Alessio Bonfietti

An efficient and complete approach for throughput-maximal SDF allocation and scheduling on multi-core platforms

Our work focuses on allocating and scheduling a synchronous data-flow (SDF) graph onto a multi-core platform subject to a minimum throughput requirement. This problem has traditionally be tackled by incomplete approaches based on problem decomposition and local search, which could not guarantee optimality. Exact algorithms used to be considered reasonable only for small problem instances. We propose a complete algorithm based on Constraint Programming which solves the allocation and scheduling problem as a whole. We introduce a number of search acceleration techniques that significantly reduce run-time by aggressively pruning the search space without compromising optimality. The solver has been tested on a number of non-trivial instances and demonstrated promising run-times on SDFGs of practical size and one order of magnitude speed-up w.r.t. the fastest known complete approach.

Throughput Constraint for Synchronous Data Flow Graphs

Stream (data-flow) computing is considered an effective para-digm for parallel programming of high-end multi-core architectures for embedded applications (networking, multimedia, wireless communication). Our work addresses a key step in stream programming for embedded multicores, namely, the efficient mapping of a synchronous data-flow graph (SDFG) onto a multi-core platform subject to a minimum throughput requirement. This problem has been extensively studied in the past, and its complexity has lead researches to develop incomplete algorithms which cannot exclude false negatives. We developed a CP-based complete algorithm based on a new throughput-bounding constraint. The algorithm has been tested on a number of non-trivial SDFG mapping problems with promising results.

MPOpt-Cell: a high-performance data-flow programming environment for the CELL BE processor

We present MPOpt-Cell, an architecture-aware framework for high-productivity development and efficient execution of stream applications on the CELL BE Processor. It enables developers to quickly build Synchronous Data Flow (SDF) applications using a simple and intuitive programming interface based on a set of compiler directives that capture the key abstractions of SDF. The compiler backend and system runtime efficiently manage hardware resources.

Embedding Decision Trees and Random Forests in Constraint Programming

In past papers, we have introduced Empirical Model Learning (EML) as a method to enable Combinatorial Optimization on real world systems that are impervious to classical modeling approaches. The core idea in EML consists in embedding a Machine Learning model in a traditional combinatorial model. So far, the method has been demonstrated by using Neural Networks and Constraint Programming (CP). In this paper we add one more technique to the EML arsenal, by devising methods to embed Decision Trees (DTs) in CP. In particular, we propose three approaches: 1) a simple encoding based on meta-constraints; 2) a method using attribute discretization and a global table constraint; 3) an approach based on converting a DT into a Multi-valued Decision Diagram, which is then fed to an mdd constraint. We finally show how to embed in CP a Random Forest, a powerful type of ensemble classifier based on DTs. The proposed methods are compared in an experimental evaluation, highlighting their strengths and their weaknesses.

Global cyclic cumulative constraint

This paper proposes a global cumulative constraint for cyclic scheduling problems. In cyclic scheduling a project graph is periodically re-executed on a set of limited capacity resources. The objective is to find an assignment of start times to activities such that the feasible repetition period λ is minimized. Cyclic scheduling is an effective method to maximally exploit available resources by partially overlapping schedule repetitions. In our previous work [4], we have proposed a modular precedence constraint along with its filtering algorithm. The approach was based on the hypothesis that the end times of all activities should be assigned within the period: this allows the use of traditional resource constraints, but may introduce resource inefficiency. The adverse effects are particularly relevant for long activity durations and high resource availability. By relaxing this restriction, the problem becomes much more complicated and specific resource constrained filtering algorithms should be devised. Here, we introduce a global cumulative constraint based on modular arithmetic, that does not require the end times to be within the period. We show the advantages obtained for specific scenarios in terms of solution quality with respect to our previous approach, that was already superior with respect to state of the art techniques.

Disregarding Duration Uncertainty in Partial Order Schedules? Yes, We Can!

In the context of Scheduling under uncertainty, Partial Order Schedules (POS) provide a convenient way to build flexible solutions. A POS is obtained from a Project Graph by adding precedence constraints so that no resource conflict can arise, for any possible assignment of the activity durations. In this paper, we use a simulation approach to evaluate the expected makespan of a number of POSs, obtained by solving scheduling benchmarks via multiple approaches. Our evaluation leads us to the discovery of a striking correlation between the expected makespan and the makespan obtained by simply fixing all durations to their average. The strength of the correlation is such that it is possible to disregard completely the uncertainty during the schedule construction and yet obtain a very accurate estimation of the expected makespan. We provide a thorough empirical and theoretical analysis of this result, showing the existence of solid ground for finding a similarly strong relation on a broad class of scheduling problems of practical importance.

Precedence constraint posting for cyclic scheduling problems

Resource constrained cyclic scheduling problems consist in planning the execution over limited resources of a set of activities, to be indefinitely repeated. In such a context, the iteration period (i.e. the difference between the completion time of consecutive iterations) naturally replaces the makespan as a quality measure; exploiting inter-iteration overlapping is the primary method to obtain high quality schedules. Classical approaches for cyclic scheduling rely on the fact that, by fixing the iteration period, the problem admits an integer linear model. The optimal solution is then usually obtained iteratively, via linear or binary search on the possible iteration period values. In this paper we follow an alternative approach and provide a port of the key Precedence Constraint Posting ideas in a cyclic scheduling context; the value of the iteration period is not a-priori fixed, but results from conflict resolution decisions. A heuristic search method based on Iterative Flattening is used as a practical demonstrator; this was tested over instances from an industrial problem obtaining encouraging results.

Swarm-Based Controller for Traffic Lights Management

This paper presents a Traffic Lights control system, inspired by Swarm intelligence methodologies, in which every intersection controller makes independent decisions to pursue common goals and is able to improve the global traffic performance. The solution is low cost and widely applicable to different urban scenarios. This work is developed within the COLOMBO european project. Control methods are divided into macroscopic and microscopic control levels: the former reacts to macroscopic key figures such as mean congestion length and mean traffic density and acts on the choice of the signal program or the development of the frame signal program; the latter includes changes at short notice based on changes in the traffic flow: they include methods for signal program adaptation and development. The developed system has been widely tested on synthetic benchmarks with promising results.

CROSS cyclic resource-constrained scheduling solver

Cyclic scheduling problems consist in ordering a set of activities executed indefinitely over time in a periodic fashion, subject to precedence and resource constraints. This class of problems has many applications in manufacturing, embedded systems and compiler design, production and chemical systems. This paper proposes a Constraint Programming approach for cyclic scheduling problems, based on modular arithmetic: in particular, we introduce a modular precedence constraint and a global cumulative constraint along with their filtering algorithms. We discuss two possible formulations. The first one (referred to as CROSS) models a pure cyclic scheduling problem and makes use of both our novel constraints. The second formulation (referred to as CROSS^@?) introduces a restrictive assumption to enable the use of classical resources constraints, but may incur a loss of solution quality. Many traditional approaches to cyclic scheduling operate by fixing the period value and then solving a linear problem in a generate-and-test fashion. Conversely, our technique is based on a non-linear model and tackles the problem as a whole: the period value is inferred from the scheduling decisions. Our approach has been tested on a number of non-trivial synthetic instances and on a set of realistic industrial instances. The method proved to effective in finding high quality solutions in a very short amount of time.

The weighted average constraint

Weighted average expressions frequently appear in the context of allocation problems with balancing based constraints. In combinatorial optimization they are typically avoided by exploiting problems specificities or by operating on the search process. This approach fails to apply when the weights are decision variables and when the average value is part of a more complex expression. In this paper, we introduce a novel average constraint to provide a convenient model and efficient propagation for weighted average expressions appearing in a combinatorial model. This result is especially useful for Empirical Models extracted via Machine Learning (see [2]), which frequently count average expressions among their inputs. We provide basic and incremental filtering algorithms. The approach is tested on classical benchmarks from the OR literature and on a workload dispatching problem featuring an Empirical Model. In our experimentation the novel constraint, in particular with incremental filtering, proved to be even more efficient than traditional techniques to tackle weighted average expressions.