| Grant number: | 13/26644-1 |
| Support Opportunities: | Regular Research Grants |
| Start date: | April 01, 2014 |
| End date: | March 31, 2016 |
| Field of knowledge: | Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques |
| Principal Investigator: | Raphael Yokoingawa de Camargo |
| Grantee: | Raphael Yokoingawa de Camargo |
| Host Institution: | Centro de Matemática, Computação e Cognição (CMCC). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André , SP, Brazil |
| Associated researchers: | Alfredo Goldman vel Lejbman ; Daniel de Angelis Cordeiro ; David Corrêa Martins Junior ; Luiz Carlos da Silva Rozante ; Siang Wun Song |
Abstract
With the advent of different classes of accelerators such as GPUs (Graphical Processing Units) and Intel MIC (Many Integrated Cores), heterogeneous clusters composed of different types of accelerators and processors became reality. We consider dedicated and shared clusters, the later composed of shared workstations distributed in several laboratories. The architectural differences between processors and the various types of accelerators make it difficult to develop applications that use these clusters in an efficient way.In this project we will propose and evaluate programming models that facilitate the development and the performance prediction of aplpications for heterogeneous clusters. We will also develop mechanisms for dynamic load distribution in these clusters, which will be implemented in an existing framework, such as StarPU. The aim is to simplify both the implementation and the execution of applications for these clusters. In the area of applications, we will develop bioinformatics algorithms for the inference of gene regulatory networks (GRNs), where we need to evaluate a large number of combinations of candidate predictor genes. In computational neuroscience, we will implement support for accelerators in neuronal simulators such as MOOSE, allowing the simulation of neural networks composed of thousands of neurons. We will develop algorithms for the efficient solution of systems of linear equations, which are the basis for the simulation of neurons. For both areas, we will focus in the efficient execution on heterogeneous clusters. (AU)
| Articles published in Agência FAPESP Newsletter about the research grant: |
| More itemsLess items |
| TITULO |
| Articles published in other media outlets ( ): |
| More itemsLess items |
| VEICULO: TITULO (DATA) |
| VEICULO: TITULO (DATA) |