In this presentation, I will describe a project led by my PhD student Tyler Sorensen, on the systematic design of a testing environment that uses stressing and fuzzing to reveal errors in GPU applications that arise due to weak memory effects. This approach is evaluated across several CUDA applications that use fine-grained concurrency, on seven GPUs spanning three Nvidia architectures. The results show that applications that rarely, or never, exhibit errors related to weak memory when executed natively can readily exhibit these errors when executed in the testing environment. The testing environment also provides a means to identify the root causes of erroneous weak effects, and automatically suggests how to insert fences that experimentally eliminate these errors. This empirical fence insertion method carries significantly lower overhead, in terms of execution time and energy consumption, than a more conservative, guaranteed-sound approach.