Roberto Vaccaro

Improving search by incorporating evolution principles in parallel Tabu Search

Combinatorial optimization problems require computing efforts which grow at least exponentially with the problem dimension. Therefore, the use of the remarkable power of massively parallel systems constitutes an opportunity to be considered for solving significant applications in reasonable times. In this paper, starting from Tabu Search, a general optimization methodology, a parallel version, oriented to distributed memory multiprocessors and including evolution principles, has been introduced and discussed. The experiments have been performed on classical Traveling Salesman Problems and Quadratic Assignment Problems taken from literature. The results obtained show that the incorporation of evolution principles is very fruitful for the search strategy in terms of both convergence speed and solution precision.<<ETX>>

SIMULATION OF GENETIC ALGORITHMS ON MIMD MULTICOMPUTERS

In this paper, a Parallel Genetic Algorithm has been developed and mapped onto a coarse grain MIMD multicomputer whose processors have been configured in a fully connected chordal ring topology. In this way, parallel diffusion processes of good local information among processors have been carried out. The Parallel Genetic Algorithm has been applied, specifically, to the Travelling Salesman Problem. Many experiments have been performed with different combinations of genetic operators; the test results suggest that PMX crossover can be avoided by using only the inversion genetic operator and that a diffusion process leads to improved search in Parallel Genetic Algorithms.

A Clockless Computing System Based on the Static Dataflow Paradigm

The ambitious challenges posed by next exascale computing systems may require a critical re-examination of both architecture design and consolidated wisdom in terms of programming style and execution model, because such systems are expected to be constituted by thousands of processors with thousands of cores per chip. But how to build exascale architectures remains an open question.This paper presents a novel computing system based on a configurable architecture and a static dataflow execution model. We assume that the basic computational unit is constituted by a dataflow graph. Each processing node is constituted by an ad hoc kernel processor - designed to manage and schedule dataflow graphs, and a manycore dataflow execution engine - designed to execute such dataflow graphs.The main components of the dataflow execution engine are the Dataflow Actor Cores (DACs), which are small, identical and configurable. The major contributions of this paper are: i) the introduction of a machine language (named D#) which represents the low-level static configuration information of the system; ii) the introduction of a self-scheduled clockless mechanism to start operations on the presence of validity tokens only; iii) a design that avoids the need of temporary storage for tokens on the links of the DACs.Our preliminary tests on FPGA-based hardware show the feasibility of this approach.

A high-level dataflow system

This paper presents a new dataflow graph model, where only data tokens are allowed to flow. First we introduce a High-Level Dataflow System (HLDS) to describe a formal dataflow graph model, then we present a homogeneous HLDS (hHLDS) that formally describes our proposal. In this proposal the dataflow graph is obtained by employing only actors with homogeneous I/O conditions, that is, each actor, which executes an elemental operation, is characterised by having one output and two input arcs. Even though no control tokens are allowed, i.e. no T-gate, merge, and switch actors are present in this model, it is always possible to obtain dataflow graphs, which represent any programming structure and whose behaviour is well-behaved. As homogeneous I/O conditions are a severe restriction to represent the flow of a computation and the token flow in such dataflow graphs is completely asynchronous, proof is given to guarantee their determinacy. The main advantage of this representation is that it maps directly to hardware through a one-to-one correspondence between actors of the model and Functional Units of a dataflow machine.

A matrix multiplier case study for an evaluation of a configurable dataflow-machine

Configurable computing has become a subject of a great deal of research given its potential to greatly accelerate a wide variety of applications that require high throughput. In this context, the dataflow approach is still promising to accelerate the kernel of applications in the field of HPC. That tanks to a computational dataflow engine able to execute dataflow program graphs directly in a custom hardware. On the other hand, evaluating radically different models of computation remains yet an open issue. In this paper we present as case study the matrix multiplication that constitutes the fundamental kernel of the linear algebra. The evaluation takes into account the execution of the matrix product both in non-pipelined and pipelined modes. Results obtained running the execution of the two modes on an FPGA-based demonstrator show the validity of the configurable Dataflow-Machine. Moreover, at the same throughput, the power consumption is expected to be lower than in clock-based systems.

Position Paper: Validity of the Static Dataflow Approach for Exascale Computing Challenges

Instead of adhering to the common wisdom that endorses multi-threading, shared memory, and imperative programming as the only dominant models for efficient future architecture design, in this paper we argue that static dataflow is still a valid alternative model to guide the design of exascale machines. We explain our position showing how it is possible to directly map dataflow graphs onto a dataflow processor chip created with thousands of dataflow actor cores. This thanks to a demand-data driven approach to the co-design process that combines functional programming style with a dataflow model. This approach offers a valid solution to effectively and efficiently design, organize, and program dataflow machines.

Conditional and iterative structures using a homogeneous static dataflow graph model

This paper presents a static dataflow graph model, where only data tokens are allowed to flow. The proposed model is formally described, and the dataflow graph is obtained by employing only actors with homogeneous I/O conditions. Each actor, which executes an elemental operation, is characterized by having one output and two input arcs. Even though no control tokens are allowed, so that no T-gate, merge, and switch actors are present in this model, it is always possible to represent conditional and iterative structures whose behavior is well-behaved. As homogeneous I/O conditions are a severe restriction to represent the flow of a computation and the token flow in such dataflow graphs is completely asynchronous, proof is given to guarantee their determinacy.<<ETX>>

D 3 AS project: a different approach to the manycore challenges

The number of cores integrated onto a single die is expected to climb steadily in the foreseeable future. The main aim of Demand Data Driven Architecture System (D3AS) project is an attempt to provide a new programming model and architecture to allow efficient programming of highly parallel systems based on thousands of simple, thin cores. After a detailed description of the proposed prototype, some experimental results, obtained by a demonstrator, are discussed. Results show that the D3AS approach is feasible and promising.

Parallel tabu search versus parallel evolution strategies

There exists in scientific, industrial and financial communities a very strong request for techniques able to efficiently solve complex optimization problems. Because of this, several techniques are being currently investigated. Among them evolutionary algorithms and tabu search seem very interesting, not only for their intrinsic features but also because they both are easily parallelizable, so that they can take advantage of the parallel machines available on the market. A new parallel approach to tabu search (PTS) is introduced and compared against parallel evolution strategies on classical optimization problems taken from literature. The experimental results have shown the superiority of the PTS in both the solution quality and the convergence time.<<ETX>>

Supports for performance evaluation in heterogeneous distributed environments

In heterogeneous environments, very sophisticated modelling techniques and tools are required for the analysis of interactions between hardware and software subsystems. After a detailed discussion about performance evaluation of heterogeneous systems, an online monitoring tool for heterogeneous distributed systems is presented. Portability, modularity and a powerful graphic interface are the main features of the developed tool.<<ETX>>