Federico Aromolo is a Ph.D. student in Emerging Digital Technologies
(Embedded Systems curriculum) at the Scuola Superiore Sant'Anna (Pisa,
Italy), where he works at the Real-Time Systems Laboratory (ReTiS Lab)
under the supervision of Prof. Giorgio Buttazzo and Prof. Alessandro
Biondi. He holds a Master of Science degree in Embedded Computing Systems
from the Scuola Superiore Sant'Anna and the University of Pisa, and a
Bachelor of Science degree in Computer Engineering from the University of
Pisa, both achieved with highest honors. His research interests are in the
area of real-time embedded systems and include real-time scheduling and
synchronization algorithms, design and implementation of embedded and
cyber-physical systems, real-time operating systems, and advanced robotics
and artificial intelligence applications.

AG 1, AG 2, AG 3, INET, AG 4, AG 5, D6, SWS, RG1, MMCI   Info

AG Audience

English

60 Minutes

Kaiserslautern

Abstract:
Support for hardware acceleration and parallel software workloads on heterogeneous multiprocessor platforms is becoming increasingly relevant
in the design of high-performance and power-efficient real-time embedded systems. Communication between jobs dispatched on different cores and
specialized hardware accelerators such as FPGAs and GPUs is most often implemented using asynchronous events. The delays incurred by each task
due to the time spent waiting for such events should appropriately be accounted for in the timing analysis of the resulting scheduling behavior.
This talk presents the event-driven delay-induced (EDD) task model, which is suitable to represent and analyze the timing behavior of complex
computing workloads that incur event-related delays in the communication and synchronization between different processing elements. The EDD task
model generalizes several existing task models, providing enhanced expressiveness towards the timing analysis of parallel processing
workloads that involve both synchronous and asynchronous hardware acceleration requests. Two analysis techniques for EDD tasks executing on single-core platforms
under fixed-priority scheduling are presented; then, a model transformation technique is provided to analyze parallel real-time tasks executing under partitioned
multiprocessor scheduling by means of a set of EDD tasks. In the experiments, partitioned scheduling of parallel tasks is shown to outperform federated scheduling
when the proposed analysis approach is combined with specialized partitioning heuristics.

Please contact the office team for link information.

Vera Schreiber

+49 631 9303 9603

--email hidden

public