Resumen de A time-predictable many-core processor design for critical real-time embedded systems
Critical Real-Time Embedded Systems (CRTES) are in charge of controlling fundamental parts of embedded system, e.g. energy harvesting solar panels in satellites, steering and breaking in cars, or flight management systems in airplanes. To do so, CRTES require strong evidence of correct functional and timing behavior. The former guarantees that the system operates correctly in response of its inputs; the latter ensures that its operations are performed within a predefined time budget.
CRTES aim at increasing the number and complexity of functions. Examples include the incorporation of \smarter" Advanced Driver Assistance System (ADAS) functionality in modern cars or advanced collision avoidance systems in Unmanned Aerial Vehicles (UAVs). All these new features, implemented in software, lead to an exponential growth in both performance requirements and software development complexity. Furthermore, there is a strong need to integrate multiple functions into the same computing platform to reduce the number of processing units, mass and space requirements, etc. Overall, there is a clear need to increase the computing power of current CRTES in order to support new sophisticated and complex functionality, and integrate multiple systems into a single platform.
The use of multi- and many-core processor architectures is increasingly seen in the CRTES industry as the solution to cope with the performance demand and cost constraints of future CRTES. Many-cores supply higher performance by exploiting the parallelism of applications while providing a better performance per watt as cores are maintained simpler with respect to complex single-core processors. Moreover, the parallelization capabilities allow scheduling multiple functions into the same processor, maximizing the hardware utilization.
However, the use of multi- and many-cores in CRTES also brings a number of challenges related to provide evidence about the correct operation of the system, especially in the timing domain. Hence, despite the advantages of many-cores and the fact that they are nowadays a reality in the embedded domain (e.g. Kalray MPPA, Freescale NXP P4080, TI Keystone II), their use in CRTES still requires finding efficient ways of providing reliable evidence about the correct operation of the system.
This thesis investigates the use of many-core processors in CRTES as a means to satisfy performance demands of future complex applications while providing the necessary timing guarantees. To do so, this thesis contributes to advance the state-of-the-art towards the exploitation of parallel capabilities of many-cores in CRTES contributing in two different computing domains. From the hardware domain, this thesis proposes new many-core designs that enable deriving reliable and tight timing guarantees. From the software domain, we present efficient scheduling and timing analysis techniques to exploit the parallelization capabilities of many-core architectures and to derive tight and trustworthy Worst-Case Execution Time (WCET) estimates of CRTES.
