Meow: Memory Evaluation through Optimized Workloads - Synthesizing Litmus Tests for Memory Consistency Verification

In this paper, we propose the "Memory Evaluation through Optimized Workloads" (Meow) for memory evaluation. It is an automatic litmus test synthesis tool used to generate distinguishing litmus tests between two versions of the memory model written in cat language. The tool aims to identify the required tests in the program execution space capable of differentiating the models. It is based on the diy suite, which identifies the relaxations relaxed by the memory model. It then cherry-picks the relaxations which are capable of identifying the differences. These relaxations are used to generate litmus tests capable of showcasing the memory model change. Two different enhancement scenarios for the memory model were identified and tested. The results show a significant reduction in the time required to compute the test cases compared to other contemporary synthesis methods.