On performance analysis of heterogeneous parallel algorithms

Parallel morphological processing of hyperspectral image data on heterogeneous networks of computers

“…In order to assess the algorithm's performance, we resort to a framework for assessment of heterogeneous algorithms recently proposed by Lastovetsky and Reddy [17], who stated that a heterogeneous algorithm cannot be executed on a heterogeneous environment faster than its homogeneous prototype on the equivalent homogeneous environment. In [17], a homogeneous computing environment was considered equivalent to the heterogeneous one if: 1) both environments have the same number of processors; 2) the speed of each processor in the homogeneous environment is equal to the average speed of processors in the heterogeneous environment; and 3) the aggregate communication characteristics of the homogeneous environment are the same as those of the heterogeneous environment. As a result, the heterogeneous algorithm may be considered optimal if its efficiency is the same as that of the homogeneous prototype.…”

“…The processor cycle-time of the 16 processors in this architecture is in the heterogeneous cluster were also used to construct the homogeneous cluster, which allowed us to better control the accuracy of experiments by ensuring that these processors have the same speed in the homogeneous cluster running an homogeneous prototype and in the heterogeneous cluster running its corresponding heterogeneous algorithm. It is also important to emphasize that the configuration of the two platforms above was custom-designed to make sure that the two architectures are approximately equivalent in the context of our specific heterogeneous application, as detailed in [17].…”

Improving the Performance of Hyperspectral Image and Signal Processing Algorithms Using Parallel, Distributed and Specialized Hardware-Based Systems

“…All of them were custom-designed in order to approximate a recently proposed framework for evaluation of heterogeneous parallel algorithms [52], which relies on the assumption that a heterogeneous algorithm cannot be executed on a heterogeneous network faster than its homogeneous version on the equivalent homogeneous network. Let us assume that a heterogeneous network consists of { p i } …”

Heterogeneous Parallel Computing in Remote Sensing Applications: Current Trends and Future Perspectives

“…The networks were custom-designed in order to approximate a recently proposed framework for evaluation of heterogeneous parallel algorithms [11], which relies on the assumption that a heterogeneous algorithm cannot be executed on a heterogeneous network faster than its homogeneous version on the equivalent homogeneous network. In [11], a homogeneous computing environment is considered equivalent to the heterogeneous one in light of the three following principles:…”

“…For comparative purposes, a massively parallel Beowulf cluster at NASA's Goddard Space Flight Center is also used in experiments. The parallel algorithms are discussed in the context of a specific case study, focused on detecting fires and mapping debris compositions on a real hyperspectral data set collected by the AVIRIS sensor over the World Trade Center area in New York City after September 11,2001. Section 4 concludes with some remarks and hints at plausible future research.…”