Search for: All records

The HPC industry is inexorably moving towards an era of extremely heterogeneous architectures, with more devices configured on any given HPC platform and potentially more kinds of devices, some of them highly specialized. Writing a separate code suitable for each target system for a given HPC application is not practical. The better solution is to use directive-based parallel programming models such as OpenMP. OpenMP provides a number of options for offloading a piece of code to devices like GPUs. To select the best option from such options during compilation, most modern compilers use analytical models to estimate the cost of executing the original code and the different offloading code variants. Building such an analytical model for compilers is a difficult task that necessitates a lot of effort on the part of a compiler engineer. Recently, machine learning techniques have been successfully applied to build cost models for a variety of compiler optimization problems. In this paper, we present COMPOFF, a cost model which uses the multi-layer perceptrons to statically estimates the Cost of OpenMP OFFloading. We used six different transformations on a parallel code of Wilson Dslash Operator to support GPU offloading, and we predicted their cost of execution on different GPUs using COMPOFF during compile time. Our results show that this model can predict offloading costs with a root mean squared error in prediction of less than 0.5 seconds. Our preliminary findings indicate that this work will make it much easier and faster for scientists and compiler developers to port legacy HPC applications that use OpenMP to new heterogeneous computing environments. 

Despite enormous structural variability exhibited in 3D chromosomal conformations at a global scale, there is a significant commonality of structures visible at smaller, local levels. We hypothesize that chromosomal conformations are representable as concatenations of a handful of prototypical shapelets, termed shape letters. This is akin to expressing complicated sentences in a language using only a small set of letters. Our goal is to organize the vast variability of 3D chromosomal conformation by constructing a set of predominant shape letters, termed a shape alphabet, using statistical shape analysis of curvelets taken from training conformations. This paper utilizes conformations generated from Integrative Genome Modeling to develop a shape alphabet as follows: it first segments 3D conformations into curvelets according to their Topologically Associated Domains. It then clusters these segments, estimates mean shapes, and refines and reorders these shapes into a Chromosome Shape Alphabet. The paper demonstrates effectiveness of this construction by successfully representing independent test conformations taken from IGM and other methods such as SIMBA3D, both symbolically and structurally, using the constructed alphabet. 

Abstract. A global in situ data set for validation of ocean colour productsfrom the ESA Ocean Colour Climate Change Initiative (OC-CCI) is presented.This version of the compilation, starting in 1997, now extends to 2021,which is important for the validation of the most recent satellite opticalsensors such as Sentinel 3B OLCI and NOAA-20 VIIRS. The data set comprisesin situ observations of the following variables: spectral remote-sensingreflectance, concentration of chlorophyll-a, spectral inherent opticalproperties, spectral diffuse attenuation coefficient, and total suspendedmatter. Data were obtained from multi-project archives acquired via openinternet services or from individual projects acquired directly from dataproviders. Methodologies were implemented for homogenization, qualitycontrol, and merging of all data. Minimal changes were made on the originaldata, other than conversion to a standard format, elimination of some points,after quality control and averaging of observations that were close in timeand space. The result is a merged table available in text format. Overall,the size of the data set grew with 148 432 rows, with each row representing aunique station in space and time (cf. 136 250 rows in previous version;Valente et al., 2019). Observations of remote-sensing reflectance increasedto 68 641 (cf. 59 781 in previous version; Valente et al., 2019). There wasalso a near tenfold increase in chlorophyll data since 2016. Metadata ofeach in situ measurement (original source, cruise or experiment, principalinvestigator) are included in the final table. By making the metadataavailable, provenance is better documented and it is also possible toanalyse each set of data separately. The compiled data are available athttps://doi.org/10.1594/PANGAEA.941318 (Valente et al., 2022). 

Abstract. A global compilation of in situ data is useful to evaluate thequality of ocean-colour satellite data records. Here we describe the datacompiled for the validation of the ocean-colour products from the ESA OceanColour Climate Change Initiative (OC-CCI). The data were acquired fromseveral sources (including, inter alia, MOBY, BOUSSOLE, AERONET-OC, SeaBASS, NOMAD,MERMAID, AMT, ICES, HOT and GeP&CO) and span the period from 1997 to 2018.Observations of the following variables were compiled: spectralremote-sensing reflectances, concentrations of chlorophyll a, spectralinherent optical properties, spectral diffuse attenuation coefficients andtotal suspended matter. The data were from multi-project archives acquiredvia open internet services or from individual projects, acquired directlyfrom data providers. Methodologies were implemented for homogenization,quality control and merging of all data. No changes were made to theoriginal data, other than averaging of observations that were close in timeand space, elimination of some points after quality control and conversionto a standard format. The final result is a merged table designed forvalidation of satellite-derived ocean-colour products and available in textformat. Metadata of each in situ measurement (original source, cruise orexperiment, principal investigator) was propagated throughout the work andmade available in the final table. By making the metadata available,provenance is better documented, and it is also possible to analyse each setof data separately. This paper also describes the changes that were made tothe compilation in relation to the previous version (Valente et al., 2016).The compiled data are available athttps://doi.org/10.1594/PANGAEA.898188 (Valente et al., 2019).