Estimation of Time Behavior of Selected Autonomous Driving Algorithms using GPGPU-Sim

Inhalt:
In order to achieve the goal of autonomous driving, more and more computing power has to be installed in vehicles. In addition to the already existing architectures, GPU manufacturers are trying to ensure that their products are used in vehicles. In the area of estimating non-intrinsic properties, such as execution time, there is still room for improvement, especially affecting GPUs. This work is to examining whether it is sufficient in this domain, that the execution time determination is conducted by an functional emulation and simple instruction counting. For this reason gpgpu-sim is extended by this simple method.