Search

We propose a methodology to address the programmability issues derived from the emergence of new-generation shared-memory NUMA architectures. For this purpose, we employ dense matrix factorizations and matrix inversion (DMFI) as a use case, and we target two modern architectures (AMD Rome and Huawei Kunpeng 920) that exhibit configurable NUMA topologies. Our methodology pursues performance portability across different NUMA configurations by proposing multi-domain implementations for DMFI plus a hybrid task- and loop-level parallelization that configures multi-threaded executions to fix core-to-data binding, exploiting locality at the expense of minor code modifications. In addition, we introduce a generalization of the multi-domain implementations for DMFI that offers support for virtually any NUMA topology in present and future architectures. Our experimentation on the two target architectures for three representative dense linear algebra operations validates the proposal, reveals insights on the necessity of adapting both the codes and their execution to improve data access locality, and reports performance across architectures and inter- and intra-socket NUMA configurations competitive with state-of-the-art message-passing implementations, maintaining the ease of development usually associated with shared-memory programming. This research was sponsored by project PID2019-107255GB of Ministerio de Ciencia, Innovación y Universidades; project S2018/TCS-4423 of Comunidad de Madrid; project 2017-SGR-1414 of the Generalitat de Catalunya and the Madrid Government under the Multiannual Agreement with UCM in the line Program to Stimulate Research for Young Doctors in the context of the V PRICIT, project PR65/19-22445. This project has also received funding from the European High-Performance Computing Joint Undertaking (JU) under grant agreement No 955558. The JU receives support from the European Union’s Horizon 2020 research and innovation programme, and Spain, Germany, France, Italy, Poland, Switzerland, Norway. The work is also supported by grants PID2020-113656RB-C22 and PID2021-126576NB-I00 of MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe.

We have created a dataset of 269 perovskite solar cells, containing information about their perovskite family, cell architecture, and multiple hole-transporting materials features, including fingerprints, additives, and structural and electronic features. We propose a predictive machine learning model that is trained on these data and can be used to screen possible candidate hole-transporting materials. Our approach allows us to predict the performance of perovskite solar cells with reasonable accuracy and is able to successfully identify most of the top-performing and lowest-performing hole-transporting materials in the dataset. We discuss the effect of data biases on the distribution of perovskite families/architectures on the model's accuracy and offer an analysis with a subset of the data to accurately study the effect of the hole-transporting material on the solar cell performance. Finally, we discuss some chemical fragments, like arylamine and aryloxy groups, which present a relatively large positive correlation with the efficiency of the cell, whereas other groups, like thiophene groups, display a negative correlation with power conversion efficiency (PCE).

Supplemental Material for Compressed basis GMRES on high-performance graphics processing units by José I Aliaga, Hartwig Anzt, Thomas Grützmacher, Enrique S Quintana-Ortí, and Andrés E Tomás in The International Journal of High Performance Computing Applications.

Supplemental Material, 20200430reviewers.comments for Reproducibility of parallel preconditioned conjugate gradient in hybrid programming environments by Roman Iakymchuk, Maria Barreda Vayá, Stef Graillat, José I. Aliaga and Enrique S. Quintana-Ortí in The International Journal of High Performance Computing Applications

Supplemental Material, 20200430diff for Reproducibility of parallel preconditioned conjugate gradient in hybrid programming environments by Roman Iakymchuk, Maria Barreda Vayá, Stef Graillat, José I. Aliaga and Enrique S. Quintana-Ortí in The International Journal of High Performance Computing Applications

We address the parallelization of the LU factorization of hierarchical matrices ($\mathcal{H}$-matrices) arising from boundary element methods. Our approach exploits task-parallelism via the OmpSs programming model and runtime, which discovers the data-flow parallelism intrinsic to the operation at execution time, via the analysis of data dependencies based on the memory addresses of the tasks' operands. This is especially challenging for $\mathcal{H}$-matrices, as the structures containing the data vary in dimension during the execution. We tackle this issue by decoupling the data structure from that used to detect dependencies. Furthermore, we leverage the support for weak operands and early release of dependencies, recently introduced in OmpSs-2, to accelerate the execution of parallel codes with nested task-parallelism and fine-grain tasks.

Donor-acceptor alkynes, endowed with 11,11,12,12-tetracyano-9,10-anthraquinodimethane (TCAQ) and N,N-dimethylaniline (DMA) units, have been further functionalized by a [2+2] tetracyanoethylene (TCNE) cycloaddition followed by a subsequent retroelectrocyclization to form distorted non-planar molecular structures with 1,1,4,4-tetracyanobuta-1,3-diene (TCBD) bridge ligands. Comprehensive spectroscopic, electrochemical, and computational studies have been carried out to compare the electronic communication in these planar (alkyne) and nonplanar (with TCBD units) TCAQ-based push-pull chromophores. Cyclic voltammetry and UV-Vis absorption measurements confirm a noticeable electronic communication between the TCAQ and DMA units regardless the quasi-orthogonal arrangement of the two dicyanovinyl halves of the TCBD groups, which partially hinder the electronic communication. The experimental trends are strongly supported by theoretical calculations performed at the density functional theory level, which further evidence an active electron-withdrawing role of the TCBD bridge both in the formation of the charged species and in the lowest-lying absorption features. The novel push-pull TCAQ-based derivatives including the TCBD bridge show a broad absorption in the whole visible range while having a structure highly distorted from planarity. These chromophores may therefore be viewed as appealing candidates to be exploited in photovoltaic devices with minimal aggregation phenomena.

En este artículo se presenta una implementación de BLAS, basada en la biblioteca BLIS, para AMPs. La evaluación de esta versión consciente de la asimetría se lleva a cabo a través de tres operaciones comunes de la biblioteca LAPACK: la factorización LU, la factorización Cholesky y la reducción a la forma tridiagonal. Los tests iniciales que emplean la implementación para AMPs directamente muestran las mejoras obtenidas con la adaptación del software al utilizarlo como base de las operaciones LAPACK, obteniendo hasta un 90% del rendimiento máximo esperado. Además, estas mejoras se ven incrementadas al combinar la versión diseñada para AMPs con un runtime, en cuyo caso se obtienen rendimientos hasta un 30% superiores respecto a la utilización directa de la versión para AMPs.

Proceedings of the First PhD Symposium on Sustainable Ultrascale Computing Systems (NESUS PhD 2016) Timisoara, Romania. February 8-11, 2016. Energy consumption is a key aspect that conditions the proper functioning of nowadays data centers and high performance computing just like the launch of new services, due to its environmental negative impact and the increasing economic costs of energy. The energy efficiency of the applications used in these data centers could be improved, especially when systems’ utilization rate is low or moderate, or when targeting memory bounded applications. In this sense, energy proportionality stands for systems which power consumption is in line with the amount of work performed in each moment. As a response to these needs, the main objective of this project is to study, design, develop and analyze experimental solutions (models, programs, tools and techniques) aware of energy proportionality for scientific and engineering applications on low-power architectures. With the aim of showing the benefits of this contribution, two applications, coming from the image processing and dynamic molecular simulation fields, have been chosen. European Cooperation in Science and Technology. COST

The synthesis and characterization of a series of novel small-molecule hole-transporting materials (HTMs) based on an anthra[1,2-b:4,3-b′:5,6-b′′:8,7-b′′′]tetrathiophene (ATT) core are reported. The new compounds follow an easy synthetic route and have no need of expensive purification steps. The novel HTMs were tested in perovskite solar cells (PSCs) and power conversion efficiencies (PCE) of up to 18.1 % under 1 sun irradiation were 2 measured. This value is comparable with the 17.8 % efficiency obtained using spiroOMeTAD as a reference compound. Similarly, a significant quenching of the Photoluminescence in the first nanosecond is observed, indicative of effective hole transfer.Additionally, the influence of introducing aliphatic alkyl chains acting as solubilizers on the device performance of the ATT molecules is investigated. Replacing the methoxy groups on the triarylamine sites by butoxy-, hexoxy- or decoxy-substituents greatly improved the solubility of the compounds without changing the energy levels, yet at the same time significantly decreasing the conductivity as well as the PCE, 17.3 % for ATT-OBu, 15.7 % for ATT-OHex and 9.7 % for ATT-ODec.

Dealing with asymmetry in the architecture opens a plethora of questions from the perspective of scheduling task-parallel applications, and there exist early attempts to address this problem via ad-hoc strategies embedded into a runtime framework. In this paper we take a different path, which consists in addressing the complexity of the problem at the library level, via a few asymmetry-aware fundamental kernels, hiding the architecture heterogeneity from the task scheduler. For the specific domain of dense linear algebra, we show that this is not only possible but delivers much higher performance than a naive approach based on an asymmetry-oblivious scheduler. Furthermore, this solution also outperforms an ad-hoc asymmetry-aware scheduler furnished with sophisticated scheduling techniques.

El día 1 de octubre de 2014 se celebró en la Feria Valencia, “Vegetal World”, una Jornada Técnica sobre “Inspección de equipos de aplicación de fitosanitarios", (ITEAF). La jornada estaba organizada por la Universidad Politécnica de Valencia (UPV) y el Instituto Valenciano de Investigaciones Agrarias (IVIA), en colaboración con el Ministerio de Agricultura, Alimentación y Medio Ambiente, MAGRAMA, la Consellería de Agricultura, Pesca, Alimentación y Agua, la Asociación de Fabricantes de Maquinaria Agrícola de la Comunidad Valencia, ASMAVA, y la Asociación Nacional de Maquinaria Agropecuaria, Forestal y de Espacios Verdes, ANSEMAT. El objetivo de esta jornada técnica era dar a conocer la normativa para la inspección técnica de los equipos de aplicación de tratamientos, sus repercusiones a nivel de usuario y las condiciones de su aplicación.

El uso de las TIC en agricultura está permitiendo mejorar la eficiencia de algunas de sus operaciones. En concreto, la protección de cultivos se puede ver beneficiada por el uso de sensores para la caracterización electrónica de las copas de los cultivos arbóreos, por la generación de herramientas de soporte a la toma de decisiones y por la utilización de tecnologías de dosificación variable. El objetivo final de estas tecnologías es reducir al máximo las pérdidas de productos fitosanitarios mejorando la competitividad de las explotaciones, la calidad de sus productos y reduciendo al máximo el impacto que generan en su entorno.

Entre los objetivos formativos de los cursos avanzados de arquitectura de computadores suele estar el de que los estudiantes sean capaces de describir y analizar el funcionamiento de los protocolos de coherencia de cachés. Aunque dichos protocolos son relativamente sencillos, es necesario analizar muchas situaciones diferentes para entender cómo abordan todos los detalles del problema que quieren resolver. Lo que hace que sean complejos de explicar y de comprender. Una herramienta que ilustrara gráficamente el funcionamiento de dichos protocolos facilitaría enormemente su enseñanza/aprendizaje. Con objeto de mejorar la docencia de dicha materia, hemos desarrollado tres animaciones interactivas que muestran cómo funcionan tres de los protocolos de coherencia de caché más frecuentemente utilizados. Para cada protocolo, una serie de operaciones de lectura/escritura ilustran todas las posibles situaciones que pueden darse. Las animaciones permiten avanzar y retroceder para poder entender/estudiar mejor las acciones que tienen lugar en cada paso. SUMMARY: Among the educational objectives in advanced courses of computers architecture there is usually one that states that students should be able to describe and analyze how the cache coherence protocols work. Although these protocols are relatively simple, it is necessary to analyze many different situations to understand how they address all the details of the problem they solve. This makes them complex to be explained and to be understood. A tool that illustrates graphically the operation of these protocols should greatly facilitate the teaching/learning of these protocols. With the aim of improving the teaching on this subject, we have developed three interactive animations that show how some of the most frequently used cache coherence protocols work. For each protocol, a sequence of read and write operations illustrates all possible situations that can take place in each protocol. The tool is interactive in that the student can go forward and backward to understand/ study the different actions that occur at each step.

El calculo del rango numerico matricial surge en numerosas aplicaciones de la ciencia y de la ingenieria. Actualmente existen tres aproximaciones numericas basicas para efectuar este calculo: la descomposicion SVD, la descomposicion URV y las descomposiciones QR reveladoras de rango (QRRR). En este trabajo se analizan experimentalmente varios algoritmos secuenciales, basados en las tres aproxima- ciones anteriores para el calculo del rango numerico matricial. Asi, en el estudio comparativo experimental se emplea una implementacion propia para el calculo de la descomposicion URV y dos nuevas rutinas para el calculo de la descomposicion QRRR. Ademas se utilizan las rutinas de la libreria LAPACK para el calculo de la descomposicion SVD y la descomposicion QR con pivotamiento de columnas. Los resultados experimentales muestran que la descomposicion QRRR es en la practica tan fiable como las costosas descomposiciones SVD y URV. Ademas, estas descomposiciones QRRR presentan la ventaja fundamental de su bajo coste computacional.

The promise of future many-core processors, with hundreds of threads running concurrently, has led the developers of linear algebra libraries to rethink their design in order to extract more parallelism, further exploit data locality, attain better load balance, and pay careful attention to the critical path of computation. In this paper we describe how existing serial libraries such as (C)LAPACK and FLAME can be easily parallelized using the SMPSs tools, consisting of a few OpenMP-like pragmas and a runtime system. In the LAPACK case, this usually requires the development of blocked algorithms for simple BLAS-level operations, which expose concurrency at a finer grain. For better performance, our experimental results indicate that column-major order, as employed by this library, needs to be abandoned in benefit of a block data layout. This will require a deeper rewrite of LAPACK or, alternatively, a dynamic conversion of the storage pattern at run-time. The parallelization of FLAME routines using SMPSs is simpler as this library includes blocked algorithms (or algorithms-by-blocks in the FLAME argot) for most operations and storage-by-blocks (or block data layout) is already in place.

A flexible, saddle-like, π-conjugated skeleton composed of four fused thiophene rings forming a cyclooctatetrathiophene (CoTh) with four triphenylamines (CoTh-TTPA) is presented as a hole-transporting material (HTM) for perovskite solar cells. The new HTM shows a bright red color stemming from a direct conjugation between the TPA groups and the central CoTh scaffold. This results in a charge transfer band due to the combination of the weak acceptor moiety, the CoTh unit, and the electron-donating p-methoxytriphenylamine groups. CoTh-TTPA exhibits a suitable highest-occupied molecular orbital (HOMO) level in relation to the valence band edge of the perovskite, which ensures efficient hole extraction at the perovskite/HTM interface. It has been applied as the HTM in combination with a mixed perovskite ([FAPbI3]0.85[MAPbBr3]0.15) and a state-of-the-art triple cation perovskite ([(FAPbI3)0.87(MAPbBr3)0.13]0.92[CsPbI3]0.08) reaching noticeable light-to-energy conversion efficiencies of 16.3 and 15.9%, respectively. These values are slightly lower than those measured for the benchmark spiro-OMeTAD HTM. The HTM properties have been analyzed by means of photoluminescence and conductivity experiments, which demonstrated a better hole extraction and conductivity for spiro-OMeTAD.

Two new hole‐transporting materials (HTMs), BX‐OMeTAD and BTX‐OMeTAD , based on xanthene and thioxanthene units, respectively, and bearing p ‐methoxydiphenylamine peripheral groups, are presented for their use in perovskite solar cells (PSCs). The novelty of the newly designed molecules relies on the use of a single carbon‐carbon bond ‘C−C’ as a linker between the two functionalized heterocycles, which increases the flexibility of the molecule compared with the more rigid structure of the widely used HTM spiro‐OMeTAD. The new HTMs display a limited absorbance in the visible region, due to the lack of conjugation between the two molecular halves, and the chemical design used has a remarkably impact on the thermal properties when compared to spiro‐OMeTAD. BX‐OMeTAD and BTX‐OMeTAD have been tested in ([(FAPbI3)0.87(MAPbBr3)0.13]0.92[CsPbI3]0.08)‐based PSC devices exhibiting power conversion efficiencies of 14.19 and 16.55 %, respectively. The efficiencies reached, although lower than those measured for spiro‐OMeTAD (19.63 %), are good enough to consider the chemical strategy used as an interesting via to design HTMs for PSCs.

We investigate the use of the SMPSs programming model to leverage task parallelism in the execution of a message-pas sing implementation of the symmetric rank- k update on clusters equipped with multicore processors. Our experience shows that the major difficulties to adapt the code to the MPI/SMPSs instance of this programming model are due to the usage of the conventional column-major layout of matrices in numerical libraries. On the other hand, the experimental results show a considerable increase in the performance and scalability of our solution when compared with the standard options based on the use of a pure MPI approach or a hybrid one that combines MPI/multi-threaded BLAS.

With the breakdown of Dennard scaling in the mid-2000s and the end of Moore's law on the horizon, the high performance computing community is turning its attention towards unconventional accelerator hardware to ensure the continued growth of computational capacity. This dissertation presents several contributions related to the iterative solution of sparse linear systems on the most widely used general purpose accelerator - the Graphics Processing Unit (GPU). Specifically, it accelerates the major building blocks of Krylov solvers, and describes their realization as part of a software library of reusable building blocks. The first part of the dissertation focuses on the sparse matrix-vector product and effective load balancing in the presence of irregular sparsity patterns. The second part describes the design of high-performance preconditioners. Finally, the third part demonstrates the potential of adaptive precision techniques for constructing preconditioners with lower memory footprint, and accuracy comparable to their full precision equivalents. Con el final de la ley de Dennard y el cercano fin de la ley de Moore, la comunidad en computación de altas prestaciones se está centrando en tecnologías de aceleración no convencionales para asegurar el crecimiento exponencial de la capacidad de computación. Esta tesis contribuye a la solución iterativa de sistemas lineales dispersos en el acelerador más difundido: el procesador gráfico. Específicamente, el trabajo acelera los bloques fundamentales de los métodos de Krylov, y describe su implementación como parte de una biblioteca de bloques reutilizables. La primera parte del trabajo se centra en el producto matriz-vector disperso y el equilibrado de la carga ante patrones de dispersidad irregulares. La segunda parte describe el diseño de precondicionadores de alto rendimiento. Finalmente, la tercera parte demuestra el potencial de las técnicas de precisión adaptativa para construir precondicionadores con menor consumo de memoria, y fiabilidad comparable con las versiones de precisión completa. Programa de Doctorat en Informàtica

Hoy en dia, en las aplicaciones cientificas y de ingenieria surgen con mucha frecuencia problemas de resolucion de sistemas de ecuaciones lineales dispersos de gran dimension. El incesante crecimiento de los tamanos de estos problemas, entre otros factores, propicia un creciente incremento del consumo energetico en los grandes centros de calculo. Las restricciones energeticas imperantes, y la concienciacion con el respeto al medio ambiente, propicio la aparicion de Green Computing, con el objetivo de disenar aplicaciones conscientes de la energia y considerar la eficiencia energetica (GFLOPS/W) como un parametro prioritario. Tradicionalmente, se ha considerado que el rendimiento es directamente proporcional al consumo energetico, pero la aplicacion de estas nuevas tecnicas permite incrementar el rendimiento que requieren las aplicaciones cientificas con un consumo energetico ajustado. Los objetivos de esta Tesis Doctoral se enfocan hacia el estudio y desarrollo de tecnicas de ahorro de energia en sistemas de computo heterogeneo, CPU-GPU, para disenar aplicaciones que consigan minimizar el consumo energetico sin sacrificar el rendimiento, en la resolucion de sistemas lineales dispersos mediante metodos iterativos. Para ello se explota que las GPUs muestran un ahorro de energia muy significativo cuando se sincronizan con la CPU en modo blocking, aunque su rendimiento disminuye de forma considerable. Por el contrario, las GPUs alcanzan maximas prestaciones cuando estan sincronizadas en modo polling, pero ello supone un sumidero de energia por parte de la CPU al permanecer en un estado activo. El diseno de “tecnicas de fusionado de kernels CUDA”, propicia una reduccion del numero de kernels, eliminandose tiempos de lanzamiento de kernels, y tiempos de transferencia de informacion entre los espacios de memoria de la CPU y la GPU. Ademas, si la GPU esta sincronizada en modo blocking, la reduccion del numero de kernels tambien decrementa el numero de cambios de un estado de bajo consumo energetico de la CPU hasta otro activo, eliminando los correspondientes tiempos de espera de recuperacion de la CPU, que permite que el coste de ejecucion del modo blocking sea muy parecido al del modo polling. Debido a su complejidad, las “tecnicas de fusionado de kernels CUDA”, solo son utilizadas por programadores expertos, por lo que el diseno de una metodologia ayudara de forma sistematica al proceso de “fusionado de kernels CUDA”. Esta tarea es la principal aportacion de esta Tesis Doctoral, extendiendo su aplicacion sobre arquitecturas de GPU que soporten paralelismo dinamico, que permite optimizar resolutores para que puedan ejecutarse de forma desacoplada de la CPU. Los resultados experimentales obtenidos validan mejoras destacables tanto en rendimiento como en eficiencia energetica, obteniendo un compromiso entre rendimiento y consumo energetico constante y equilibrado.