Abstract
The second session on the topic of Multiprocessor Issues reviewed and evaluated the efficacy of the Ada 2012 support in the area of multiprocessor resource control. Andy Wellings, the Chair presented a proposal of an API that allowed controlling and extending the queue locks, and implementing the access control protocols. In the second part of the session, Miguel Pinho started by presenting an overview of Transactional Memory (TM), providing a quick overview of how in this approach atomic sections are executed concurrently and speculatively, in isolation. In the third topic of the session, the Chair started by providing an overview of the Reference Manual wordings concerning the access and control protocols for Protected Objects, noting that both the RM and the Annotated Reference Manual (ARM) do not fully define the access protocol for a protected object on a multiprocessor system. Finally, in the last topic (parallel barriers in Protected Objects), the workshop concluded that this would be a good mechanism to have, but that a suitable approach needs further investigation.

Abstract

Abstract
Designing cost-efficient multi-core real-time systems requires efficient techniques to allocate applications to cores while satisfying their timing constraints. However, existing approaches typically allocate using a First-Fit algorithm, which does not consider the execution time and potential parallelism of paths in the applications, resulting in over-dimensioned systems. This work addresses this problem by proposing a new heuristic algorithm, Critical-Path-First, for the allocation of real-time streaming applications modeled as dataflow graphs on 2D mesh multi-core processors. The main criteria of the algorithm is to allocate paths that have the highest impact on the execution time of the application first. It is also able to exploit parallelism in the application by allocating parallel paths on different cores. Experimental evaluation shows that the proposed heuristic improves the resource utilization by allocating up to 7% more applications and it minimizes the average end-to-end worst-case response time of the allocated applications by up to 31%.

Abstract
Preemptions via virtual channels have been proposed as a means to introduce the notion of priorities and real-time concepts in wormhole-switched NoC-based architectures. This work presents a holistic approach, which utilises a novel three-staged mapping method in order to assess what should be the physical characteristics of the platform (and its interconnect), such that real-time guarantees can be provided, assuming a given workload. We estimate the "gap" between platform characteristics required for the real-time analysis and those of currently available many-core platforms and propose to employ the existing feature of many-core platforms in order to significantly reduce this gap. The experiments demonstrate that virtual channels, an essential prerequisite for the real-time analysis, are not the bottle-neck. The approach presented in this paper can help system designers to select/design the most suitable platform for a given workload, such that all temporal constraints are met and over-dimensioning is avoided. © 2013 ACM.

Abstract
Multi-core platforms are challenging the way software is developed, in all application domains. For the particular case of real-time systems, models for the development of parallel software must be able to be shown correct in both functional and non-functional properties at design-time. In particular, issues such as concurrency, timing behaviour and interaction with the environment need to be addressed with the same caution as for the functional requirements. This paper proposes an execution model for the parallelization of real-time software, based upon a fine-grained parallelism support being proposed to Ada, a programming language particularly suited to the development of critical, concurrent software. We also show the correctness of the proposed model in terms of satisfying constraints related to execution order and unbounded priority inversions. © 2013 ACM.

Abstract
Experts provided information about parallel and multicore systems in papers submitted and discussed at a workshop. Discussion followed about the wisdom of giving any directive further than with parallel for the programmers to control the details of how parallelism was configured, executed, and potentially mapped in the runtime. The counter argument was raised that in real-time systems there was a need for the programmer to specify such control to directly specify the behavior, which was required for behavior analysis and timing behavior analysis. Questions were raised about the memory model of the proposal, and it was decided that the general model was that which supported a shared memory system, with cache coherency and uniform access to memory= within a single partition.