For over half a century, microprocessor research has focused on improving performance. Such performance optimizations either reduce the latency in specific operations or enable the simultaneous execution of multiple programs on the same hardware. While these optimizations significantly enhance the capabilities of modern-day microprocessors, they also lead to contention among applications for shared hardware resources, such as cache memory. Due to this resource contention, the execution time of a program can be influenced by other contending programs. This leads to non-deterministic delays in execution, known as program interference. In applications with real-time constraints, such as automotive, space, and medical devices, operations must adhere to strict deadlines. In such scenarios, non-determinism in execution time can be fatal. In order to guarantee safety when the application has real-time constraints, a critical requirement is to estimate the worst-case interference from other executing programs. However, the complexity of multi-core hardware inhibits precisely determining the Worst-Case Program Interference (WCPI). Existing solutions are either prone to overestimate the interference or are not scalable to different hardware sizes and designs.

In this talk, we discuss Kryptonite, an automated framework to synthesize WCPI environments for multi-core systems. Fundamental to Kryptonite is a set of tiny hardware-specific code gadgets that are crafted to maximize interference locally. The gadgets are arranged using a greedy approach and then molded using a Reinforcement Learning algorithm to create the WCPI environment. We demonstrate Kryptonite on the automotive grade Infineon AURIX TC399 processor with a wide range of programs that includes a commercial real-time automotive application. We show that, while being easily scalable and tunable, Kryptonite creates WCPI environments in a black box setup without the requirement of proprietary information.