4 research outputs found

    Implementazione MPI di Parmod

    No full text
    Parmod (Modulo Parallelo) e' il nuovo costrutto di programmazione parallela strutturata introdotto con ASSIST. Esso generalizza e amplia le caratteristiche degli skeletons classici, offrendo una maggiore potenza espressiva, un alto livello di astrazione e la possibilità di esprimere facilmente nuove forme di parallelismo. Questa tesi nasce dall'idea di facilitare la diffusione di ParMod anche al di fuori del modello di programmazione Assist. A tale scopo e' stata implementata una libreria che permettesse l'uso del costrutto da normali programmi C++ utilizzando un supporto MPI. Un punto di forza della libreria è la distribuzione dinamica dei processi paralleli sulle risorse reali. La progettazione e l'implementazione hanno richiesto due fasi principali: 1) l'analisi e la scelta di un'interfaccia che garantisse un buon livello di usabilita' e programmabilita'. 2) la creazione di un supporto a run-time efficiente, che permettesse di supportare l'esecuzione corretta del modulo parallelo. La parte successiva all'implementazione è stata quella di tests, in cui è stato individuato un insieme di benchmarks necessario per analizzare le prestazione del sistema, in termini di scalabilità, efficienza e tempo di completamento; gli stessi benchmarks sono stati provati anche su Assist (vers. 1.1), al fine di effettuarne un confronto. La versione attuale di Assist effettua un mapping statico delle risorse logiche su quelle fisiche, quindi il confronto ha permesso di valutare anche il comportamento di un approccio dinamico rispetto a quello statico. I risultati ottenuti hanno evidenziato degli ottimi valori di scalabilità ed efficienza, a volte anche vicini all'ideale, reggendo bene il confronto con Assist

    An Insight into Chemistry and Structure of Colloidal 2D-WS<sub>2</sub> Nanoflakes: Combined XPS and XRD Study

    No full text
    The surface and structural characterization techniques of three atom-thick bi-dimensional 2D-WS2 colloidal nanocrystals cross the limit of bulk investigation, offering the possibility of simultaneous phase identification, structural-to-morphological evaluation, and surface chemical description. In the present study, we report a rational understanding based on X-ray photoelectron spectroscopy (XPS) and structural inspection of two kinds of dimensionally controllable 2D-WS2 colloidal nanoflakes (NFLs) generated with a surfactant assisted non-hydrolytic route. The qualitative and quantitative determination of 1T’ and 2H phases based on W 4f XPS signal components, together with the presence of two kinds of sulfur ions, S22− and S2−, based on S 2p signal and related to the formation of WS2 and WOxSy in a mixed oxygen-sulfur environment, are carefully reported and discussed for both nanocrystals breeds. The XPS results are used as an input for detailed X-ray Diffraction (XRD) analysis allowing for a clear discrimination of NFLs crystal habit, and an estimation of the exact number of atomic monolayers composing the 2D-WS2 nanocrystalline samples

    A dataflow-like implementation of ASSIST parmod

    No full text
    ASSIST is a structured parallel programming environment targeting networks/clusters of workstations and grids. It introduced the parmod parallel construct, supporting a variety of parallelism exploitation patterns, including classical ones. The original implementation of parmod relies on static assignment of parallel activities to the processing elements at hand. In this work we discuss an alternative implementation of the parmod construct that implements completely dynamic assignment of parallel activities to the processing elements. We show that the new implementation introduces very limited overhead in case of regular computations, whereas it performs much better than the original one in case of irregular applications. The whole implementation of parmod is available as a C++/MPI library

    Synthesis of reduced graphite oxide by a novel green process based on UV light irradiation

    No full text
    We have developed a novel and straightforward approach for the green synthesis of reduced graphite oxide (rGO). First, graphite oxide (GO) was prepared by the Hummers' oxidation method, starting from high-surfacearea graphite. Then, rGO was generated from GO in aqueous suspension through a UV-irradiation treatment. The influence of different process parameters (including type of UV source, irradiation time and atmosphere) on the GO reduction efficiency was explored and evaluated on the basis of the data acquired by several experimental techniques, such as infrared spectroscopy in attenuated total reflectance mode, X-ray diffraction, UV-vis absorption spectrophotometry, X-ray photoelectron spectroscopy and thermogravimetry. The acquired results allowed identifying appropriate sets of reaction conditions under which GO reduction yield could be maximized. In particular, the highest reduction degree was obtained by exposing GO to UV light in a UV oven for 48 h under inert atmosphere. The reduction strategy developed by us represents an innovative low-cost and easy route to graphene-based nanomaterials, which does not require any stabilizer, photocatalyst or reducing agent. For this reason, our method represents an attractive environmentally friendly alternative approach for the preparation of stable rGO dispersions in large-scale amounts, to be utilizable in disparate engineering applications
    corecore