Research Overview
Hardware / Software System Engineering is the common denominator of research projects pursued at the Chair of Integrated Systems. Our manner of hardware / software systems engineering covers the development of new architectures for application-specific multicore processors (MPSoC) and distributed embedded systems, as well as techniques and methods for complexity governance and co-optimization of functional and different non-functional performance metrics. Last but not least, we devise tools to assess performance, power and reliability characteristics of such systems during early phases of design (HW/SW Codesign) for the sake of a successive improved architecture development.
Architectures
Architecture deals with the structural composition of a system from its base components. MPSoC components consist of RISC CPU cores, memory and I/O controllers, DMA units, hardware accelerators and coprocessors as well as standardized interconnect structures and interfaces. Application requirements determine type, number and topology of the components which are assembled to MPSoC architectures.
Internet packet processing
- MUCOS (finished)
Scalable Routing Platform for Demonstrating Model-Based Software Development Flow - FlexPath (finished)
Network processor architectures with flexible data paths - NPU100 (finished)
Design of a Network Processor Datapath for 100Gbit/s Carrier Grade Ethernet
Visual Computing
- AutoVision (finished)
A Run-time Reconfigurable MPSoC Architecture for Future Driver Assistance Systems
Self-adaptive workload, reliability and power optimization
- ASoC (finished)
Autonomic Systems on Chip - InvasIC B3
Invasive Loosely-Coupled MPSoCs - EndorA
Energy Efficient Design of a Future Oriented E/E-Architecture - RELY
Design for RELIABILITY of SoCs for Applications
Thread Management & Message Passing
- MAPCO
Multicore Architecture and Programming Model Co-Optimization
Software Defined Radio (SDR)
- PROTON (finished)
Software Defined Radio platform for automotive telematics applications
I/O Virtualization
- VirTherm 3D
Communication Virtualization Enabling Systen Management for Dependable 3D MPSoCs - ARAMiS
Automotive, Railway and Avionic Multicore System
CPU Data-/Controlpath Protection against transient Runtime and Single Event Upsets
- AIS (finished)
Autonome Integrierte Systeme
FPGA Demonstrator Platforms for Multicore
- OpTiMSoC
Open Tiled Manycore System-on-Chip - InvasIC Z2
Validation and Demonstrator
Network-on-Chip (NoC):
Multi-topology hierarchical NoC
- InvasIC B5
Invasive NoCs - Autonomous, Self-Optimising Communication Infrastructures for MPSoCs - RapidMPSoC (finished)
Rapid system prototyping and plattform-based design for mixed-signal multi-processor SoC: Future on-chip communication structures
3D NoC
- NEEDS
Interconnection Networks for 3D-ICs
Methods
Methods are concerned with the conceptual approaches and techniques to achieve specified optimization goals. Safety of a technical system can, for example, be improved by triplication of components followed by a majority voting on the individual results (Triple Modular Redundancy). Alternatively, machine learning techniques can be used to sense deviations from the expected behavior as part of an overall self-organizing system. These are two conceptually different approaches, or methods, for achieving the same goal. Methods, in general, aren’t isolated or independent from architectures and tools. Methods are frequently interlinked with architectures and tools on and with which they are implemented.
Invasive Computing
- InvasIC B3
Invasive Loosely-Coupled MPSoCs - InvasIC B5
Invasive NoCs - Autonomous, Self-Optimising Communication Infrastructures for MPSoCs - InvasIC C2
Simulation of Invasive Applications and Invasive Architectures - InvasIC D1
Invasive Software-Hardware Architectures for Robotics
Reconfigurable Computing
- PROTON (finished)
Software Defined Radio platform for automotive telematics applications - ARAMiS
Automotive, Railway and Avionic Multicore System - AutoVision (finished)
A Run-time Reconfigurable MPSoC Architecture for Future Driver Assistance Systems - FlexPath (finished)
Network processor architectures with flexible data paths
Bio-inspired Techniques for Self-Organization
- ASoC (finished)
Autonomic Systems on Chip
Fault Tolerance
- AIS (finished)
Autonome Integrierte Systeme
Hardware-based Machine Learning
- ASoC (finished)
Autonomic Systems on Chip
Energy/Power Management
- EndorA
Energy Efficient Design of a Future Oriented E/E-Architecture
FPGA Prototyping
- InvasIC Z2
Validation and Demonstrator - OpTiMSoC
Open Tiled Manycore System-on-Chip
Tools
Simulation tools at a high level of abstraction are prerequisite for early feedback on design decisions during hardware / software system design. Qualitative evaluations are necessary for both performance related system characteristics (data throughput, transaction latency, operation frequency) as well as for energy consumption and robustness or reliability perspectives. The rapid assessment of different architecture alternatives enables MPSoC developers to cover a significantly larger design space than time and resource intensive realizations at lower levels of abstraction.
Simulation-based Design Space Exploration of MPSoC
- McSIM
Trace-driven simulator for multiprocessor systems-on-chip - EndorA
Energy Efficient Design of a Future Oriented E/E-Architecture - TAPES (finished)
Trace-based Architecture Performance Evaluation with SystemC
Software Performance Estimation
- SciSim (finished)
SciSim
Vulnerability Analyse für Soft-Errors
- Soft Error Rate Estimation (finished)
Soft Error Rate Estimation