Tomo-GPU: a problem-solving environment for analysing tomographic data of structural characterization of materials

• Autor(a) principal: Quaresma, Paulo Jorge
• Data de Publicação: 2018
• Outros Autores: Lopes, Adriano, Medeiros, Pedro, Preto, Bruno, Velhinho, Alexandre
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://doi.org/10.34627/rcc.v5i0.30

The micro-structural characterization of composite materials uses X-ray tomography to collect information about the internal characteristics of the samples, in order to educate the researcher about their intrinsic properties. The raw tomographic data, need to go through several steps of computational processing, starting with the elimination of noise and other artifacts. The experience gained by the authors has shown that in some cases the required processing time is to long and therefore not easy for a materials scientist to interact with the program in order to define the most adequate computing parameters and the correct sequence of operations.This article describes a Problem Solving Environment (PSE), called Tomo-GPU, dedicated specifically to the field of tomography,and targeted to run on a desktop computer equipped with one or more GPGPUs. Tomo-GPU intends to exploit a hardware configuration in which is added to a traditional multi-core CPU, the processing capabilities of one or more GPUs to achieve, even with large volumes of data, execution times that are short enough to allow for an interactive use. One of the main contributions of this work is the support of a three-layer memory hierarchy where, in addition to the disk and RAM levels, traditionally used in the so called out-of-core memory computing, it is added the memory of the GPUs. The work is in a development stage with a number of features related to data processing and visualization already implemented.