages, each with features designed to reduce effort or improve performance for certain types of input models, and a range of verification tools, each with a different set of strategies available for reducing time and memory requirements. It can be difficult to determine which verification strategy is best for a particular software system. Others have observed complementary relationships between tools and have argued that there is no single best tool—that as users’ needs change the choice of tool should change as well. This dissertation provides further evidence for complementary relationships between verification tools, specifically considering tools available for specifications of synchronous software systems

Citations