Summary We analyze the feasibility and energy efficiency of using an unconventional cluster of low‐power Advanced RISC Machines processors to execute two scientific parallel applications. For this purpose, we have selected two applications that present high computational and communication cost: the Ondes3D that simulates geophysical events, and the all‐pairs N‐Body that simulates astrophysical events. We compare and discuss the impact of different compilation directives and processor frequency and how they interfere in Time‐to‐Solution and Energy‐to‐Solution. Our results demonstrate that by correctly tuning the application at compile time, for the Advanced RISC Machines architecture, we can considerably reduce the execution time and the energy spent by computing simulations. Furthermore, we observe reductions of up to 54.14% in Time‐to‐Solution and gains of up to 53.65% in Energy‐to‐Solution with two cores. Additionally, we consider the impact of two processor frequency governors on these metrics. Results indicate that the powersave governor presents a smaller instantaneous power consumption. However, it spends more time executing tasks, increasing the energy needed to achieve the solution. Finally, we correlate the energy consumption with the execution time in the experimental results using Pareto. These findings suggest that it is possible to explore low‐powered clusters for high‐performance computing applications by tuning application and hardware configuration to achieve energy efficiency. Copyright © 2016 John Wiley & Sons, Ltd.

No abstract

No abstract

Neste artigo buscamos analisar a viabilidade e a eficiência energética ao utilizar um cluster não convencional de processadores ARM para executar uma aplicação científica real. Utilizamos, para este fim, o Ondes3D que permite simular eventos geofísicos. Apresentamos um comparativo entre a execução da aplicação empregando diferentes flags de compilação e diferentes valores para a frequência do processador. Os resultados demonstraram ganhos de até 54.24% no tempo para a solução e 53.24% na energia para a solução, quando comparado com a execução sem nenhuma otimização. Estes resultados apontam que é possível explorar tal plataforma através da correta configuração do ambiente para se obter um bom equilíbrio entre desempenho e custo energético.

This work presents a novel unsupervised method to segment skin lesions in macroscopic images, grouping the pixels into three disjoint categories, namely 'skin lesion', 'suspicious region' and 'healthy skin'. These skin region categories are obtained by analyzing the agreement of adaptative thresholds applied to the different skin image color channels. In the sequence we use stochastic texture features to refine the suspicious regions. Our preliminary results are promising, and suggest that skin lesions can be segmented successfully with the proposed approach. Also, 'suspicious regions' are identified correctly, where it is uncertain if they belong to skin lesions or to the surrounding healthy skin. Motivation to Segment Skin Lesion Regions into Three CategoriesThere are different methods in the literature presenting approaches to segment pigmented skin lesions, with their effectiveness already confirmed [1]. However, most skin lesion segmentation methods generate a deterministic skin lesion rim, even when the lesion is affecting some of the surrounding areas with less intensity or has retracted beneath the skin. The proposed method tends to detect accurately 'suspicious regions' and refine their detection using stochastic features. Proposed Skin Lesion Segmentation ApproachThe three RGB color channels are used to obtain an initial estimate of the three types of skin regions, and then the initial segmentation is refined using stochastic textures as described in Section 3. We propose segmenting independently each of the RGB color channels using the Otsu's thresholding method. When analyzing the color information in the RGB color space, the lesion area is assumed to be darker than the background skin. For a given pixel, when all the three thresholded channels agree with the pixel categorization as 'healthy skin' or 'skin lesion', we assign it to the 'healthy skin' or 'skin lesion' categories. However, when there is no unanimous agreement among the three thresholded channels, the pixel is assigned to the 'suspicious region' category. More sophisticated algorithms could be used for the initial segmentation. Skin Lesion Segmentation RefinementSince pigmented skin lesions have no specific shape, color or texture, we consider the 'healthy skin' and 'skin lesion' as having different stochastic textures. In order to qualify these different stochastic textures within a skin lesion image, we assume the gradient distribution in these different skin regions as represented by asymmetric distributions (e.g. Gamma or Rayleigh), and use stochastic texture features to measure how much they deviate from randomness [2] (e.g. the regions within a skin lesion tend to be more clustered and less random than the surrounding healthy skin, which tends to be less clustered and more random). Also, describing the local gradient magnitudes by an asymmetric distribution facilitates the discrimination between the stochastic textures inside and outside the pigmented 'skin lesions' [3]. In our approach, the discrimination between the 'skin...