Calculating differential cross-sections of particle interactions is an important tool in high-energy physics to predict experimental data and verify theoretical models. Generally, the computational complexity of these calculations scales factorially in the number of particles, therefore posing a considerable challenge. In the presentation, I will present a way to reduce the computational complexity significantly. Furthermore, I show the results obtained from the software developed during my diploma thesis, which represents the problem as a directed acyclic graph that can be optimized at a high level before being compiled into architecture-independent, scalable code. Finally, I will show how these optimizations complement compiler optimizations differently depending on the type of workload and target architecture and the subsequent effects on the generated code's runtime.