The HPC group has been involved in numerous research activities, whose details are briefly presented below.
Research projects (some websites in Spanish)
- Algoritmos Evolutivos para la generación automática de Inteligencias Artificiales
Currently in progress.
- Reconocimiento facial robusto al envejecimiento
Currently in progress.
- Diseño de redes de contenido en plataformas cloud
Currently in progress.
- Inteligencia computacional aplicada a problemas de optimización de transporte urbano
Currently in progress.
- Algoritmos de inteligencia computacional para la detección de patrones de movimiento de personas
- Planificación de procesos en sistemas heterogéneos utilizando hwloc
- Energy-aware workflow scheduling in distributed datacenters
We study scheduling algorithms for large workflows in distributed datacenters.
- Pattern recognition on reckless driving
The project proposes a study about drivers on public roads in order to detect patterns of reckless driving through footage in real time. A study on pattern recognition in traffic patterns will be performed. The ultimate goal is develop a software product to generate alerts to detect patterns of reckless driving. These alerts will be shown in video comparison.
Partners: Universidad de la República (Uruguay).
- Evolutionary algorithms for traffic lights synchronization on Garzón Avenue
This project proposes the study of the problem of public transport planning by synchronizing traffic lights and design and implementation of an evolutionary algorithm to solve numerical computational performance and high efficiency.
Partners: Universidad de la República (Uruguay).
- Cloud Computing for Embryonic Development
The project proposes applying HPC/distributed computing techniques over cluster and cloud computing platforms for studying cell biology processes.
Partners: Universidad de la República (Uruguay), Universidad de Buenos Aires (Argentina).
- An evolutionary algorithm for taxi sharing optimization
The project presents the application of parallel micro evolutionary algorithms for solving the problem of distributing a group of passengers travelling from the same origin to different destinations in several taxis.
- Adaptive Deployment of MapReduce-based Applications over Pervasive and Desktop Grid Infrastructures (PERMARE)
The project proposes to improve the behavior of MapReduce-based applications on pervasive grids and desktop grids
Partners: Université de Reims Champagne-Ardenne (France), Universidad de la República (Uruguay), Universidade Federal de Santa Maria (Brasil), Université Paris 1 – Panthéon Sorbonne (France)
- Scheduling evaluation in heterogeneous computing systems with hwloc (SEHLOC)
The project proposes to develop runtime systems that combine application characteristics with topology information to automatically offer scheduling hints that try to respect hardware and software affinities.
Partners: Runtime group, INRIA Bordeaux-Sud-Ouest, France (local coordinator Brice Goglin), High Performance Computing Group, Universidad de la República, Uruguay (local coordinator Sergio Nesmachnow), LIDIC laboratory, Universidad Nacional de San Luis (local coordinator Marcela Printista).
- Virtual Machine Planning in Cloud Computing Systems
The project study efficient allocation algorithms for of a set of customers virtual machine requests into available pre-booked resources from a cloud broker, in order to maximize the broker profit..
- Energy-aware scheduling in heterogeneous computing systems
The project proposes the design and implementation of energy-aware scheduling algorithms for heterogeneous computing systems.
- Soft Computing and HPC for image processing
The project proposes the use of soft computing techniques implemented over high performance computing infrastructures to solve image processing problems.
- Facial Recognition Using Neural Networks over GPGPU
The project proposes designing a parallel neural network approach implemented over Graphic Processing Units (GPU) to solve a facial recognition problem, which consists in deciding where the face of a person in a certain image is pointing. The proposed method uses the parallel capabilities of GPU in order to train and evaluate a neural network used to solve the aforementioned problem.
- High Performance Computing for Self-gravity Calculation on Small Solar System Bodies
The project proposes using high performance computing techniques for efficient and accurate computing the self-gravity of granular solar system bodies (asteroids, comets, etc.).
- Greedy heterogeneous resource allocation in OurGrid
The project consists in designing a greedy resource allocation policy for OurGrid middleware in heterogeneous scenarios. The main purpose is the reduction of the makespan of jobs submitted by users.
- Recovering historical climatic data (Digi-Clima)
The aim of the Digi-Clima project is to provide semi-automatic processing capabilities for historical graphical rain intensity records. A stable version of the algorithm is available, but new features are under development. Digi-Clima been developed since early 2011 and by mid-2011 a running version was already available. The Digi-Clima application was conceived as a script written in Matlab-compatible language and should run with no problems using Matlab or Octave.
- Empirical Time Analysis of Evolutionary Algorithms as Software Programs
The project characterizes the computational efficiency behavior of a standard evolutionary algorithm as a software program. The experimental analysis allows concluding that significant improvemets in efficiency can be gained by applying simple guidelines on how to best program an evolutionary algorithm.
- Multiobjective Evolutionary Algorithms for Heterogeneous Computing Scheduling
The project implements and analyzes efficent multiobjective evolutionary algorithms for scheduling in large heterogeneous computing systems.
- Grid Initiatives for e-Science virtual communities in Europe and Latin America (GISELA)
The objective of GISELA is to ensure the long-term sustainability of the European–Latin American (EU-LA) e-Infrastructure, thus guaranteeing the continuity and the expected enhancement of the EU-LA Virtual Research Communities. The project focuses on two inter-related goals: i) implement the Latin American Grid Initiative (LGI) rooted on National Grid Initiatives (NGI)1, and ii) provide the Communities with the suited e-Infrastructure and Application-related Services required to improve the effectiveness of their research.
- High Performance Computing for Granular Media
The project proposes using high performance computing techniques for designing efficient and accurate algorithms for granular media simulations.
Cluster FING is the high performance scientific computing infrastructure in Facultad de Ingeniería. It has been used in nearly 50 research projects from Universidad de la República. The design, administration and operation of cluster FING is done by people from the HPC group at CeCal.
- Efficient heuristics and metaheuristics for heterogeneous computing scheduling
This project intends to design and implement efficent methods for scheduling in large heterogeneous computing systems, able to perform the planning in reduced times (i.e., less than a minute). The main concerns are the efficiency, accuracy and scalability of the studied methods.
- DigiClima: Parallel computing applied to the recovery of historical climate data
The project studies the application of parallel computing techniques to the recovery of historical climate data.
- Parallel quantum search algorithm for the 3-SAT
The project studies the application of parallel quantum computing algorithms for solving the 3-SAT problem.
- E-science Grid facility for Europe and Latin America (EELA-2)
EELA-2 aims at i) building a powerful, functional and well supported grid facility by expanding the e-Infrastructure in 41 Resource centers (13 in Europe and 28 in Latin America), by building the support of the e-Infrastructure to provide high reliability services to partnerships, and creating a comprehensive knowledge repository on Grid operation; ii) addressing a large community of users; iii) asserting the financial & management schemes to operate and support the e-Infrastructure on the long range; and iv) anticipating the handover of the e-Infrastructure operation and support.
- MOE framework (parallel multiobjective evolutionary algorithms
MOE is an object oriented framework for use of evolutionary algorithms in the resolution of multiobjective optimization problems..
- Parallel evolutionary algorithms for scheduling on heterogeneous computing environments
The project proposes the application of parallel evolutionary algorithms for solving the scheduling problem in large heterogenous clusters and grid environments. Information available: project website
- Parallel evolutionary algorithms for multiobjective scheduling on heterogeneous computing environments
The project proposes the application of parallel multiobjective evolutionary algorithms for solving the scheduling problem in large heterogenous clusters and grid environments. Information available: project website
(in Spanish), project report
- Metaheuristics applied to phylogenetic tree reconstruction
The project proposes the application of evolutionary algorithms for solving the phylogenetic tree reconstruction problem. Information available: project website
- High performance computing cluster with remote access
The main objective of this project is to design and implement an interface for the high performance cluster Medusa.
- High performance implementation of a real-time radiosity algorithm using GPUs
The project's goal is the implementation of a radiosity algorithm for fixed geometry and variable light scenarios, with a complexity order of nxm, (n is the number of patches and m<<n), running on a Cuda GPU architecture. The expected results includes the development of a code that works in real time (30 frames per second) with 10,000 patches scenes and with the hardware available.
- Laboratory for numerical simulation of free-surface flows
The project proposes the implementation of a high performance cluster for solving complex numerical simulation models related to free surface flows problems.
- Parallel computing applied to bioinformatics
- Parallel multiobjective evolutionary algorithms
- Design, configuration, administration and performance evaluation of a HPC cluster
The project proposes the implementation of a high performance cluster for solving applications with high CPU requirements. Information available: project website (in Spanish), project report (in Spanish), project website (in Spanish), technical report: performance evaluation (pdf, in Spanish)
- Improving the efficiency of a Río de la Plata numerical model
The project proposed to apply high performance techniques to improve the computational efficiency of the RMA-10 shallow water model. Information available: project website (in Spanish), Master thesis (pdf, in Spanish).
- Parallel genetic algorithms applied to frequency assignment problems in wireless networks
This project was an international cooperation with Universidad Nacional de la Patagonia San Juan Bosco, Argentina. Its main objectives involved studying the frequency assignment problems in wireless networks and the aplicability of parallel genetic algorithms to solve real-size scenarios. Information available: project website (in Spanish), graduate thesis (pdf, in Spanish), research article (pdf).
The project present the design and implementation of a parallel version of the multiobjective evolutionary algorithm NSGA-II and its application to reliable communication network design. Information available: project website (in Spanish, it includes software, documentation, test instances and results), technical report (pdf, in Spanish).
- Incremental genetic algoritms
- Hydrosedimentologic model of Río de la Plata
- Parallel genetic algorithms applied to reliable communication network design
- Parallel programming on shared memory architectures and its application to the Steiner problem in graphs
- Study of the architecture and the operation of the Sun2000E mutiprocessor system
- Parallel genetic algorithms for training neural nets
- HPCChess: Parallel access of GNUChess games database
- A parallel version of the PovRay computer graphics library
- Parallel genetic algorithms
- Development of a generic genetic engine (serial version 1.0)
- Parallel frontal solver for PDE in a shallow water model