Pasar al contenido principal

UPC-COMPUTACION DE ALTAS PRESTACIONES VIII

PID2019-107255GB-C22

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2019
Plan Estatal 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITECNICA DE CATALUNYA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 92
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Contribución de congreso (ConferenceOutput)

How2Sign: A large-scale multimodal dataset for continuous American sign language

  • Duarte, Amanda
  • Palaskar, Shruti
  • Ventura, Lucas
  • Ghadiyaram, Deepti
  • DeHaan, Kenneth
  • Metze, Florian
  • Torres, Jordi
  • Giró i Nieto, Xavier
Trabajo presentado en la IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), celebrada de forma virtual del 19 al 25 de junio de 2021, One of the factors that have hindered progress in the areas of sign language recognition, translation, and production is the absence of large annotated datasets. Towards this end, we introduce How2Sign, a multimodal and multiview continuous American Sign Language (ASL) dataset, consisting of a parallel corpus of more than 80 hours of sign language videos and a set of corresponding modalities including speech, English transcripts, and depth. A three-hour subset was further recorded in the Panoptic studio enabling detailed 3D pose estimation. To evaluate the potential of How2Sign for real-world impact, we conduct a study with ASL signers and show that synthesized videos using our dataset can indeed be understood. The study further gives insights on challenges that computer vision should address in order to make progress in this field. Dataset website: http://how2sign.github.io/, This work received funding from Facebook through gifts to CMU and UPC; through projects TEC2016-75976-R, TIN2015- 65316-P, SEV-2015-0493 and PID2019-107255GB-C22 of the Spanish Government and 2017-SGR-1414 of Generalitat de Catalunya. This work used XSEDE’s “Bridges” system at the Pittsburgh Supercomputing Center (NSF award ACI1445606). Amanda Duarte has received support from la Caixa Foundation (ID 100010434) under the fellowship code LCF/BQ/IN18/11660029. Shruti Palaskar was supported by the Facebook Fellowship program.
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Energy optimization and analysis with EAR

  • Corbalán González, Julita|||0000-0002-3926-5634
  • Alonso Jane, Lluís
  • Aneas Gómez, Jordi
  • Brochard, Luigi
EAR is an energy management framework which offers three main services: energy accounting, energy control, and energy optimization. The latter is done through the EAR runtime library (EARL). EARL is a dynamic, transparent, and lightweight runtime library that provides energy optimisation and control. EARL optimises energy by selecting the optimal CPU frequency, based on the energy policy selected and application runtime characteristics without any application modification or user input. Currently EARL only works for MPI applications but EAR itself can still operate for non-MPI applications. It automatically (and transparently) identifies iterative regions (loops) and computes a set of metrics per iteration, application signature, and, together with the system signature, applies energy models to estimate the execution time and power for the CPU frequencies available. System signature is a set of coefficients per-node computed during EAR installation via a learning phase. Given time and power projections, EARL selects the best frequency based on policy settings. This papers shows how to optimize energy using the EAR library with min_time_to_solution energy policy and how to analyse applications through EAR framework. Evaluation includes eight applications with different sizes and application signatures. Results show how EARL computes each application signature on the fly and applies the CPU frequency selected by the min_time_to_solution policy., This work has been mostly funded by the BSC-Lenovo collaboration agreement and partially by the Spanish Government through Programa Severo Ochoa (SEV-2015-0493), by the Spanish Ministry of Science and Technology through PID2019-107255GB project, by the Generalitat de Catalunya (2017- SGR-1414), Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

HyCSim: A rapid design space exploration tool for emerging hybrid last-level caches

  • Escuín Blasco, Carlos
  • Ali Khan, Asif
  • Ibáñez Marín, Pablo Enrique
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Viñals Yúfera, Victor
  • Castrillón, Jerónimo
Recent years have seen a rising trend in the exploration of non-volatile memory (NVM) technologies in the memory subsystem. Particularly in the cache hierarchy, hybrid last-level cache (LLC) solutions are proposed to meet the wide-ranging performance and energy requirements of modern days applications. These emerging hybrid solutions need simulation and detailed exploration to fully understand their capabilities before exploiting them. Existing simulation tools are either too slow or incapable of prototyping such systems and optimizing for NVM devices. To this end, we propose HyCSim, a trace-driven simulation infrastructure that enables rapid comparison of various hybrid LLC configurations for different optimization objectives. Notably, HyCSim makes it possible to quickly estimate the impact of various hybrid LLC insertion and replacement policies, disabling of a cache region at byte or cache frame granularity for different fault maps. In addition, HyCSim allows to evaluate the impact of various compression schemes on the overall performance (hit and miss rate) and the number of writes to the LLC. Our evaluation on ten multi-program workloads from the SPEC 2006 benchmarks suite shows that HyCSim accelerates the simulation time by 24×, compared to the cycle-accurate Gem5 simulator, with high-fidelity., This work was partially funded by the HiPEAC collaboration grant 2020, the German Research Council (DFG) through the TraceSymm project (366764507) and the Co4RTM project (450944241), MCIN/AEI/10.13039/501100011033 (grants PID2019-105660RB-C21 and PID2019- 107255GB-C22), and by Aragón Government (T5820R research group)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Hypervisor extension for a RISC-V processor

  • Gauchola Vilardell, Jaume
  • Costa Prats, Juan José|||0000-0003-2479-0230
  • Morancho Llena, Enrique|||0000-0003-2403-8145
  • Canal Corretger, Ramon|||0000-0003-4542-204X
  • Carril Gil, Xavier|||0009-0008-8392-7579
  • Doblas Font, Max|||0000-0002-8909-3033
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Pajuelo González, Manuel Alejandro|||0000-0002-5510-6860
  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Salamero Sanz, Javier
  • Verdú Mulà, Javier|||0000-0003-4485-2419
This paper describes our experience implementing a Hypervisor extension for a 64-bit RISC-V processor. We describe the design process and the main required parts with a brief explanation of each one., This work is partially supported by the DRAC (IU16-011591), the HORIZON Vitamin-V (101093062) and the Computación de Altas Prestaciones VIII (PID2019-107255GB) projects., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Parallel and distributed protein processing for 3D-protein pattern discovery and clustering

  • Valdés Jiménez, Alejandro Mauricio
  • Núñez Vivanco, Gabriel
  • Jiménez González, Daniel|||0000-0001-6064-7883
The discovery and clustering of three-dimensional protein patterns in a set of protein structures, without relying on predefined search patterns, can be highly beneficial for predicting the functions of unknown proteins and facilitating rational multi-target drug design. This work introduces a novel OpenMP parallelization of the 3D-PP algorithm using explicit and nested tasks, which balances data sharing synchronization and load unbalance, improving previous implementations based on implicit tasks. Given the vast number of protein structures available in the Protein Data Bank (over 231,000 from PDB and more than 1,068,000 from AlphaFold), processing large datasets locally can be constrained by available resources such as main memory and secondary storage. To address this challenge, we propose parallel approaches to the 3D-PP algorithm for distributed memory systems (using MPI) and hybrid systems (combining MPI and OpenMP). The evaluated strategies distribute the workload based on the entire protein structure. Experimental results using a dataset of 8,344 protein structures show that the new OpenMP taskified version is 1.6x faster than the previous best OpenMP implementation. Moreover, the hybrid MPI+OpenMP version achieves a speedup of up to 162.5x, demonstrating its scalability and efficiency for large-scale protein structure analysis., This research was funded by the Spanish Government (Grants PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007). We also want to thank to the Barcelona Supercomputing Center for the use of the CTE-AMD cluster., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Pronóstico de capacidad efectiva y prestaciones en una cache no volátil de último nivel

  • Escuín Blasco, Carlos
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Llaberia Griñó, José M.|||0000-0002-3753-4108
  • Ibáñez Marín, Pablo Enrique
  • Viñals Yúfera, Victor
La degradación debida a las escrituras que sufren las bitcells implementadas con tecnologi´as de memoria no volátil (NVM) es uno de los principales escollos que se presentan a la hora de construir la cache de último nivel (LLC) con estas tecnologi´as. Aunque en la literatura se recogen diferentes propuestas para hacer frente a esta degradación, la metodologi´a usada en los trabajos previos no permite estudiar en detalle la evolución de la capacidad efectiva ni de las prestaciones en una cache no volátil de último nivel (NV-LLC). Por ello, en este trabajo se propone un procedimiento de pronóstico que combina simulación y predicción con el objetivo de estudiar dicha evolución. Por otra parte, la compresión es una de las técnicas propuestas en la literatura para lidiar con la degradación de las memorias no volátiles. En primer lugar, la compresión diminuye la cantidad de información escrita en una NV-LLC. En segundo lugar, cuando los contenedores ven mermada su capacidad debido a la degradación, la compresión permite mantener su funcionalidad albergando bloques de tamaño reducido. El procedimiento de pronóstico desarrollado en este trabajo permite evaluar el impacto de diferentes técnicas y mecanismos de gestión de contenidos en la esperanza de vida y las prestaciones de una NV-LLC de manera detallada. El mecanismo de compresión adoptado en este trabajo multiplica hasta por 5 veces la esperanza de vida de una NV-LLC., Este trabajo ha sido financiado por MINECO/AEI/FEDER (UE) (proyectos PID2019-105660RB-C21 y PID2019-107255GB-C22 / AEI / 10.13039/501100011033), el Gobierno de Aragon (Grupo T58 20R) y FEDER 2014-2020 “Construyendo Europa desde Aragón”., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Empirical evidence for MPSoCs in critical systems: The case of NXP’s T2080 cache coherence

  • Pujol Torramorell, Roger
  • Tabani, Hamid
  • Abella Ferrer, Jaume|||0000-0001-7951-4028
  • Hassan, Mohamed
  • Cazorla Almeida, Francisco Javier
The adoption of complex MPSoCs in critical real-time embedded systems mandates a detailed analysis their architecture to facilitate certification. This analysis is hindered by the lack of a thorough understanding of the MPSoC system due to the unobvious and/or insufficiently documented behavior of some key hardware features. Confidence on those features can only be regained by building specific tests to both, assess whether their behavior matches specifications and unveil their behavior when it is not fully known a priori. In this work, we introduce a systematic approach that constructs this thorough understanding of the MPSoC architecture-- and assess against its specification in processor documentation -- with a focus on the cache coherence protocol in the avionics-relevant NXP T2080 architecture as our use-case. Our approach covers all transitions in the MESI cache coherence protocol, with emphasis on the coherence between DMA and processing cores. We build evidence of their behavior based on available debug support and performance monitors. Our analysis discloses unexpected behavior for coherence-related notifications as well as some hardware monitors., This work has been partially supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB; the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 878752 (MASTECS) and the European Research Council (ERC) grant agreement No. 772773 (SuPerCom); the HiPEAC Network of Excellence; and the Natural Sciences and Engineering Research Council of Canada (NSERC)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Task-based programming models for heterogeneous recurrent workloads

  • Bosch Pons, Jaume|||0000-0002-4040-3416
  • Vidal, Miquel|||0000-0002-1973-8289
  • Filgueras Izquierdo, Antonio|||0000-0001-6071-3254
  • Jiménez González, Daniel|||0000-0001-6064-7883
  • Álvarez Martínez, Carlos|||0000-0003-0536-5183
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
This paper proposes the extension of task-based programming models with recurrent workloads concepts. The proposal introduces new clauses in the OmpSs task directive to efficiently model recurrent workloads. The clauses define the task period and/or the number of task body repetitions. Despite the new clauses are suitable for any device, their support has been implemented using the capabilities of FPGA devices in embedded systems. These heterogeneous systems are common in industrial applications that usually develop recurrent workloads. The evaluation shows a huge gap in the applications' programmability, saving lines of code, and increasing the code readability. Besides, it shows the efficient management of recurrent tasks when performed in FPGA devices, which can support one order of magnitude finer tasks. All these improvements perfectly suit the needs of cyber-physical heterogeneous systems, which are frequently used in industrial environments to run recurrent workloads., This work is partially supported by the European Union H2020 Research and Innovation Action (project 801051), by the Spanish Government (projects SEV-2015-0493 and PID2019-107255GB, grant BES-2016-078046), and by the Generalitat de Catalunya (contracts 2017-SGR-1414 and 2017-SGR-1328)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

How2Sign: A large-scale multimodal dataset for continuous American sign language

  • Cardoso Duarte, Amanda
  • Palaskar, Shruti
  • Ventura Ripol, Lucas
  • Ghadiyaram, Deepti
  • DeHaan, Kenneth
  • Metze, Florian
  • Torres Viñals, Jordi|||0000-0003-1963-7418
  • Giró Nieto, Xavier|||0000-0002-9935-5332
One of the factors that have hindered progress in the areas of sign language recognition, translation, and production is the absence of large annotated datasets. Towards this end, we introduce How2Sign, a multimodal and multiview continuous American Sign Language (ASL) dataset, consisting of a parallel corpus of more than 80 hours of sign language videos and a set of corresponding modalities including speech, English transcripts, and depth. A three-hour subset was further recorded in the Panoptic studio enabling detailed 3D pose estimation. To evaluate the potential of How2Sign for real-world impact, we conduct a study with ASL signers and show that synthesized videos using our dataset can indeed be understood. The study further gives insights on challenges that computer vision should address in order to make progress in this field. Dataset website: http://how2sign.github.io/, This work received funding from Facebook through gifts to CMU and UPC; through projects TEC2016-75976-R, TIN2015- 65316-P, SEV-2015-0493 and PID2019-107255GB-C22 of the Spanish Government and 2017-SGR-1414 of Generalitat de Catalunya. This work used XSEDE’s “Bridges” system at the Pittsburgh Supercomputing Center (NSF award ACI- 1445606). Amanda Duarte has received support from la Caixa Foundation (ID 100010434) under the fellowship code LCF/BQ/IN18/11660029. Shruti Palaskar was supported by the Facebook Fellowship program., Peer Reviewed, Objectius de Desenvolupament Sostenible::10 - Reducció de les Desigualtats, Objectius de Desenvolupament Sostenible::4 - Educació de Qualitat::4.5 - Per a 2030, eliminar les disparitats de gènere en l’educació i garantir l’accés en condicions d’igualtat a les persones vulnerables, incloses les persones amb discapacitat, els pobles indígenes i els nens i nenes en situacions de vulnerabilitat, a tots els nivells de l’ensenyament i la formació professional, Objectius de Desenvolupament Sostenible::10 - Reducció de les Desigualtats::10.2 - Per a 2030, potenciar i promoure la inclusió social, econòmica i política de totes les persones, independentment de l’edat, sexe, discapacitat, raça, ètnia, origen, religió, situació econòmica o altra condició, Objectius de Desenvolupament Sostenible::4 - Educació de Qualitat
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

AI-powered personalised learning platforms for EFL learning: preliminary results

  • Folgieri, Raffaella
  • Gil, Marisa|||0000-0001-9755-3311
  • Bait, Miriam
  • Lucchiari, Claudio
Artificial intelligence (AI) has been increasingly integrated into the field of education, including personalised learning platforms. However, concerns have been raised about the potential of AI to replace human teachers and the impact on student agency and autonomy. In this research, we discuss the development of an AI-powered platform as a helper, not a substitution, for self-directed personal and professional growth. The present study investigates the effectiveness of an AI-powered personalised learning platform in enhancing self-directed learning and personal and professional growth. We also explored the role of human teaching and the ethical considerations of AI in education. A mixed-methods approach was used, including surveys, interviews, and qualitative analysis of participant feedback. The participants were randomly assigned to either an AI group or the traditional learning group. Findings suggest that the AI-powered personalised learning platform that we used is a promising approach for enhancing self-directed learning and personal and professional growth. However, it is important to note that these are just preliminary findings, and further research is needed to confirm our results and to understand the mechanisms by which a specific use of AI in education may lead to positive effects., This research was funded by the Department of Philosophy “Piero Martinetti” of the University of Milan under the Project “Departments of Excellence 2023-2027” awarded by the Ministry of University and Research (MUR). Marisa Gil's work has been supported by the Spanish Ministry of Education (PID2019-107255GBC22) and the Generalitat de Catalunya (2021-SGR01007)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

On the relation between open project-based learning in undergraduate computer science education and contemporary technological trends

  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Freitag, Fèlix|||0000-0001-5438-479X
  • Berral García, Josep Lluís|||0000-0003-3037-3580
Amidst rapid and constant technological change, keeping IT higher education curricula up to date is becoming increasingly challenging. For a decade, the course”Project on Information Technologies” within the undergraduate computer science studies offered by a major technical university has been pursuing an effort of continuous curriculum adaptation and active learning practices based on an open project-based learning (PBL) methodology. This article researches the implications of employing an open-ended PBL approach, with a specific focus on the alignment of acquired skills with the trends in the professional landscape. Our analysis has identified a strong correlation between the technologies utilized in the projects and the contemporary technological trends in the areas of global technological focus, programming languages, server-side technologies, database management systems, and DevOps-related tools. For the study, we analyzed empirical data gathered from over 100 projects involving more than 400 students who enrolled in this reference course, belonging to the last year of the IT higher education programme in the last 10 years. The results suggest the open project-based design as a teaching means in the student’s study plan for fostering the student the learning of prevailing current practical technologies., This work is partially supported by the Span-ish Ministry of Science under contracts PID2019-107255GB, PID2019-106774RB-C21, PID2021-126248OB-I00 (MCIN/AEI/10.13039/501100011033/FEDER), PDC2023-145809-I00 (PDC/AEI/10.13039/501100011033), the Recovery and Resilience Facility of the European Union, and by the SGR programme of the Catalan Government under contracts 2021-SGR-00478 and 2021-SGR-01059., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Efficiently running SpMV on long vector architectures

  • Gómez Crespo, Constantino
  • Mantovani, Filippo|||0000-0003-3559-4825
  • Focht, Erich
  • Casas, Marc|||0000-0003-4564-2093
Sparse Matrix-Vector multiplication (SpMV) is an essential kernel for parallel numerical applications. SpMV displays sparse and irregular data accesses, which complicate its vectorization. Such difficulties make SpMV to frequently experiment non-optimal results when run on long vector ISAs exploiting SIMD parallelism. In this context, the development of new optimizations becomes fundamental to enable high performance SpMV executions on emerging long vector architectures. In this paper, we improve the state-of-the-art SELL-C-s sparse matrix format by proposing several new optimizations for SpMV. We target aggressive long vector architectures like the NEC Vector Engine. By combining several optimizations, we obtain an average 12% improvement over SELL-C-s considering a heterogeneous set of 24 matrices. Our optimizations boost performance in long vector architectures since they expose a high degree of SIMD parallelism., The authors would like to acknowledge the support of NEC Corporation. This work is partially supported by the Spanish Ministry of Science and Technology through PID2019-107255GB project and by the Generalitat de Catalunya (contract 2017-SGR-1414). Marc Casas has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramon y Cajal fellowship number RYC-2017-23269., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

An FPGA accelerator of the wavefront algorithm for genomics pairwise alignment

  • Haghi, Abbas
  • Marco Sola, Santiago|||0000-0001-7951-3914
  • Álvarez Martí, Lluc|||0000-0003-0506-8867
  • Diamantopoulos, Dionysios
  • Hagleitner, Christoph
  • Moretó Planas, Miquel|||0000-0002-9848-8758
In the last years, advances in next-generation sequencing technologies have enabled the proliferation of genomic applications that guide personalized medicine. These applications have an enormous computational cost due to the large amount of genomic data they process. The first step in many of these applications consists in aligning reads against a reference genome. Very recently, the wavefront alignment algorithm has been introduced, significantly reducing the execution time of the read alignment process. This paper presents the first FPGA- based hardware/software co-designed accelerator of such relevant algorithm. Compared to the reference WFA CPU-only implementation, the proposed FPGA accelerator achieves performance speedups of up to 13.5× while consuming up to 14.6× less energy.ed medicine. These applications have an enormous computational cost due to the large amount of genomic data they process. The first step in many of these applications consists in aligning reads against a reference genome. Very recently, the wavefront alignment algorithm has been introduced, significantly reducing the execution time of the read alignment process. This paper presents the first FPGA- based hardware/software co-designed accelerator of such relevant algorithm. Compared to the reference WFA CPU-only imple- mentation, the proposed FPGA accelerator achieves performance speedups of up to 13.5× while consuming up to 14.6× less energy., This work has been supported by the European HiPEAC Network of Excellence, by the Spanish Ministry of Science and Innovation (contract PID2019-107255GB-C21/AEI/10.13039/501100011033), by the Generalitat de Catalunya (contracts 2017-SGR-1414 and 2017-SGR-1328), by the IBM/BSC Deep Learning Center initiative, and by the DRAC project, which is co-financed by the European Union Regional Development Fund within the framework of the ERDF Operational Program of Catalonia 2014-2020 with a grant of 50% of total eligible cost. Ll. Alvarez has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under the Juan de la Cierva Formacion fellowship No. FJCI-2016-30984. M. Moreto has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramon y Cajal fellowship No. RYC-2016-21104., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

GPU4S: Major project outcomes, lessons learnt and way forward

  • Kosmidis, Leonidas
  • Rodríguez Ferrández, Iván
  • Jover Álvarez, Álvaro
  • Alcaide Portet, Sergi|||0000-0001-8561-7795
  • Lachaize, Jérôme
  • Notebaert, Olivier
  • Certain, Antoine
  • Steenari, David
Embedded GPUs have been identified from both private and government space agencies as promising hardware technologies to satisfy the increased needs of payload processing. The GPU4S (GPU for Space) project funded from the European Space Agency (ESA) has explored in detail the feasibility and the benefit of using them for space workloads. Currently at the closing phases of the project, in this paper we describe the main project outcomes and explain the lessons we learnt. In addition, we provide some guidelines for the next steps towards their adoption in space., This work is funded by ESA under the GPU4S (GPU for Space) project (ITT AO/1-9010/17/NL/AF). It is also partially supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under grants PID2019-107255GB and FJCI-2017-34095 and HiPEAC., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Partitioned reduction for heterogeneous environments

  • De Rango, Alessio
  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Gil, Marisa|||0000-0001-9755-3311
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
  • Giordano, Andrea
  • D'Ambrosio, Donato
  • Mendicino, Giuseppe
Nowadays, performance in HPC applications focuses on MPI efficiency as the de facto message-passing library to exploit parallelism. Features such as multithread and communication and processing overlap are continuously studied to adapt to new platforms and a more significant number of processing units like GPU platforms. In this sense, recently, the MPI-4.0 standard introduced the partitioned point-to-point communication primitives to potentiate computation and communication overlapping. This paper introduces an innovative extension to MPI, specifically addressing partitioned communication for MPI-reduction primitives. Traditional reduction tasks conventionally involve processing the complete input vector following the conclusion of GPU computations. In contrast, our proposed methodology exploits message partitioning to process reduction tasks in real-time incrementally. This approach allows the system to process individual partitions of the input vector as they become available, removing the necessity to await the full completion of GPU computations before initiating the reduction. Our results demonstrate promising benefits, particularly for large message sizes. However, it is essential to acknowledge that optimizations at synchronization points remain potential bottlenecks, requiring meticulous analysis and consideration., This work was supported by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR-01007). This work was partially supported by PON ”Research and Innovation” 2014-2020 CUP H25F21001220006, C. I. 1062 R22 INNOVAZIONE, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Improving the discovery and clustering of three-dimensional protein patterns with OpenMP

  • Valdés Jiménez, Alejandro Mauricio
  • Reyes Parada, Miguel
  • Núñez Vivanco, Gabriel
  • Durán-Verdugo, Fabio
  • Jiménez González, Daniel|||0000-0001-6064-7883
The discovery of conserved three-dimensional (3D) amino-acid patterns among a set of protein structures can be useful, for instance, to predict the functions of unknown proteins or for the rational design of multi-target drugs. There are several applications that perform a three-dimensional search of patterns in the structures of proteins. However, discovering conserved 3D patterns in a set of proteins with no other baseline patterns is a challenge. In this paper, we analyze and improve a state-of-the-art algorithm, 3D-PP, that implements this discovery. In this algorithm, the 3D patterns are detected and clustered using the root mean square deviation value, measured among each pair of 3D patterns (topological variability indicator). Even when 3D-PP deals with this task, the simultaneous processing of high amounts of proteins becomes a computational challenge with the size and the number of proteins to be evaluated. In this work, we present and analyze different shared memory parallel strategies of 3D-PP, using OpenMP. Those strategies improve the overall performance of the original implementation by reducing parallel load unbalance among threads and overall increasing parallelism. The results show significant performance improvements compared to the original version, achieving up to 13x speedup for a small number of proteins and 17.7× for a larger set., This research was funded by DICYT-USACH grant 5392102RP-AC and the Fondo Nacional de Desarrollo Cientifico y Tecnologico (FONDECYT) grants 1191133, Proyecto Fondecyt (Chile) 1220656 (MR-P), and by the Spanish Government (Grants PCI2021-121964 - TEXTAROSSA, PID2019-107255GB-C21/AEI/10.13039/501100011033, PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007). Additional tests were performed in the cluster obtained with the grant CONICYT-FONDEQUIP-EQM160063., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Improving the discovering and comparing of three-dimensional protein patterns

  • Valdés Jiménez, Alejandro Mauricio
  • Núñez Vivanco, Gabriel
  • Jiménez González, Daniel|||0000-0001-6064-7883
The discovery and comparison of three-dimensional (3D) patterns in protein structures are critical tasks in structural bioinformatics, particularly for identifying functional sites and potential drug targets. In this work, we present significant enhancements of Geomfinder, a ligand-independent algorithm for identifying and comparing 3D protein patterns. Our contribu- tions include the following key improvements: i) a mechanism that reduces the number of unnecessary site comparisons; ii) an optimization that lowers the computational complexity of partial score calculations from O(N×M) to O(N+M); and iii) an updated OpenMP-based parallel version to accelerate the site discovery and comparison processes further. Our results show significant performance improvements for a base dataset of proteins. Results show that the optimized sequential Geomfinder achieves up to 26.3x speedup, compared to the original sequential Geomfinder, for relatively large proteins (reducing from approximately 1 hour to approximately 2 and a half minutes). And our OpenMP Geomfinder proposal, combined with the sequential optimization, reaches a speedup of more than 1000x compared to the sequential original Geomfinder. This novel optimized and parallel Geomfinder version allows for processing large protein structures that were not reasonable to compare before., This research was funded by the Spanish Government (Grants PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148- S), by Generalitat de Catalunya (2021 SGR 01007), and by Vicerrectoría de Investigación y Postgrado de la Universidad del Bío-Bío (Proyecto Iniciación 2025 IN2537025)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Feature engineering for agents: an adaptive cognitive architecture for interpretable ML monitoring

  • Bravo Rocca, Gusseppe Jesus|||0000-0001-6824-1124
  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Carrillo Larco, Rodrigo M.
  • Dholakia, Ajay
  • Ellison, David
Monitoring Machine Learning (ML) models in production environments is crucial, yet traditional approaches often yield verbose, low-interpretability outputs that hinder effective decision-making. We propose a cognitive architecture for ML monitoring that applies feature engineering principles to agents based on Large Language Models (LLMs), significantly enhancing the interpretability of monitoring outputs. Central to our approach is a Decision Procedure module that simulates feature engineering through three key steps: Refactor, Break Down, and Compile. The Refactor step improves data representation to better capture feature semantics, allowing the LLM to focus on salient aspects of the monitoring data while reducing noise and irrelevant information. Break Down decomposes complex information for detailed analysis, and Compile integrates sub-insights into clear, interpretable outputs. This process leads to a more deterministic planning approach, reducing dependence on LLM-generated planning, which can sometimes be inconsistent and overly general. The combination of feature engineering-driven planning and selective LLM utilization results in a robust decision support system, capable of providing highly interpretable and actionable insights. Experiments using multiple LLMs demonstrate the efficacy of our approach, achieving significantly higher accuracy compared to various baselines across several domains., This work was partially supported by Lenovo and Intel® as part of the Lenovo AI Innovators University Research program, by Spanish Ministry of Science (MICINN), the Research State Agency (AEI) and European Regional Development Funds (ERDF/FEDER) under contracts PID2019- 107255GB-C22 and PID2021-126248OB-I00, and by the Generalitat de Catalunya (AGAUR) under contract 2021-SGR-00478., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

MNEMOSENE++: Scalable multi-tile design with enhanced buffering and VGSOT-MRAM based compute-in-memory crossbar array

  • Escuín Blasco, Carlos
  • García Redondo, Fernando
  • Zahedi, Mahdi
  • Ibáñez Marín, Pablo Enrique
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Viñals Yúfera, Victor
  • Llaberia Griñó, José M.|||0000-0002-3753-4108
  • Myers, James
  • Ryckaert, Julien
  • Biswas, Dwaipayan
  • Catthoor, Francky
This paper optimizes the MNEMOSENE architecture, a compute-in-memory (CiM) tile design integrating computation and storage for increased efficiency. We identify and address bottlenecks in the Row Data (RD) buffer that cause losses in performance. Our proposed approach includes mitigating these buffering bottlenecks and extending MNEMOSENE’s single-tile design to a multi-tile configuration for improved parallel processing. The proposal is validated through comprehensive analyses exploring the mapping of diverse neural networks evaluated on CiM crossbar arrays based on NVM technologies. These proposed enhancements lead up to 55% reduction in execution time compared to the original single-tile architecture for any general matrix multiplication (GEMM) operation. Our evaluation shows that while ReRAM and PCM offer notable energy advantages, their integration with scaled CMOS is limited, which leads to VGSOT-MRAM emerging as a promising alternative due to its good balance between energy efficiency and superior integration capabilities. The VGSOT-MRAM crossbar arrays provide 12×,49×, and 346× more energy efficiency than PCM, ReRAM, and STT-MRAM ones, respectively. It translates, on average for the considered workload, in 1.5×,3×, and 14.5× better energy efficiency of the entire system., This project has received funding from the ECSEL Joint Undertaking (JU) under grant agreement No 876925, from MCIN/AEI/10.13039/501100011033 (grants PID2019-105660RB-C21 and PID2019-107255GB-C22), and from Aragon Government (T58_23R research group)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Transparent load balancing of MPI programs using OmpSs-2@Cluster and DLB

  • Aguilar Mena, Jimmy|||0000-0001-6802-2247
  • Ali, Omar Shaaban Ibrahim
  • López Herrero, Víctor
  • Garcia Casulla, Marta
  • Carpenter, Paul Matthew
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
Load imbalance is a long-standing source of inefficiency in high performance computing. The situation has only got worse as applications and systems increase in complexity, e.g., adaptive mesh refinement, DVFS, memory hierarchies, power and thermal management, and manufacturing processes. Load balancing is often implemented in the application, but it obscures application logic and may need extensive code refactoring. This paper presents an automated and transparent dynamic load balancing approach for MPI applications with OmpSs-2 tasks, which relieves applications from this burden. Only local and trivial changes are required to the application. Our approach exploits the ability of OmpSs-2@Cluster to offload tasks for execution on other nodes, and it reallocates compute resources among ranks using the Dynamic Load Balancing~(DLB) library. It employs LeWI to react to fine-grained load imbalances and DROM to address coarse-grained load imbalances by reserving cores on other nodes that can be reclaimed on demand. We use an expander graph to limit the amount of point-to-point communication and state. The results show 46% reduction in time-to-solution for micro-scale solid mechanics on 32 nodes and a 20% reduction beyond DLB for $n$-body on 16 nodes, when one node is running slow. A synthetic benchmark shows that performance is within 10% of optimal for an imbalance of up to 2.0 on 8 nodes. All software is released open source., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement No 955606 (DEEP-SEA) and 754337 (EuroEXA). It is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255GB and Ramon y Cajal fellowship RYC2018-025628-I) and by the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Experience replay and zero-shot clustering for continual learning in diabetic retinopathy detection

  • Bravo Rocca, Gusseppe Jesus|||0000-0001-6824-1124
  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Dholakia, Ajay
  • Ellison, David
  • Carrillo Larco, Rodrigo M.
We present an approach to mitigate catastrophic forgetting in Continual Learning (CL), focusing on domain incremental scenarios in medical imaging. Our method leverages Large Language Models (LLMs) to generate task-agnostic descriptions from multimodal inputs, enabling zero-shot clustering of tasks without supervision. This clustering underpins an enhanced Experience Replay (ER) strategy, strategically sampling data points to refresh the model’s memory while preserving privacy. By incrementally updating a multi-head classifier using only data embeddings, our approach maintains both efficiency and data confidentiality. Evaluated on a challenging diabetic retinopathy dataset, our method demonstrates significant improvements over traditional CL techniques, including Elastic Weight Consolidation (EWC), Gradient Episodic Memory (GEM), and Learning Without Forgetting (LWF). Extensive experiments across Multi-Layer Perceptron (MLP), Residual, and Attention architectures show consistent performance gains (up to 3.1% in Average Mean Class Accuracy) and reduced forgetting, with only 6% computational overhead. These results highlight our approach’s potential for privacy-preserving, efficient CL in sensitive domains like healthcare, offering a promising direction for developing adaptive AI systems that can learn continuously while respecting data privacy constraints., This work was partially supported by Lenovo and Intel as part of the Lenovo AI Innovators University Research program, by the Spanish Ministry of Science (MICINN), the Research State Agency (AEI) and European Regional Development Funds (ERDF/FEDER) under grant agreements PID2019-107255GB-C22 and PID2021-126248OBI00, MCIN/AEI/10.13039/ 501100011033/FEDER, UE, and by the Generalitat de Catalunya under contract 2021-SGR-00478.
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Exploring hierarchical MPI reduction collective algorithms targeted to multicore node clusters

  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Gil, Marisa|||0000-0001-9755-3311
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
  • Spataro, William
  • Giordano, Andrea
High-performance computing applications heavily rely on message-passing mechanisms for data sharing in cluster environments, and the MPI library stands as the default communication library for parallel applications. Significant efforts have been directed toward optimizing data distribution and buffering based on size. This optimization aims to enhance communication performance and prevent issues such as running out of memory on the target node. Furthermore, the emergence of multicore clusters with larger node sizes has stimulated the investigation of hierarchical collective algorithms that consider the placement of processes within the cluster and the memory hierarchy. This paper studies and compares the performance of the algorithm of the reduction collective from the literature, specifically several implementations that do not form part of the current MPI standard, which tackle this issue. We implement the algorithms on top of Intel MPI and OpenMPI libraries using the MPI profiling interface. Experimental results with the Intel MPI Benchmarks on a multicore cluster, Intel Platinum processor-based and OmniPath interconnection network show much room for improvement in the performance of collectives depending on the message sizes., The authors acknowledge the support of the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR01007)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Performance characterization of multi-container deployment schemes for online learning inference

  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Taherkordi, Amir
Online machine learning (ML) inference services provide users with an interactive way to request for predictions in realtime. To meet the notable computational requirements of such services, they are increasingly being deployed in the Cloud. In this context, the efficient provisioning and optimization of ML inference services in the Cloud is critical to achieve the required performance and meet the dynamic queries by end-users. Existing provisioning solutions focus on framework parameter tuning and infrastructure resources scaling, without considering deployments based on containerization technologies. The latter promises reproducibility and portability features for ML inferences services. There is limited knowledge about the impact of distinct deployment schemes at the container-level on the performance of online ML inference services, particularly on how to exploit multi-container deployments and its relation with processor and memory affinity. In light of this, in this paper we investigate experimentally the containerization of ML inference services and analyze the performance of multi-container deployments that partition the threads belonging to an online learning application into multiple containers in each node. This paper shares the findings and lessons learned from conducting realistic client patterns on an image classification model across numerous deployment configurations, especially including the impact of container granularity and its potential to exploit processor and memory affinity. Our results indicate that fine-grained multi-container deployments and affinity are useful for improving performance (both throughput and latency). In particular, our experiments on single-node and four-node clusters show up to 69% and 87% performance improvement compared to the single-container deployment, respectively., This work was partially supported by Lenovo as part of Lenovo-BSC collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2021-SGR-00478 and under grant 2020 FI-B 00257., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Fine-grained scheduling for containerized HPC workloads in Kubernetes clusters

  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
Containerization technology offers lightweight OS-level virtualization, and enables portability, reproducibility, and flexibility by packing applications with low performance overhead and low effort to maintain and scale them. Moreover, container orchestrators (e.g., Kubernetes) are widely used in the Cloud to manage large clusters running many containerized applications. However, scheduling policies that consider the performance nuances of containerized High Performance Computing (HPC) workloads have not been well-explored yet. This paper conducts fine-grained scheduling policies for containerized HPC workloads in Kubernetes clusters, focusing especially on partitioning each job into a suitable multi-container deployment according to the application profile. We implement our scheduling schemes on different layers of management (application and infrastructure), so that each component has its own focus and algorithms but still collaborates with others. Our results show that our fine-grained scheduling policies outperform baseline and baseline with CPU/memory affinity enabled policies, reducing the overall response time by 35% and 19%, respectively, and also improving the makespan by 34% and 11%, respectively. They also provide better usability and flexibility to specify HPC workloads than other comparable HPC Cloud frameworks, while providing better scheduling efficiency thanks to their multi-layered approach., This work was partially supported by Lenovo as part of Lenovo-BSC collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Software-only based diverse redundancy for ASIL-D automotive applications on embedded HPC platforms

  • Alcaide Portet, Sergi|||0000-0001-8561-7795
  • Kosmidis, Leonidas
  • Hernández Luz, Carles
  • Abella Ferrer, Jaume|||0000-0001-7951-4028
©2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., High-Performance Computing (HPC) platforms become a must in automotive systems to enable autonomous driving. However, automotive platforms must avoid Common Cause Failures (CCFs), as indicated by the ISO26262 automotive safety standard. CCFs can be avoided enforcing diverse redundancy. Unfortunately, HPC platforms fail to provide such support. This paper proposes a flexible and efficient software-based scheme to implement diverse redundancy on HPC platforms. A software implementation on a Commercial Off-The-Shelf ARM multicore proves the effectiveness of this scheme to guarantee diverse redundancy with negligible performance degradation. Our solution is the first step towards an automotive-compliant HPC platform., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB and the European Union’s Horizon 2020 Research and Innovation programme under Grant Agreement No 826647 (EPI)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

TADIL: task-agnostic domain-incremental learning through task-ID inference using transformer nearest-centroid embeddings

  • Bravo Rocca, Gusseppe Jesus|||0000-0001-6824-1124
  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Dholakia, Ajay
  • Ellison, David
Classical Machine Learning models have problems when faced with learning from data that changes over time or across domains due to factors such as noise, occlusion, illumination, or frequency variation, which humans can adapt to without being given independent and identically distributed data. Therefore, a Continual Learning (CL) approach is essential, particularly in this case, the Domain-Incremental Learning. This paper presents a novel pipeline for identifying tasks in domain-incremental learning scenarios without supervision. The pipeline consists of four steps. First, we obtain base embeddings from the raw data through a transformer-based existing model. Second, we group the embedding densities based on similarity and extract few nearest points to each cluster centroid. Third, we train an incremental task classifier using only the previous few points. Finally, we make the pipeline lightweight in terms of computational requirements and build an algorithm that decides when to learn a new task in an online way using the task classifier and drift detector. We experiment with the real-world driving dataset SODA10M and several CL strategies. We conclude that the performance of CL strategies with our pipeline is better not only when the task boundaries are given, but also in the more general practical case of task-agnostic strategies that demand identifying new tasks on-the-fly., We thank Lenovo for providing the technical infrastructure to run the experiments in this paper. This work was partially supported by Lenovo and Intel as part of the Lenovo AI Innovators University Research program, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2021-SGR-00478, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Towards data-flow parallelization for adaptive mesh refinement applications

  • Sala Penadés, Kevin|||0000-0001-8233-1185
  • Rico Carro, Alejandro
  • Beltran Querol, Vicenç
Adaptive Mesh Refinement (AMR) is a prevalent method used by distributed-memory simulation applications to adapt the accuracy of their solutions depending on the turbulent conditions in each of their domain regions. These applications are usually dynamic since their domain areas are refined or coarsened in various refinement stages during their execution. Thus, they periodically redistribute their workloads among processes to avoid load imbalance. Although the defacto standard for scientific computing in distributed environments is MPI, in recent years, pure MPI applications are being ported to hybrid ones, attempting to cope with modern multi-core systems. Recently, the Task-Aware MPI library was proposed to efficiently integrate MPI communications and tasking models, providing also the transparent management of communications issued by tasks. In this paper, we demonstrate the benefits of porting AMR applications to data-flow programming models leveraging that novel hybrid approach. We exploit most of the application parallelism by taskifying all stages, allowing their natural overlap. We employ these techniques on the miniAMR proxy application, which mimics the refinement, load balancing, communication, and computation patterns of general AMR applications. We evaluate how this approach reduces the time in its computation and communication phases while achieving better programmability than other conventional hybrid techniques., This work has been supported by the European Union H2020 Programme through the DEEP-EST project (agreement No. 754304), the Spanish Government through the Severo Ochoa Program (SEV-2015-0493), the Spanish Ministry of Science and Innovation (PID2019-107255GB), and the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Towards on-demand I/O forwarding in HPC platforms

  • Bez, Jean Luca
  • Zanon Boito, Francieli
  • Miranda Bueno, Alberto
  • Nou Castell, Ramon|||0000-0003-0464-1473
  • Cortés, Toni|||0000-0002-2537-8937
  • Navaux, Philippe O.A.
I/O forwarding is an established and widely-adopted technique in HPC to reduce contention and improve I/O performance in the access to shared storage infrastructure. On such machines, this layer is often physically deployed on dedicated nodes, and their connection to the clients is static. Furthermore, the increasingly heterogeneous workloads entering HPC installations stress the I/O stack, requiring tuning and reconfiguration based on the applications' characteristics. Nonetheless, it is not always feasible in a production system to explore the potential benefits of this layer under different configurations without impacting clients. In this paper, we investigate the effects of I/O forwarding on performance by considering the application's I/O access patterns and system characteristics. We aim to explore when forwarding is the best choice for an application, how many I/O nodes it would benefit from, and whether not using forwarding at all might be the correct decision. To gather performance metrics, explore, and understand the impact of forwarding I/O requests of different access patterns, we implemented FORGE, a lightweight I/O forwarding layer in user-space. Using FORGE, we evaluated the optimal forwarding configurations for several access patterns on MareNostrum 4 (Spain) and Santos Dumont (Brazil) supercomputers. Our results demonstrate that shifting the focus from a static system-wide deployment to an on-demand reconfigurable I/O forwarding layer dictated by application demands can improve I/O performance on future machines., This study was financed by the Coordenaçao de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001. It has also received support from the Conselho Nacional de Desenvolvimento Científico e Tecnologico (CNPq), Brazil; It is also partially supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under grants PID2019-107255GB; and the Generalitat de Catalunya under contract 2014–SGR–1051., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Human-in-the-loop online multi-agent approach to increase trustworthiness in ML models through trust scores and data augmentation

  • Bravo Rocca, Gusseppe Jesus|||0000-0001-6824-1124
  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Increasing a ML model accuracy is not enough, we must also increase its trustworthiness. This is an important step for building resilient AI systems for safety-critical applications such as automotive, finance, and healthcare. For that purpose, we propose a multi-agent system that combines both machine and human agents. In this system, a checker agent calculates a trust score of each instance (which penalizes overconfidence in predictions) using an agreement-based method and ranks it; then an improver agent filters the anomalous instances based on a human rule-based procedure (which is considered safe), gets the human labels, applies geometric data augmentation, and retrains with the augmented data using transfer learning. We evaluate the system on corrupted versions of the MNIST and FashionMNIST datasets. We get an improvement in accuracy and trust score with just few additional labels compared to a baseline approach., This work was supported by Lenovo as part of LenovoBSC 2020 collaboration agreement, by the Spanish Government under contracts PID2019-107255GB-C21 and PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Distributing deep learning hyperparameter tuning for 3D medical image segmentation

  • Berral García, Josep Lluís|||0000-0003-3037-3580
  • Aranda Llorens, Oriol
  • Domínguez Bermúdez, Juan Luis
  • Torres Viñals, Jordi|||0000-0003-1963-7418
Most research on novel techniques for 3D Medical Image Segmentation (MIS) is currently done using Deep Learning with GPU accelerators. The principal challenge of such technique is that a single input can easily cope computing resources, and require prohibitive amounts of time to be processed. Distribution of deep learning and scalability over computing devices is an actual need for progressing on such research field. Conventional distribution of neural networks consist in “data parallelism”, where data is scattered over resources (e.g., GPUs) to parallelize the training of the model. However, “experiment parallelism” is also an option, where different training processes (i.e., on a hyper-parameter search) are parallelized across resources. While the first option is much more common on 3D image segmentation, the second provides a pipeline design with less dependence among parallelized processes, allowing overhead reduction and more potential scalability. In this work we present a design for distributed deep learning training pipelines, focusing on multi-node and multi-GPU environments, where the two different distribution approaches are deployed and benchmarked. We take as proof of concept the 3D U-Net architecture, using the MSD Brain Tumor Segmentation dataset, a state-of-art problem in medical image segmentation with high computing and space requirements. Using the BSC MareNostrum supercomputer as benchmarking environment, we use TensorFlow and Ray as neural network training and experiment distribution platforms. We evaluate the experiment speed-up when parallelizing, showing the potential for scaling out on GPUs and nodes. Also comparing the different parallelism techniques, showing how experiment distribution leverages better such resources through scaling, e.g. by a speed-up factor from x12 to x14 using 32 GPUs. Finally, we provide the implementation of the design open to the community, and the non-trivial steps and methodology for adapting and deploying a MIS case as the here presented., This work has been partially financed by the European Commission (EU-H2020 INCISIVE GA.952179, and CALLISTO GA.101004152). Also the Spanish Ministry of Science (PID2019- 107255GB-C22/ AEI / 10.13039/501100011033), and Generalitat de Catalunya through the 2017-SGR-1414 project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Improving performance of HPC kernels on FPGAs using high-level resource management

  • Filgueras Izquierdo, Antonio|||0000-0001-6071-3254
  • Vidal, Miquel|||0000-0002-1973-8289
  • Jiménez González, Daniel|||0000-0001-6064-7883
  • Álvarez Martínez, Carlos|||0000-0003-0536-5183
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
In state-of-the-art FPGA, especially in chiplet-based devices, place and route has become an important challenge due to an increase in device size and complexity. In the same way, off-chip memory resources have grown in size and number of memory modules. Making efficient use of them has become a difficult task., This work is supported by the TEXTAROSSA project G.A. n.956831, as part of the EuroHPC initiative, by the Spanish Government (Grants PCI2021-121964 - TEXTAROSSA, PID2019-107255GB-C21, PID2019-107255GB-C22, MCIN/AEI/10.13039/501100011033, and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Scanflow: an end-to-end agent-based autonomic ML workflow manager for clusters

  • Liu, Peini|||0000-0003-0058-8732
  • Bravo Rocca, Gusseppe Jesus|||0000-0001-6824-1124
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Machine Learning (ML) is more than just training models, the whole life-cycle must be considered. Once deployed, a ML model needs to be constantly managed, supervised and debugged to guarantee its availability, validity and robustness in dynamic contexts. This demonstration presents an agent-based ML workflow manager so-called Scanflow1, which enables autonomic management and supervision of the end-to-end life-cycle of ML workflows on distributed clusters. The case study on a MNIST project2 shows that different teams can collaborate using Scanflow within a ML project at different phases, and the effectiveness of agents to maintain the model accuracy and throughput of the model serving while running in production., This work was partially supported by Lenovo as part of LenovoBSC 2020 collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257.
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Analyzing the performance of hierarchical collective algorithms on ARM-based multicore clusters

  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Gil, Marisa|||0000-0001-9755-3311
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
MPI is the de facto communication standard library for parallel applications in distributed memory architectures. Collective operations performance is critical in HPC applications as they can become the bottleneck of their executions. The advent of larger node sizes on multicore clusters has motivated the exploration of hierarchical collective algorithms aware of the process placement in the cluster and the memory hierarchy. This work analyses and compares several hierarchical collective algorithms from the literature that do not form part of the current MPI standard. We implement the algorithms on top of OpenMPI using the shared-memory facility provided by MPI-3 at the intra-node level and evaluate them on ARM-based multicore clusters. From our results, we evidence aspects of the algorithms that impact the performance and applicability of the different algorithms. Finally, we propose a model that helps us to analyze the scalability of the algorithms., This work has been supported by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Compression-aware and performance-efficient insertion policies for long-lasting hybrid LLCs

  • Escuín Blasco, Carlos
  • Ali Khan, Asif
  • Ibáñez, Pablo Enrique
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Castrillón, Jerónimo
  • Viñals Yúfera, Victor
Emerging non-volatile memory (NVM) technologies can potentially replace large SRAM memories such as the last-level cache (LLC). However, despite recent advances, NVMs suffer from higher write latency and limited write endurance. Recently, NVM-SRAM hybrid LLCs are proposed to combine the best of both worlds. Several policies have been proposed to improve the performance and lifetime of hybrid LLCs by intelligently steering the incoming LLC blocks into either the SRAM or NVM part, regarding the cache behavior of the LLC blocks and the SRAM/NVM device properties. However, these policies neither consider compressing the contents of the cache block nor using partially worn-out NVM cache blocks.This paper proposes new insertion policies for byte-level fault-tolerant hybrid LLCs that collaboratively optimize for lifetime and performance. Specifically, we leverage data compression to utilize partially defective NVM cache entries, thereby improving the LLC hit rate. The key to our approach is to guide the insertion policy by both the reuse properties of the block and the size resulting from its compression. A block is inserted in NVM only if it is a read-reuse block or its compressed size is lower than a threshold. It will be inserted in SRAM if the block is a write-reuse or its compressed size is greater than the threshold. We use set-dueling to tune the compression threshold at runtime. This compression threshold provides a knob to control the NVM write rate and, together with a rule-based mechanism, allows balancing performance and lifetime.Overall, our evaluation shows that, with affordable hardware overheads, the proposed schemes can nearly reach the performance of an SRAM cache with the same associativity while improving lifetime by 17× compared to a hybrid NVM-unaware LLC. Our proposed scheme outperforms the state-of-the-art insertion policies by 9% while achieving a comparative lifetime. The rule-based mechanism shows that by compromising, for instance, 1.1% and 1.9% performance, the NVM lifetime can be further increased by 28% and 44%, respectively., This work was partially funded by the HiPEAC collaboration grant 2020, the Center for Advancing Electronics Dresden (cfaed), the German Research Council (DFG) through the HetCIM project (502388442) under the Priority Program on ‘Disruptive Memory Technologies’ (SPP 2377), and from grants (1) PID2019-105660RB-C21 and PID2019-107255GB- C22/AEI/10.13039/501100011033 from Agencia Estatal de Investigación (AEI), and (2) gaZ: T5820R research group from Dept. of Science, University and Knowledge Society, Government of Aragon., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Cellular automata on a multi-GPU architecture: A technical overview

  • Giordano, Andrea
  • De Rango, Alessio
  • D'Ambrosio, Donato
  • Gil, Marisa|||0000-0001-9755-3311
  • Macri, Davide
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
  • Rongo, Rocco
  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Mendicino, Giuseppe
  • Spataro, William
This work is focused on the transparent execution of Cellular Automata models on a multi-GPU architecture. Although Cellular Automata models can be easily parallelized on a single GPU, the domain size and transition function complexity may require the use of multiple GPUs. Our goal is to allow modellers to be completely unaware of the parallel execution context, i.e., the code implementing the Cellular Automata model remains the same regardless if the execution is performed on CPU, single GPU, or multi-GPU systems. This paper supplies meaningful technical insights on how to ensure both transparency and efficiency in multi-GPU execution of Cellular Automata models. In particular, an object-oriented approach is exploited in which a transparent layer is devised that abstracts the parallelization details and allows a strong “separation of concerns” between the execution parallelism issues and the model implementation. Preliminary experiments have been carried out on the multi-GPU cluster CTE-POWER available at the Barcelona Supercomputing Center (BSC), witnessing good speedups notwithstanding the transparency feature supplied by our approach., This research was partially funded by the Italian “ICSC National Center for HPC, Big Data and Quantum Computing” Project, CN00000013 (approved under the Call M42C – Investment 1.4 – Avviso “Centri Nazionali” - D.D. n. 3138 of 16.12.2021, admitted to financing with MUR Decree n. 1031 of 06.17.2022). This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ’Innovation Ecosystems’, building ’Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work was supported by Italian Ministry of Research (MUR) under PNRR projects FAIR “Future AI ResearcH” - Spoke 9 - CUP H23C22000860006. This research was partially funded by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR-01007)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Exploring image transformations with diffusion models: A survey of applications and implementation code

  • Arellano García, Silvia
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Tous Liesa, Rubén|||0000-0002-1409-5843
Diffusion Models have become increasingly popular in recent years and their applications span a wide range of fields. This survey focuses on the use of diffusion models in computer vision, specially in the branch of image transformations. The objective of this survey is to provide an overview of state-of-the-art applications of diffusion models in image transformations, including image inpainting, super-resolution, restoration, translation, and editing. This survey presents a selection of notable papers and repositories including practical applications of diffusion models for image transformations. The applications are presented in a practical and concise manner, facilitating the understanding of concepts behind diffusion models and how they function. Additionally, it includes a curated collection of GitHub repositories featuring popular examples of these subjects., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB and PID2021-124463OB- IOO, by the Generalitat de Catalunya under grants 2021-SGR-00478 and 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

DO-178C certification of general-purpose GPU software: review of existing methods and future directions

  • Trompouki, Matina Maria
  • Kosmidis, Leonidas
—General-Purpose GPU software is considered for use in avionics to satisfy the increased computational requirements of future systems. Therefore, it needs to be certified following the DO-178C guidance as all airborne software. In this work, we review the existing methods in the literature, we analyse their advantages and disadvantages, and we discuss how they can be combined to obtain certification with lower effort and cost. Our focus is restricted on application-level software, under the premise that successful completion of verification of avionics graphics GPU software products has been demonstrated, so their GPU compiler has been considered acceptable for these already DO-178C certified products, or existing qualified GPU compilers exist. Finally, we discuss upcoming solutions for certified general purpose GPU computing ., This work was performed within the Airbus TANIAGPU Project ADS (E/200) in collaboration with the project partners Airbus Defence and Space, Madrid, Spain and CoreAVI, Canada. It was also partially supported by the European Space Agency (ESA) through the GPU4S (GPU for Space) activity, the Spanish Ministry of Economy and Competitiveness under grants PID2019-107255GB and FJCI-2017-34095 (Spanish State Research Agency / http://dx.doi.org/10.13039/501100011033) and the HiPEAC Network of Excellence., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

An educational tool to analyze the hardware/software integration in RISC-V systems

  • Castells Rufas, David
  • Novo, David
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
The analysis and verification of the integration of hardware and software components is challenging. Traditional HDL simulators, which are often slow and provide low-level analysis tools, are not always effective for simulating the software component of processor-based systems. This work presents an interactive co-simulation framework that enhances students’ understanding of computer architecture concepts typically obscured by hardware intricacies. Our framework facilitates a unified environment for iterative design, enabling detailed analysis and generation of reports and traces to identify system performance and bottlenecks. By prioritizing analysis features over performance, our framework serves as a valuable tool for educational purposes and comprehensive examination of system interactions., This project is partly funded by the Catalan Government Research groups 2021-SGR-01623, 2021-SGR-01007, by the Spanish Ministry of Science, Innovation and Universities under grants PID2023-148717OB-C22, PID2019-107255BGC22, PCI2021-121964, and the F3CAS project (ANR-20- CE25-0010) of the French National Research Agency., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

MUCH: exploiting pairwise hardware event monitor correlations for improved timing analysis of complex MPSoCs

  • Vilardell Moreno, Sergi
  • Serra Mochales, Isabel
  • Mezzetti, Enrico|||0000-0002-1886-2931
  • Abella Ferrer, Jaume|||0000-0001-7951-4028
  • Cazorla Almeida, Francisco Javier
Measurement-based timing analysis techniques increasingly rely on the Performance Monitoring Units (PMU) of MPSoCs, as these units implement specialized Hardware Event Monitors (HEMs) that convey detailed information about multicore interference in hardware shared resources. Unfortunately, there is an evident mismatch between the large number of HEMs (typically several hundreds) and the comparatively small number (normally less than ten) of Performance Monitoring Counters (PMCs) that can be configured to track HEMs in the PMU. Timing analysis normally require to observe a non-negligible number of HEMs per task from the same execution. However, due to the small number of PMCs, HEMs are necessarily collected across multiple runs that, despite intended to repeat the same experiment, carry out some significant variability (above 50% for some HEMs in relevant MPSoCs) caused by platform-intrinsic execution conditions. Therefore, blindly merging HEMs from different runs is not acceptable since they may easily correspond to significantly different conditions. To tackle this issue, the HRM approach has been proposed recently to merge HEMs from different runs accurately preserving their correlation w.r.t. one anchor HEM (i.e. processor cycles) building on order statistics. However, HRM do not always preserves the correlation between other pairs of HEMs that might be lost to a large extent. This paper copes with HRM limitations by proposing the MUlti-Correlation HEM reading and merging approach (MUCH). MUCH builds on multivariate Gaussian distributions to merge HEMs from different runs while preserving pairwise correlations across each individual pair of HEMs simultaneously. Our results on an NXP T2080 MPSoC used for avionics systems show that MUCH largely outperforms HRM for an identical number of input runs., This work has been partially supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB, the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 772773) and the HiPEAC Network of Excellence., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Deploying vision foundation AI models on the edge. The SAM2 experience

  • Lin, Zheshuo
  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
As AI-driven applications expand across industries, the need for efficient edge computing solutions becomes increasingly critical. Traditional AI models are designed for high-performance cloud infrastructures, but emerging constraints—such as privacy requirements, network limitations, and real-time processing needs—necessitate optimized deployment on resource-constrained edge devices. This study presents a practical experience in adapting Segment Anything Model 2 (SAM2), a vision foundation model, for edge AI environments. The adaptation process involved translating the model to C++ using ONNX Runtime, enabling efficient execution on heterogeneous hardware. Experimental evaluations demonstrate that deploying SAM2 at the edge enhances processing efficiency, reduces reliance on network stability, and improves real-time responsiveness. This research provides valuable insights into AI in pervasive computing environments, contributing to the sustainable and scalable deployment of foundation models on edge devices., This work is partially funded by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, as well as by the SGR programs 2021-SGR-00478 and 2021-SGR-00326 of the Catalan government. Finally, the research that led to these results has received funding from the Collaboration Grant for the 2024/2025 period from the Spanish Ministry of Education., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Scanflow-K8s: agent-based framework for autonomic management and supervision of ML workflows in Kubernetes clusters

  • Liu, Peini|||0000-0003-0058-8732
  • Bravo Rocca, Gusseppe
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Machine Learning (ML) projects are currently heavily based on workflows composed of some reproducible steps and executed as containerized pipelines to build or deploy ML models efficiently because of the flexibility, portability, and fast delivery they provide to the ML life-cycle. However, deployed models need to be watched and constantly managed, supervised, and debugged to guarantee their availability, validity, and robustness in unexpected situations. Therefore, containerized ML workflows would benefit from leveraging flexible and diverse autonomic capabilities. This work presents an architecture for autonomic ML workflows with abilities for multi-layered control, based on an agent-based approach that enables autonomic management and supervision of ML workflows at the application layer and the infrastructure layer (by collaborating with the orchestrator). We redesign the Scanflow ML framework to support such multi-agent approach by using triggers, primitives, and strategies. We also implement a practical platform, so-called Scanflow-K8s, that enables autonomic ML workflows on Kubernetes clusters based on the Scanflow agents. MNIST image classification and MLPerf ImageNet classification benchmarks are used as case studies to show the capabilities of Scanflow-K8s under different scenarios. The experimental results demonstrate the feasibility and effectiveness of our proposed agent approach and the Scanflow-K8s platform for the autonomic management of ML workflows in Kubernetes clusters at multiple layers., This work was supported by Lenovo as part of Lenovo-BSC 2020 collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Page size aware cache prefetching

  • Vavouliotis, Georgios
  • Chacon, Gino
  • Álvarez Martí, Lluc|||0000-0003-0506-8867
  • Gratz, Paul V.
  • Jiménez, Daniel A.
  • Casas, Marc|||0000-0003-4564-2093
The increase in working set sizes of contemporary applications outpaces the growth in cache sizes, resulting in frequent main memory accesses that deteriorate system per- formance due to the disparity between processor and memory speeds. Prefetching data blocks into the cache hierarchy ahead of demand accesses has proven successful at attenuating this bottleneck. However, spatial cache prefetchers operating in the physical address space leave significant performance on the table by limiting their pattern detection within 4KB physical page boundaries when modern systems use page sizes larger than 4KB to mitigate the address translation overheads. This paper exploits the high usage of large pages in modern systems to increase the effectiveness of spatial cache prefetch- ing. We design and propose the Page-size Propagation Module (PPM), a µarchitectural scheme that propagates the page size information to the lower-level cache prefetchers, enabling safe prefetching beyond 4KB physical page boundaries when the accessed blocks reside in large pages, at the cost of augmenting the first-level caches’ Miss Status Holding Register (MSHR) entries with one additional bit. PPM is compatible with any cache prefetcher without implying design modifications. We capitalize on PPM’s benefits by designing a module that consists of two page size aware prefetchers that inherently use different page sizes to drive prefetching. The composite module uses adaptive logic to dynamically enable the most appropriate page size aware prefetcher. Finally, we show that the proposed designs are transparent to which cache prefetcher is used. We apply the proposed page size exploitation techniques to four state-of-the-art spatial cache prefetchers. Our evalua- tion shows that our proposals improve single-core geomean performance by up to 8.1% (2.1% at minimum) over the original implementation of the considered prefetchers, across 80 memory-intensive workloads. In multi-core contexts, we report geomean speedups up to 7.7% across different cache prefetchers and core configurations., This work is supported by the Spanish Ministry of Science and Technology through the PID2019-107255GB project, the Generalitat de Catalunya (contract 2017-SGR-1414), the European Union Horizon 2020 research and innovation program under grant agreement No 955606 (DEEP-SEA EU project), the National Science Foundation through grants CNS-1938064 and CCF-1912617, and the Semiconductor Research Corporation project GRC 2936.001. Georgios Vavouliotis has been supported by the Spanish Ministry of Economy, Industry, and Competitiveness and the European Social Fund under the FPI fellowship No. PRE2018-087046. Marc Casas has been partially supported by the Grant RYC2017-23269 funded by MCIN/AEI/10.13039/501100011033 and ESF ‘Investing in your future’., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Epicentral region estimation using convolutional neural networks

  • Cruz de la Cruz, Stalin Leonel
  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Alvarado Bermúdez, Leonardo
  • Mus León, Sergi
  • Rojas Ulacio, Otilio Jose
Recent works have assessed the capability of deep neural networks of estimating the epicentral source region of a seismic event from a single-station three-channel signal. In all the cases, the geographical partitioning is performed by automatic tessellation algorithms such as the Voronoi decomposition. This paper evaluates the hypothesis that the source region estimation accuracy is significantly increased if the geographical partitioning is performed considering the regional geological characteristics such as the tectonic plate boundaries. Also, it raises the transformation of the training data to increase the accuracy of the predictive model based on a Projected Coordinate Reference (PCR) System. A deep convolutional neural network (CNN) is applied over the data recorded by the broadband stations of the Venezuelan Foundation of Seismological Research (FUNVISIS) in the region of 9.5 to 11.5ºN and 67.0 to 69.0ºW between April 2018 and April 2019. In order to estimate the epicentral source region of a detected event, several geographical tessellations provided by seismologists from the area are employed. These tessellations, with different number of partitions, consider the fault systems of the study region (San Sebastián, La Victoria and Morón fault systems). The results are compared to the ones obtained with automatic partitioning performed by the k-means algorithm., This work is partially supported by the Spanish Ministry of Economy and Competitivity under contract TIN2015-65316-P, by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB-C22, and by the SGR programmes (2014-SGR-1051 and 2017-SGR-962) of the Catalan Government and has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 777778 (MATHROCKS)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

High-resolution image generation using artificial intelligence and diffusion modelling

  • Scoles García, Amanda
  • Sionis Ponte, Giulia
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
Super-Resolution via Repeated Refinement (SR3) is a state-of-the-art super-resolution algorithm based on diffusion model that can enhance the resolution of images. This method is used to pre-trained models on large datasets and can be used for various tasks without requiring training from scratch. Training SR3 from scratch using the ImageNet dataset involves a complex process that requires substantial computational resources and expertise. The idea is applied the trained SR3 model to new images by feeding the low-resolution inputs and obtaining the high-resolution outputs. It's important to note that training SR3 from scratch is a resource-intensive process that requires powerful GPUs and significant computation time. If you do not have access to such resources, an alternative is to use pre-trained models that are already available and fine-tune them on specific datasets or tasks. The paper shows the result of comparing the resolution of the preprocessed images using a significantly smaller number of images to perform the training with those obtained using the pre-trained model. The results obtained show acceptable results without having to perform on large datasets minimizing the computation time to obtain the resolution of images., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-100 and PID2019-107255GB-C22 by the Generalitat de Catalunya under grant 2021-SGR-00326 and 2021-SGR-01007. Finally, the research leading to these results also has received funding from the European Union's Horizon 2020 research and innovation program under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

OmpSs-2@Cluster: Distributed memory execution of nested OpenMP-style tasks

  • Aguilar Mena, Jimmy|||0000-0001-6802-2247
  • Ali, Omar Shaaban Ibrahim
  • Beltran Querol, Vicenç
  • Carpenter, Paul Matthew
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
State-of-the-art programming approaches generally have a strict division between intra-node shared memory parallelism and inter-node MPI communication. Tasking with dependencies offers a clean, dependable abstraction for a wide range of hardware and situations within a node, but research on task offloading between nodes is still relatively immature. This paper presents a flexible task offloading extension of the OmpSs-2 programming model, which inherits task ordering from a sequential version of the code and uses a common address space to avoid address translation and simplify the use of data structures with pointers. It uses weak dependencies to enable work to be created concurrently. The program is executed in distributed dataflow fashion, and the runtime system overlaps the construction of the distributed dependency graph, enforces dependencies, transfers data, and schedules tasks for execution. Asynchronous task parallelism avoids synchronization that is often required in MPI+OpenMP tasks. Task scheduling is flexible, and data location is tracked through the dependencies. We wish to enable future work in resiliency, scalability, load balancing and malleability, and therefore release all source code and examples open source., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement No 955606 (DEEP-SEA) and 754337 (EuroEXA). It is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255GB and Ramon y Cajal fellowship RYC2018-025628-I) and by the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Comparison of GPU computing methodologies for safety-critical systems: an avionics case study

  • Benito Bermúdez, Marc
  • Trompouki, Matina Maria
  • Kosmidis, Leonidas
  • García Martín, David
  • Carretero Jiménez, Sergio
  • Wenger, Ken
Introducing advanced functionalities in safety-critical systems requires using more powerful architectures such as GPUs. However software in safety-critical industries is subject to functional certification, which cannot be achieved using standard GPU programming languages such as CUDA and OpenCL. Fortunately, GPUs are already used in certified critical systems for display tasks, using safety-certified solutions such as OpenGL SC 2.0. In this paper, we compare two state-of-the art graphics-based methodologies, OpenGL SC 2.0 and Brook Auto/BRASIL for the implementation of a prototype avionics case study. We evaluate both methods on a realistic industrial setup, composed by an avionics-grade GPU and a safety-certified GPU driver in terms of development metrics and performance, showing their feasibility., This work was funded by the Airbus TANIA-GPU Project ADS (E/200). It was also partially supported by the Spanish Ministry of Economy and Competitiveness under grants PID2019-107255GB and FJCI-2017-34095 and HiPEAC., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

OmpSs-2@Cluster: Distributed memory execution of nested OpenMP-style tasks

  • Aguilar Mena, Jimmy
  • Ali, Omar Shaaban Ibrahim
  • Beltran Querol, Vicenç
  • Carpenter, Paul Matthew
  • Ayguadé Parra, Eduard
  • Labarta Mancho, Jesús José
State-of-the-art programming approaches generally have a strict division between intra-node shared memory parallelism and inter-node MPI communication. Tasking with dependencies offers a clean, dependable abstraction for a wide range of hardware and situations within a node, but research on task offloading between nodes is still relatively immature. This paper presents a flexible task offloading extension of the OmpSs-2 programming model, which inherits task ordering from a sequential version of the code and uses a common address space to avoid address translation and simplify the use of data structures with pointers. It uses weak dependencies to enable work to be created concurrently. The program is executed in distributed dataflow fashion, and the runtime system overlaps the construction of the distributed dependency graph, enforces dependencies, transfers data, and schedules tasks for execution. Asynchronous task parallelism avoids synchronization that is often required in MPI+OpenMP tasks. Task scheduling is flexible, and data location is tracked through the dependencies. We wish to enable future work in resiliency, scalability, load balancing and malleability, and therefore release all source code and examples open source., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement No 955606 (DEEP-SEA) and 754337 (EuroEXA). It is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255GB and Ramon y Cajal fellowship RYC2018-025628-I) and by the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

An educational tool to analyze the hardware/software integration in RISC-V systems

  • Castells Rufas, David
  • Novo, David
  • Martorell Bofill, Xavier
The analysis and verification of the integration of hardware and software components is challenging. Traditional HDL simulators, which are often slow and provide low-level analysis tools, are not always effective for simulating the software component of processor-based systems. This work presents an interactive co-simulation framework that enhances students’ understanding of computer architecture concepts typically obscured by hardware intricacies. Our framework facilitates a unified environment for iterative design, enabling detailed analysis and generation of reports and traces to identify system performance and bottlenecks. By prioritizing analysis features over performance, our framework serves as a valuable tool for educational purposes and comprehensive examination of system interactions., This project is partly funded by the Catalan Government Research groups 2021-SGR-01623, 2021-SGR-01007, by the Spanish Ministry of Science, Innovation and Universities under grants PID2023-148717OB-C22, PID2019-107255BGC22, PCI2021-121964, and the F3CAS project (ANR-20- CE25-0010) of the French National Research Agency., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Compression-aware and performance-efficient insertion policies for long-lasting hybrid LLCs

  • Escuín Blasco, Carlos
  • Ali Khan, Asif
  • Ibáñez, Pablo Enrique
  • Monreal Arnal, Teresa
  • Castrillón, Jerónimo
  • Viñals Yúfera, Victor
Emerging non-volatile memory (NVM) technologies can potentially replace large SRAM memories such as the last-level cache (LLC). However, despite recent advances, NVMs suffer from higher write latency and limited write endurance. Recently, NVM-SRAM hybrid LLCs are proposed to combine the best of both worlds. Several policies have been proposed to improve the performance and lifetime of hybrid LLCs by intelligently steering the incoming LLC blocks into either the SRAM or NVM part, regarding the cache behavior of the LLC blocks and the SRAM/NVM device properties. However, these policies neither consider compressing the contents of the cache block nor using partially worn-out NVM cache blocks.This paper proposes new insertion policies for byte-level fault-tolerant hybrid LLCs that collaboratively optimize for lifetime and performance. Specifically, we leverage data compression to utilize partially defective NVM cache entries, thereby improving the LLC hit rate. The key to our approach is to guide the insertion policy by both the reuse properties of the block and the size resulting from its compression. A block is inserted in NVM only if it is a read-reuse block or its compressed size is lower than a threshold. It will be inserted in SRAM if the block is a write-reuse or its compressed size is greater than the threshold. We use set-dueling to tune the compression threshold at runtime. This compression threshold provides a knob to control the NVM write rate and, together with a rule-based mechanism, allows balancing performance and lifetime.Overall, our evaluation shows that, with affordable hardware overheads, the proposed schemes can nearly reach the performance of an SRAM cache with the same associativity while improving lifetime by 17× compared to a hybrid NVM-unaware LLC. Our proposed scheme outperforms the state-of-the-art insertion policies by 9% while achieving a comparative lifetime. The rule-based mechanism shows that by compromising, for instance, 1.1% and 1.9% performance, the NVM lifetime can be further increased by 28% and 44%, respectively., This work was partially funded by the HiPEAC collaboration grant 2020, the Center for Advancing Electronics Dresden (cfaed), the German Research Council (DFG) through the HetCIM project (502388442) under the Priority Program on ‘Disruptive Memory Technologies’ (SPP 2377), and from grants (1) PID2019-105660RB-C21 and PID2019-107255GB- C22/AEI/10.13039/501100011033 from Agencia Estatal de Investigación (AEI), and (2) gaZ: T5820R research group from Dept. of Science, University and Knowledge Society, Government of Aragon., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Software-only based diverse redundancy for ASIL-D automotive applications on embedded HPC platforms

  • Alcaide Portet, Sergi
  • Kosmidis, Leonidas
  • Hernández Luz, Carles
  • Abella Ferrer, Jaume
©2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., High-Performance Computing (HPC) platforms become a must in automotive systems to enable autonomous driving. However, automotive platforms must avoid Common Cause Failures (CCFs), as indicated by the ISO26262 automotive safety standard. CCFs can be avoided enforcing diverse redundancy. Unfortunately, HPC platforms fail to provide such support. This paper proposes a flexible and efficient software-based scheme to implement diverse redundancy on HPC platforms. A software implementation on a Commercial Off-The-Shelf ARM multicore proves the effectiveness of this scheme to guarantee diverse redundancy with negligible performance degradation. Our solution is the first step towards an automotive-compliant HPC platform., This work has been supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB and the European Union’s Horizon 2020 Research and Innovation programme under Grant Agreement No 826647 (EPI)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

An FPGA accelerator of the wavefront algorithm for genomics pairwise alignment

  • Haghi, Abbas
  • Marco Sola, Santiago
  • Álvarez Martí, Lluc
  • Diamantopoulos, Dionysios
  • Hagleitner, Christoph
  • Moretó Planas, Miquel
In the last years, advances in next-generation sequencing technologies have enabled the proliferation of genomic applications that guide personalized medicine. These applications have an enormous computational cost due to the large amount of genomic data they process. The first step in many of these applications consists in aligning reads against a reference genome. Very recently, the wavefront alignment algorithm has been introduced, significantly reducing the execution time of the read alignment process. This paper presents the first FPGA- based hardware/software co-designed accelerator of such relevant algorithm. Compared to the reference WFA CPU-only implementation, the proposed FPGA accelerator achieves performance speedups of up to 13.5× while consuming up to 14.6× less energy.ed medicine. These applications have an enormous computational cost due to the large amount of genomic data they process. The first step in many of these applications consists in aligning reads against a reference genome. Very recently, the wavefront alignment algorithm has been introduced, significantly reducing the execution time of the read alignment process. This paper presents the first FPGA- based hardware/software co-designed accelerator of such relevant algorithm. Compared to the reference WFA CPU-only imple- mentation, the proposed FPGA accelerator achieves performance speedups of up to 13.5× while consuming up to 14.6× less energy., This work has been supported by the European HiPEAC Network of Excellence, by the Spanish Ministry of Science and Innovation (contract PID2019-107255GB-C21/AEI/10.13039/501100011033), by the Generalitat de Catalunya (contracts 2017-SGR-1414 and 2017-SGR-1328), by the IBM/BSC Deep Learning Center initiative, and by the DRAC project, which is co-financed by the European Union Regional Development Fund within the framework of the ERDF Operational Program of Catalonia 2014-2020 with a grant of 50% of total eligible cost. Ll. Alvarez has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under the Juan de la Cierva Formacion fellowship No. FJCI-2016-30984. M. Moreto has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramon y Cajal fellowship No. RYC-2016-21104., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Page size aware cache prefetching

  • Vavouliotis, Georgios
  • Chacon, Gino
  • Álvarez Martí, Lluc
  • Gratz, Paul V.
  • Jiménez, Daniel A.
  • Casas, Marc
The increase in working set sizes of contemporary applications outpaces the growth in cache sizes, resulting in frequent main memory accesses that deteriorate system per- formance due to the disparity between processor and memory speeds. Prefetching data blocks into the cache hierarchy ahead of demand accesses has proven successful at attenuating this bottleneck. However, spatial cache prefetchers operating in the physical address space leave significant performance on the table by limiting their pattern detection within 4KB physical page boundaries when modern systems use page sizes larger than 4KB to mitigate the address translation overheads. This paper exploits the high usage of large pages in modern systems to increase the effectiveness of spatial cache prefetch- ing. We design and propose the Page-size Propagation Module (PPM), a µarchitectural scheme that propagates the page size information to the lower-level cache prefetchers, enabling safe prefetching beyond 4KB physical page boundaries when the accessed blocks reside in large pages, at the cost of augmenting the first-level caches’ Miss Status Holding Register (MSHR) entries with one additional bit. PPM is compatible with any cache prefetcher without implying design modifications. We capitalize on PPM’s benefits by designing a module that consists of two page size aware prefetchers that inherently use different page sizes to drive prefetching. The composite module uses adaptive logic to dynamically enable the most appropriate page size aware prefetcher. Finally, we show that the proposed designs are transparent to which cache prefetcher is used. We apply the proposed page size exploitation techniques to four state-of-the-art spatial cache prefetchers. Our evalua- tion shows that our proposals improve single-core geomean performance by up to 8.1% (2.1% at minimum) over the original implementation of the considered prefetchers, across 80 memory-intensive workloads. In multi-core contexts, we report geomean speedups up to 7.7% across different cache prefetchers and core configurations., This work is supported by the Spanish Ministry of Science and Technology through the PID2019-107255GB project, the Generalitat de Catalunya (contract 2017-SGR-1414), the European Union Horizon 2020 research and innovation program under grant agreement No 955606 (DEEP-SEA EU project), the National Science Foundation through grants CNS-1938064 and CCF-1912617, and the Semiconductor Research Corporation project GRC 2936.001. Georgios Vavouliotis has been supported by the Spanish Ministry of Economy, Industry, and Competitiveness and the European Social Fund under the FPI fellowship No. PRE2018-087046. Marc Casas has been partially supported by the Grant RYC2017-23269 funded by MCIN/AEI/10.13039/501100011033 and ESF ‘Investing in your future’., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Scanflow: an end-to-end agent-based autonomic ML workflow manager for clusters

  • Liu, Peini
  • Bravo Rocca, Gusseppe Jesus
  • Guitart Fernández, Jordi
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Machine Learning (ML) is more than just training models, the whole life-cycle must be considered. Once deployed, a ML model needs to be constantly managed, supervised and debugged to guarantee its availability, validity and robustness in dynamic contexts. This demonstration presents an agent-based ML workflow manager so-called Scanflow1, which enables autonomic management and supervision of the end-to-end life-cycle of ML workflows on distributed clusters. The case study on a MNIST project2 shows that different teams can collaborate using Scanflow within a ML project at different phases, and the effectiveness of agents to maintain the model accuracy and throughput of the model serving while running in production., This work was partially supported by Lenovo as part of LenovoBSC 2020 collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Scanflow-K8s: agent-based framework for autonomic management and supervision of ML workflows in Kubernetes clusters

  • Liu, Peini
  • Bravo Rocca, Gusseppe
  • Guitart Fernández, Jordi
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Machine Learning (ML) projects are currently heavily based on workflows composed of some reproducible steps and executed as containerized pipelines to build or deploy ML models efficiently because of the flexibility, portability, and fast delivery they provide to the ML life-cycle. However, deployed models need to be watched and constantly managed, supervised, and debugged to guarantee their availability, validity, and robustness in unexpected situations. Therefore, containerized ML workflows would benefit from leveraging flexible and diverse autonomic capabilities. This work presents an architecture for autonomic ML workflows with abilities for multi-layered control, based on an agent-based approach that enables autonomic management and supervision of ML workflows at the application layer and the infrastructure layer (by collaborating with the orchestrator). We redesign the Scanflow ML framework to support such multi-agent approach by using triggers, primitives, and strategies. We also implement a practical platform, so-called Scanflow-K8s, that enables autonomic ML workflows on Kubernetes clusters based on the Scanflow agents. MNIST image classification and MLPerf ImageNet classification benchmarks are used as case studies to show the capabilities of Scanflow-K8s under different scenarios. The experimental results demonstrate the feasibility and effectiveness of our proposed agent approach and the Scanflow-K8s platform for the autonomic management of ML workflows in Kubernetes clusters at multiple layers., This work was supported by Lenovo as part of Lenovo-BSC 2020 collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Task-based programming models for heterogeneous recurrent workloads

  • Bosch Pons, Jaume
  • Vidal, Miquel
  • Filgueras Izquierdo, Antonio
  • Jiménez González, Daniel
  • Álvarez Martínez, Carlos
  • Martorell Bofill, Xavier
  • Ayguadé Parra, Eduard
This paper proposes the extension of task-based programming models with recurrent workloads concepts. The proposal introduces new clauses in the OmpSs task directive to efficiently model recurrent workloads. The clauses define the task period and/or the number of task body repetitions. Despite the new clauses are suitable for any device, their support has been implemented using the capabilities of FPGA devices in embedded systems. These heterogeneous systems are common in industrial applications that usually develop recurrent workloads. The evaluation shows a huge gap in the applications' programmability, saving lines of code, and increasing the code readability. Besides, it shows the efficient management of recurrent tasks when performed in FPGA devices, which can support one order of magnitude finer tasks. All these improvements perfectly suit the needs of cyber-physical heterogeneous systems, which are frequently used in industrial environments to run recurrent workloads., This work is partially supported by the European Union H2020 Research and Innovation Action (project 801051), by the Spanish Government (projects SEV-2015-0493 and PID2019-107255GB, grant BES-2016-078046), and by the Generalitat de Catalunya (contracts 2017-SGR-1414 and 2017-SGR-1328)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Human-in-the-loop online multi-agent approach to increase trustworthiness in ML models through trust scores and data augmentation

  • Bravo Rocca, Gusseppe Jesus
  • Liu, Peini
  • Guitart Fernández, Jordi
  • Dholakia, Ajay
  • Ellison, David
  • Hodak, Miroslav
Increasing a ML model accuracy is not enough, we must also increase its trustworthiness. This is an important step for building resilient AI systems for safety-critical applications such as automotive, finance, and healthcare. For that purpose, we propose a multi-agent system that combines both machine and human agents. In this system, a checker agent calculates a trust score of each instance (which penalizes overconfidence in predictions) using an agreement-based method and ranks it; then an improver agent filters the anomalous instances based on a human rule-based procedure (which is considered safe), gets the human labels, applies geometric data augmentation, and retrains with the augmented data using transfer learning. We evaluate the system on corrupted versions of the MNIST and FashionMNIST datasets. We get an improvement in accuracy and trust score with just few additional labels compared to a baseline approach., This work was supported by Lenovo as part of LenovoBSC 2020 collaboration agreement, by the Spanish Government under contracts PID2019-107255GB-C21 and PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

GPU4S: Major project outcomes, lessons learnt and way forward

  • Kosmidis, Leonidas
  • Rodríguez Ferrández, Iván
  • Jover Álvarez, Álvaro
  • Alcaide Portet, Sergi
  • Lachaize, Jérôme
  • Notebaert, Olivier
  • Certain, Antoine
  • Steenari, David
Embedded GPUs have been identified from both private and government space agencies as promising hardware technologies to satisfy the increased needs of payload processing. The GPU4S (GPU for Space) project funded from the European Space Agency (ESA) has explored in detail the feasibility and the benefit of using them for space workloads. Currently at the closing phases of the project, in this paper we describe the main project outcomes and explain the lessons we learnt. In addition, we provide some guidelines for the next steps towards their adoption in space., This work is funded by ESA under the GPU4S (GPU for Space) project (ITT AO/1-9010/17/NL/AF). It is also partially supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under grants PID2019-107255GB and FJCI-2017-34095 and HiPEAC., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Towards data-flow parallelization for adaptive mesh refinement applications

  • Sala Penadés, Kevin
  • Rico Carro, Alejandro
  • Beltran Querol, Vicenç
Adaptive Mesh Refinement (AMR) is a prevalent method used by distributed-memory simulation applications to adapt the accuracy of their solutions depending on the turbulent conditions in each of their domain regions. These applications are usually dynamic since their domain areas are refined or coarsened in various refinement stages during their execution. Thus, they periodically redistribute their workloads among processes to avoid load imbalance. Although the defacto standard for scientific computing in distributed environments is MPI, in recent years, pure MPI applications are being ported to hybrid ones, attempting to cope with modern multi-core systems. Recently, the Task-Aware MPI library was proposed to efficiently integrate MPI communications and tasking models, providing also the transparent management of communications issued by tasks. In this paper, we demonstrate the benefits of porting AMR applications to data-flow programming models leveraging that novel hybrid approach. We exploit most of the application parallelism by taskifying all stages, allowing their natural overlap. We employ these techniques on the miniAMR proxy application, which mimics the refinement, load balancing, communication, and computation patterns of general AMR applications. We evaluate how this approach reduces the time in its computation and communication phases while achieving better programmability than other conventional hybrid techniques., This work has been supported by the European Union H2020 Programme through the DEEP-EST project (agreement No. 754304), the Spanish Government through the Severo Ochoa Program (SEV-2015-0493), the Spanish Ministry of Science and Innovation (PID2019-107255GB), and the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Experience replay and zero-shot clustering for continual learning in diabetic retinopathy detection

  • Bravo Rocca, Gusseppe Jesus
  • Liu, Peini
  • Guitart Fernández, Jordi
  • Dholakia, Ajay
  • Ellison, David
  • Carrillo Larco, Rodrigo M.
We present an approach to mitigate catastrophic forgetting in Continual Learning (CL), focusing on domain incremental scenarios in medical imaging. Our method leverages Large Language Models (LLMs) to generate task-agnostic descriptions from multimodal inputs, enabling zero-shot clustering of tasks without supervision. This clustering underpins an enhanced Experience Replay (ER) strategy, strategically sampling data points to refresh the model’s memory while preserving privacy. By incrementally updating a multi-head classifier using only data embeddings, our approach maintains both efficiency and data confidentiality. Evaluated on a challenging diabetic retinopathy dataset, our method demonstrates significant improvements over traditional CL techniques, including Elastic Weight Consolidation (EWC), Gradient Episodic Memory (GEM), and Learning Without Forgetting (LWF). Extensive experiments across Multi-Layer Perceptron (MLP), Residual, and Attention architectures show consistent performance gains (up to 3.1% in Average Mean Class Accuracy) and reduced forgetting, with only 6% computational overhead. These results highlight our approach’s potential for privacy-preserving, efficient CL in sensitive domains like healthcare, offering a promising direction for developing adaptive AI systems that can learn continuously while respecting data privacy constraints., This work was partially supported by Lenovo and Intel as part of the Lenovo AI Innovators University Research program, by the Spanish Ministry of Science (MICINN), the Research State Agency (AEI) and European Regional Development Funds (ERDF/FEDER) under grant agreements PID2019-107255GB-C22 and PID2021-126248OBI00, MCIN/AEI/10.13039/ 501100011033/FEDER, UE, and by the Generalitat de Catalunya under contract 2021-SGR-00478., Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

On the relation between open project-based learning in undergraduate computer science education and contemporary technological trends

  • Tous Liesa, Rubén
  • Freitag, Fèlix
  • Berral García, Josep Lluís
Amidst rapid and constant technological change, keeping IT higher education curricula up to date is becoming increasingly challenging. For a decade, the course”Project on Information Technologies” within the undergraduate computer science studies offered by a major technical university has been pursuing an effort of continuous curriculum adaptation and active learning practices based on an open project-based learning (PBL) methodology. This article researches the implications of employing an open-ended PBL approach, with a specific focus on the alignment of acquired skills with the trends in the professional landscape. Our analysis has identified a strong correlation between the technologies utilized in the projects and the contemporary technological trends in the areas of global technological focus, programming languages, server-side technologies, database management systems, and DevOps-related tools. For the study, we analyzed empirical data gathered from over 100 projects involving more than 400 students who enrolled in this reference course, belonging to the last year of the IT higher education programme in the last 10 years. The results suggest the open project-based design as a teaching means in the student’s study plan for fostering the student the learning of prevailing current practical technologies., This work is partially supported by the Span-ish Ministry of Science under contracts PID2019-107255GB, PID2019-106774RB-C21, PID2021-126248OB-I00 (MCIN/AEI/10.13039/501100011033/FEDER), PDC2023-145809-I00 (PDC/AEI/10.13039/501100011033), the Recovery and Resilience Facility of the European Union, and by the SGR programme of the Catalan Government under contracts 2021-SGR-00478 and 2021-SGR-01059., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

How2Sign: A large-scale multimodal dataset for continuous American sign language

  • Cardoso Duarte, Amanda
  • Palaskar, Shruti
  • Ventura Ripol, Lucas
  • Ghadiyaram, Deepti
  • DeHaan, Kenneth
  • Metze, Florian
  • Torres Viñals, Jordi
  • Giró Nieto, Xavier
One of the factors that have hindered progress in the areas of sign language recognition, translation, and production is the absence of large annotated datasets. Towards this end, we introduce How2Sign, a multimodal and multiview continuous American Sign Language (ASL) dataset, consisting of a parallel corpus of more than 80 hours of sign language videos and a set of corresponding modalities including speech, English transcripts, and depth. A three-hour subset was further recorded in the Panoptic studio enabling detailed 3D pose estimation. To evaluate the potential of How2Sign for real-world impact, we conduct a study with ASL signers and show that synthesized videos using our dataset can indeed be understood. The study further gives insights on challenges that computer vision should address in order to make progress in this field. Dataset website: http://how2sign.github.io/, This work received funding from Facebook through gifts to CMU and UPC; through projects TEC2016-75976-R, TIN2015- 65316-P, SEV-2015-0493 and PID2019-107255GB-C22 of the Spanish Government and 2017-SGR-1414 of Generalitat de Catalunya. This work used XSEDE’s “Bridges” system at the Pittsburgh Supercomputing Center (NSF award ACI- 1445606). Amanda Duarte has received support from la Caixa Foundation (ID 100010434) under the fellowship code LCF/BQ/IN18/11660029. Shruti Palaskar was supported by the Facebook Fellowship program., Peer Reviewed, Objectius de Desenvolupament Sostenible::10 - Reducció de les Desigualtats, Objectius de Desenvolupament Sostenible::4 - Educació de Qualitat::4.5 - Per a 2030, eliminar les disparitats de gènere en l’educació i garantir l’accés en condicions d’igualtat a les persones vulnerables, incloses les persones amb discapacitat, els pobles indígenes i els nens i nenes en situacions de vulnerabilitat, a tots els nivells de l’ensenyament i la formació professional, Objectius de Desenvolupament Sostenible::10 - Reducció de les Desigualtats::10.2 - Per a 2030, potenciar i promoure la inclusió social, econòmica i política de totes les persones, independentment de l’edat, sexe, discapacitat, raça, ètnia, origen, religió, situació econòmica o altra condició, Objectius de Desenvolupament Sostenible::4 - Educació de Qualitat, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Comparison of GPU computing methodologies for safety-critical systems: an avionics case study

  • Benito Bermúdez, Marc
  • Trompouki, Matina Maria
  • Kosmidis, Leonidas
  • García Martín, David
  • Carretero Jiménez, Sergio
  • Wenger, Ken
Introducing advanced functionalities in safety-critical systems requires using more powerful architectures such as GPUs. However software in safety-critical industries is subject to functional certification, which cannot be achieved using standard GPU programming languages such as CUDA and OpenCL. Fortunately, GPUs are already used in certified critical systems for display tasks, using safety-certified solutions such as OpenGL SC 2.0. In this paper, we compare two state-of-the art graphics-based methodologies, OpenGL SC 2.0 and Brook Auto/BRASIL for the implementation of a prototype avionics case study. We evaluate both methods on a realistic industrial setup, composed by an avionics-grade GPU and a safety-certified GPU driver in terms of development metrics and performance, showing their feasibility., This work was funded by the Airbus TANIA-GPU Project ADS (E/200). It was also partially supported by the Spanish Ministry of Economy and Competitiveness under grants PID2019-107255GB and FJCI-2017-34095 and HiPEAC., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Analyzing the performance of hierarchical collective algorithms on ARM-based multicore clusters

  • Utrera Iglesias, Gladys Miriam
  • Gil, Marisa
  • Martorell Bofill, Xavier
MPI is the de facto communication standard library for parallel applications in distributed memory architectures. Collective operations performance is critical in HPC applications as they can become the bottleneck of their executions. The advent of larger node sizes on multicore clusters has motivated the exploration of hierarchical collective algorithms aware of the process placement in the cluster and the memory hierarchy. This work analyses and compares several hierarchical collective algorithms from the literature that do not form part of the current MPI standard. We implement the algorithms on top of OpenMPI using the shared-memory facility provided by MPI-3 at the intra-node level and evaluate them on ARM-based multicore clusters. From our results, we evidence aspects of the algorithms that impact the performance and applicability of the different algorithms. Finally, we propose a model that helps us to analyze the scalability of the algorithms., This work has been supported by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Feature engineering for agents: an adaptive cognitive architecture for interpretable ML monitoring

  • Bravo Rocca, Gusseppe Jesus
  • Liu, Peini
  • Guitart Fernández, Jordi
  • Carrillo Larco, Rodrigo M.
  • Dholakia, Ajay
  • Ellison, David
Monitoring Machine Learning (ML) models in production environments is crucial, yet traditional approaches often yield verbose, low-interpretability outputs that hinder effective decision-making. We propose a cognitive architecture for ML monitoring that applies feature engineering principles to agents based on Large Language Models (LLMs), significantly enhancing the interpretability of monitoring outputs. Central to our approach is a Decision Procedure module that simulates feature engineering through three key steps: Refactor, Break Down, and Compile. The Refactor step improves data representation to better capture feature semantics, allowing the LLM to focus on salient aspects of the monitoring data while reducing noise and irrelevant information. Break Down decomposes complex information for detailed analysis, and Compile integrates sub-insights into clear, interpretable outputs. This process leads to a more deterministic planning approach, reducing dependence on LLM-generated planning, which can sometimes be inconsistent and overly general. The combination of feature engineering-driven planning and selective LLM utilization results in a robust decision support system, capable of providing highly interpretable and actionable insights. Experiments using multiple LLMs demonstrate the efficacy of our approach, achieving significantly higher accuracy compared to various baselines across several domains., This work was partially supported by Lenovo and Intel® as part of the Lenovo AI Innovators University Research program, by Spanish Ministry of Science (MICINN), the Research State Agency (AEI) and European Regional Development Funds (ERDF/FEDER) under contracts PID2019- 107255GB-C22 and PID2021-126248OB-I00, and by the Generalitat de Catalunya (AGAUR) under contract 2021-SGR-00478., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Energy optimization and analysis with EAR

  • Corbalán González, Julita
  • Alonso Jane, Lluís
  • Aneas Gómez, Jordi
  • Brochard, Luigi
EAR is an energy management framework which offers three main services: energy accounting, energy control, and energy optimization. The latter is done through the EAR runtime library (EARL). EARL is a dynamic, transparent, and lightweight runtime library that provides energy optimisation and control. EARL optimises energy by selecting the optimal CPU frequency, based on the energy policy selected and application runtime characteristics without any application modification or user input. Currently EARL only works for MPI applications but EAR itself can still operate for non-MPI applications. It automatically (and transparently) identifies iterative regions (loops) and computes a set of metrics per iteration, application signature, and, together with the system signature, applies energy models to estimate the execution time and power for the CPU frequencies available. System signature is a set of coefficients per-node computed during EAR installation via a learning phase. Given time and power projections, EARL selects the best frequency based on policy settings. This papers shows how to optimize energy using the EAR library with min_time_to_solution energy policy and how to analyse applications through EAR framework. Evaluation includes eight applications with different sizes and application signatures. Results show how EARL computes each application signature on the fly and applies the CPU frequency selected by the min_time_to_solution policy., This work has been mostly funded by the BSC-Lenovo collaboration agreement and partially by the Spanish Government through Programa Severo Ochoa (SEV-2015-0493), by the Spanish Ministry of Science and Technology through PID2019-107255GB project, by the Generalitat de Catalunya (2017- SGR-1414), Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

HyCSim: A rapid design space exploration tool for emerging hybrid last-level caches

  • Escuín Blasco, Carlos
  • Ali Khan, Asif
  • Ibáñez Marín, Pablo Enrique
  • Monreal Arnal, Teresa
  • Viñals Yúfera, Victor
  • Castrillón, Jerónimo
Recent years have seen a rising trend in the exploration of non-volatile memory (NVM) technologies in the memory subsystem. Particularly in the cache hierarchy, hybrid last-level cache (LLC) solutions are proposed to meet the wide-ranging performance and energy requirements of modern days applications. These emerging hybrid solutions need simulation and detailed exploration to fully understand their capabilities before exploiting them. Existing simulation tools are either too slow or incapable of prototyping such systems and optimizing for NVM devices. To this end, we propose HyCSim, a trace-driven simulation infrastructure that enables rapid comparison of various hybrid LLC configurations for different optimization objectives. Notably, HyCSim makes it possible to quickly estimate the impact of various hybrid LLC insertion and replacement policies, disabling of a cache region at byte or cache frame granularity for different fault maps. In addition, HyCSim allows to evaluate the impact of various compression schemes on the overall performance (hit and miss rate) and the number of writes to the LLC. Our evaluation on ten multi-program workloads from the SPEC 2006 benchmarks suite shows that HyCSim accelerates the simulation time by 24×, compared to the cycle-accurate Gem5 simulator, with high-fidelity., This work was partially funded by the HiPEAC collaboration grant 2020, the German Research Council (DFG) through the TraceSymm project (366764507) and the Co4RTM project (450944241), MCIN/AEI/10.13039/501100011033 (grants PID2019-105660RB-C21 and PID2019- 107255GB-C22), and by Aragón Government (T5820R research group)., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Improving performance of HPC kernels on FPGAs using high-level resource management

  • Filgueras Izquierdo, Antonio
  • Vidal, Miquel
  • Jiménez González, Daniel
  • Álvarez Martínez, Carlos
  • Martorell Bofill, Xavier
In state-of-the-art FPGA, especially in chiplet-based devices, place and route has become an important challenge due to an increase in device size and complexity. In the same way, off-chip memory resources have grown in size and number of memory modules. Making efficient use of them has become a difficult task., This work is supported by the TEXTAROSSA project G.A. n.956831, as part of the EuroHPC initiative, by the Spanish Government (Grants PCI2021-121964 - TEXTAROSSA, PID2019-107255GB-C21, PID2019-107255GB-C22, MCIN/AEI/10.13039/501100011033, and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Empirical evidence for MPSoCs in critical systems: The case of NXP’s T2080 cache coherence

  • Pujol Torramorell, Roger
  • Tabani, Hamid
  • Abella Ferrer, Jaume
  • Hassan, Mohamed
  • Cazorla Almeida, Francisco Javier
The adoption of complex MPSoCs in critical real-time embedded systems mandates a detailed analysis their architecture to facilitate certification. This analysis is hindered by the lack of a thorough understanding of the MPSoC system due to the unobvious and/or insufficiently documented behavior of some key hardware features. Confidence on those features can only be regained by building specific tests to both, assess whether their behavior matches specifications and unveil their behavior when it is not fully known a priori. In this work, we introduce a systematic approach that constructs this thorough understanding of the MPSoC architecture-- and assess against its specification in processor documentation -- with a focus on the cache coherence protocol in the avionics-relevant NXP T2080 architecture as our use-case. Our approach covers all transitions in the MESI cache coherence protocol, with emphasis on the coherence between DMA and processing cores. We build evidence of their behavior based on available debug support and performance monitors. Our analysis discloses unexpected behavior for coherence-related notifications as well as some hardware monitors., This work has been partially supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB; the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 878752 (MASTECS) and the European Research Council (ERC) grant agreement No. 772773 (SuPerCom); the HiPEAC Network of Excellence; and the Natural Sciences and Engineering Research Council of Canada (NSERC)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

High-resolution image generation using artificial intelligence and diffusion modelling

  • Scoles García, Amanda
  • Sionis Ponte, Giulia
  • Otero Calviño, Beatriz
  • Utrera Iglesias, Gladys Miriam
Super-Resolution via Repeated Refinement (SR3) is a state-of-the-art super-resolution algorithm based on diffusion model that can enhance the resolution of images. This method is used to pre-trained models on large datasets and can be used for various tasks without requiring training from scratch. Training SR3 from scratch using the ImageNet dataset involves a complex process that requires substantial computational resources and expertise. The idea is applied the trained SR3 model to new images by feeding the low-resolution inputs and obtaining the high-resolution outputs. It's important to note that training SR3 from scratch is a resource-intensive process that requires powerful GPUs and significant computation time. If you do not have access to such resources, an alternative is to use pre-trained models that are already available and fine-tune them on specific datasets or tasks. The paper shows the result of comparing the resolution of the preprocessed images using a significantly smaller number of images to perform the training with those obtained using the pre-trained model. The results obtained show acceptable results without having to perform on large datasets minimizing the computation time to obtain the resolution of images., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2021-124463OB-100 and PID2019-107255GB-C22 by the Generalitat de Catalunya under grant 2021-SGR-00326 and 2021-SGR-01007. Finally, the research leading to these results also has received funding from the European Union's Horizon 2020 research and innovation program under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

DO-178C certification of general-purpose GPU software: review of existing methods and future directions

  • Trompouki, Matina Maria
  • Kosmidis, Leonidas
—General-Purpose GPU software is considered for use in avionics to satisfy the increased computational requirements of future systems. Therefore, it needs to be certified following the DO-178C guidance as all airborne software. In this work, we review the existing methods in the literature, we analyse their advantages and disadvantages, and we discuss how they can be combined to obtain certification with lower effort and cost. Our focus is restricted on application-level software, under the premise that successful completion of verification of avionics graphics GPU software products has been demonstrated, so their GPU compiler has been considered acceptable for these already DO-178C certified products, or existing qualified GPU compilers exist. Finally, we discuss upcoming solutions for certified general purpose GPU computing ., This work was performed within the Airbus TANIAGPU Project ADS (E/200) in collaboration with the project partners Airbus Defence and Space, Madrid, Spain and CoreAVI, Canada. It was also partially supported by the European Space Agency (ESA) through the GPU4S (GPU for Space) activity, the Spanish Ministry of Economy and Competitiveness under grants PID2019-107255GB and FJCI-2017-34095 (Spanish State Research Agency / http://dx.doi.org/10.13039/501100011033) and the HiPEAC Network of Excellence., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Parallel and distributed protein processing for 3D-protein pattern discovery and clustering

  • Valdés Jiménez, Alejandro Mauricio
  • Núñez Vivanco, Gabriel
  • Jiménez González, Daniel
The discovery and clustering of three-dimensional protein patterns in a set of protein structures, without relying on predefined search patterns, can be highly beneficial for predicting the functions of unknown proteins and facilitating rational multi-target drug design. This work introduces a novel OpenMP parallelization of the 3D-PP algorithm using explicit and nested tasks, which balances data sharing synchronization and load unbalance, improving previous implementations based on implicit tasks. Given the vast number of protein structures available in the Protein Data Bank (over 231,000 from PDB and more than 1,068,000 from AlphaFold), processing large datasets locally can be constrained by available resources such as main memory and secondary storage. To address this challenge, we propose parallel approaches to the 3D-PP algorithm for distributed memory systems (using MPI) and hybrid systems (combining MPI and OpenMP). The evaluated strategies distribute the workload based on the entire protein structure. Experimental results using a dataset of 8,344 protein structures show that the new OpenMP taskified version is 1.6x faster than the previous best OpenMP implementation. Moreover, the hybrid MPI+OpenMP version achieves a speedup of up to 162.5x, demonstrating its scalability and efficiency for large-scale protein structure analysis., This research was funded by the Spanish Government (Grants PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007). We also want to thank to the Barcelona Supercomputing Center for the use of the CTE-AMD cluster., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Exploring hierarchical MPI reduction collective algorithms targeted to multicore node clusters

  • Utrera Iglesias, Gladys Miriam
  • Gil, Marisa
  • Martorell Bofill, Xavier
  • Spataro, William
  • Giordano, Andrea
High-performance computing applications heavily rely on message-passing mechanisms for data sharing in cluster environments, and the MPI library stands as the default communication library for parallel applications. Significant efforts have been directed toward optimizing data distribution and buffering based on size. This optimization aims to enhance communication performance and prevent issues such as running out of memory on the target node. Furthermore, the emergence of multicore clusters with larger node sizes has stimulated the investigation of hierarchical collective algorithms that consider the placement of processes within the cluster and the memory hierarchy. This paper studies and compares the performance of the algorithm of the reduction collective from the literature, specifically several implementations that do not form part of the current MPI standard, which tackle this issue. We implement the algorithms on top of Intel MPI and OpenMPI libraries using the MPI profiling interface. Experimental results with the Intel MPI Benchmarks on a multicore cluster, Intel Platinum processor-based and OmniPath interconnection network show much room for improvement in the performance of collectives depending on the message sizes., The authors acknowledge the support of the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR01007)., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

MUCH: exploiting pairwise hardware event monitor correlations for improved timing analysis of complex MPSoCs

  • Vilardell Moreno, Sergi
  • Serra Mochales, Isabel
  • Mezzetti, Enrico
  • Abella Ferrer, Jaume
  • Cazorla Almeida, Francisco Javier
Measurement-based timing analysis techniques increasingly rely on the Performance Monitoring Units (PMU) of MPSoCs, as these units implement specialized Hardware Event Monitors (HEMs) that convey detailed information about multicore interference in hardware shared resources. Unfortunately, there is an evident mismatch between the large number of HEMs (typically several hundreds) and the comparatively small number (normally less than ten) of Performance Monitoring Counters (PMCs) that can be configured to track HEMs in the PMU. Timing analysis normally require to observe a non-negligible number of HEMs per task from the same execution. However, due to the small number of PMCs, HEMs are necessarily collected across multiple runs that, despite intended to repeat the same experiment, carry out some significant variability (above 50% for some HEMs in relevant MPSoCs) caused by platform-intrinsic execution conditions. Therefore, blindly merging HEMs from different runs is not acceptable since they may easily correspond to significantly different conditions. To tackle this issue, the HRM approach has been proposed recently to merge HEMs from different runs accurately preserving their correlation w.r.t. one anchor HEM (i.e. processor cycles) building on order statistics. However, HRM do not always preserves the correlation between other pairs of HEMs that might be lost to a large extent. This paper copes with HRM limitations by proposing the MUlti-Correlation HEM reading and merging approach (MUCH). MUCH builds on multivariate Gaussian distributions to merge HEMs from different runs while preserving pairwise correlations across each individual pair of HEMs simultaneously. Our results on an NXP T2080 MPSoC used for avionics systems show that MUCH largely outperforms HRM for an identical number of input runs., This work has been partially supported by the Spanish Ministry of Science and Innovation under grant PID2019-107255GB, the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 772773) and the HiPEAC Network of Excellence., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

MNEMOSENE++: Scalable multi-tile design with enhanced buffering and VGSOT-MRAM based compute-in-memory crossbar array

  • Escuín Blasco, Carlos
  • García Redondo, Fernando
  • Zahedi, Mahdi
  • Ibáñez Marín, Pablo Enrique
  • Monreal Arnal, Teresa
  • Viñals Yúfera, Victor
  • Llaberia Griñó, José M.
  • Myers, James
  • Ryckaert, Julien
  • Biswas, Dwaipayan
  • Catthoor, Francky
This paper optimizes the MNEMOSENE architecture, a compute-in-memory (CiM) tile design integrating computation and storage for increased efficiency. We identify and address bottlenecks in the Row Data (RD) buffer that cause losses in performance. Our proposed approach includes mitigating these buffering bottlenecks and extending MNEMOSENE’s single-tile design to a multi-tile configuration for improved parallel processing. The proposal is validated through comprehensive analyses exploring the mapping of diverse neural networks evaluated on CiM crossbar arrays based on NVM technologies. These proposed enhancements lead up to 55% reduction in execution time compared to the original single-tile architecture for any general matrix multiplication (GEMM) operation. Our evaluation shows that while ReRAM and PCM offer notable energy advantages, their integration with scaled CMOS is limited, which leads to VGSOT-MRAM emerging as a promising alternative due to its good balance between energy efficiency and superior integration capabilities. The VGSOT-MRAM crossbar arrays provide 12×,49×, and 346× more energy efficiency than PCM, ReRAM, and STT-MRAM ones, respectively. It translates, on average for the considered workload, in 1.5×,3×, and 14.5× better energy efficiency of the entire system., This project has received funding from the ECSEL Joint Undertaking (JU) under grant agreement No 876925, from MCIN/AEI/10.13039/501100011033 (grants PID2019-105660RB-C21 and PID2019-107255GB-C22), and from Aragon Government (T58_23R research group)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Exploring image transformations with diffusion models: A survey of applications and implementation code

  • Arellano García, Silvia
  • Otero Calviño, Beatriz
  • Tous Liesa, Rubén
Diffusion Models have become increasingly popular in recent years and their applications span a wide range of fields. This survey focuses on the use of diffusion models in computer vision, specially in the branch of image transformations. The objective of this survey is to provide an overview of state-of-the-art applications of diffusion models in image transformations, including image inpainting, super-resolution, restoration, translation, and editing. This survey presents a selection of notable papers and repositories including practical applications of diffusion models for image transformations. The applications are presented in a practical and concise manner, facilitating the understanding of concepts behind diffusion models and how they function. Additionally, it includes a curated collection of GitHub repositories featuring popular examples of these subjects., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB and PID2021-124463OB- IOO, by the Generalitat de Catalunya under grants 2021-SGR-00478 and 2021-SGR-00326. Finally, the research leading to these results also has received funding from the European Union’s Horizon 2020 research and innovation programme under the HORIZON-EU VITAMIN-V (101093062) project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Partitioned reduction for heterogeneous environments

  • De Rango, Alessio
  • Utrera Iglesias, Gladys Miriam
  • Gil, Marisa
  • Martorell Bofill, Xavier
  • Giordano, Andrea
  • D'Ambrosio, Donato
  • Mendicino, Giuseppe
Nowadays, performance in HPC applications focuses on MPI efficiency as the de facto message-passing library to exploit parallelism. Features such as multithread and communication and processing overlap are continuously studied to adapt to new platforms and a more significant number of processing units like GPU platforms. In this sense, recently, the MPI-4.0 standard introduced the partitioned point-to-point communication primitives to potentiate computation and communication overlapping. This paper introduces an innovative extension to MPI, specifically addressing partitioned communication for MPI-reduction primitives. Traditional reduction tasks conventionally involve processing the complete input vector following the conclusion of GPU computations. In contrast, our proposed methodology exploits message partitioning to process reduction tasks in real-time incrementally. This approach allows the system to process individual partitions of the input vector as they become available, removing the necessity to await the full completion of GPU computations before initiating the reduction. Our results demonstrate promising benefits, particularly for large message sizes. However, it is essential to acknowledge that optimizations at synchronization points remain potential bottlenecks, requiring meticulous analysis and consideration., This work was supported by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR-01007). This work was partially supported by PON ”Research and Innovation” 2014-2020 CUP H25F21001220006, C. I. 1062 R22 INNOVAZIONE, Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Cost-aware prediction of uncorrected DRAM errors in the field

  • Boixaderas Coderch, Isaac
  • Živanovič, Darko
  • Moré Codina, Sergi
  • Bartolomé Rodríguez, Javier
  • Vicente Dorca, David
  • Casas, Marc
  • Carpenter, Paul Matthew
  • Radojković, Petar
  • Ayguadé Parra, Eduard
This paper presents and evaluates a method to predict DRAM uncorrected errors, a leading cause of hardware failures in large-scale HPC clusters. The method uses a random forest classifier, which was trained and evaluated using error logs from two years of production of the MareNostrum 3 supercomputer. By enabling the system to take measures to mitigate node failures, our method reduces lost compute time by up to 57%, a net saving of 21,000 node–hours per year. We release all source code as open source. We also discuss and clarify aspects of methodology that are essential for a DRAM prediction method to be useful in practice. We explain why standard evaluation metrics, such as precision and recall, are insufficient, and base the evaluation on a cost–benefit analysis. This methodology can help ensure that any DRAM error predictor is clear from training bias and has a clear cost–benefit calculation., This work was supported by the Spanish Ministry of Science and Technology (project PID2019-107255GB), Generalitat de Catalunya (contracts 2014-SGR-1051 and 2014-SGR-1272) and the European Union’s Horizon 2020 research and innovation programme and EuroEXA project (grant agreement No 754337). Paul Carpenter and Marc Casas hold the Ramon y Cajal fellowship under contracts RYC2018-025628-I and RYC2017-23269, respectively, of the Ministry of Economy and Competitiveness of Spain., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Epicentral region estimation using convolutional neural networks

  • Cruz de la Cruz, Stalin Leonel
  • Tous Liesa, Rubén
  • Otero Calviño, Beatriz
  • Alvarado Bermúdez, Leonardo
  • Mus León, Sergi
  • Rojas Ulacio, Otilio Jose
Recent works have assessed the capability of deep neural networks of estimating the epicentral source region of a seismic event from a single-station three-channel signal. In all the cases, the geographical partitioning is performed by automatic tessellation algorithms such as the Voronoi decomposition. This paper evaluates the hypothesis that the source region estimation accuracy is significantly increased if the geographical partitioning is performed considering the regional geological characteristics such as the tectonic plate boundaries. Also, it raises the transformation of the training data to increase the accuracy of the predictive model based on a Projected Coordinate Reference (PCR) System. A deep convolutional neural network (CNN) is applied over the data recorded by the broadband stations of the Venezuelan Foundation of Seismological Research (FUNVISIS) in the region of 9.5 to 11.5ºN and 67.0 to 69.0ºW between April 2018 and April 2019. In order to estimate the epicentral source region of a detected event, several geographical tessellations provided by seismologists from the area are employed. These tessellations, with different number of partitions, consider the fault systems of the study region (San Sebastián, La Victoria and Morón fault systems). The results are compared to the ones obtained with automatic partitioning performed by the k-means algorithm., This work is partially supported by the Spanish Ministry of Economy and Competitivity under contract TIN2015-65316-P, by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB-C22, and by the SGR programmes (2014-SGR-1051 and 2017-SGR-962) of the Catalan Government and has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 777778 (MATHROCKS)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Cellular automata on a multi-GPU architecture: A technical overview

  • Giordano, Andrea
  • De Rango, Alessio
  • D'Ambrosio, Donato
  • Gil, Marisa
  • Macri, Davide
  • Martorell Bofill, Xavier
  • Rongo, Rocco
  • Utrera Iglesias, Gladys Miriam
  • Mendicino, Giuseppe
  • Spataro, William
This work is focused on the transparent execution of Cellular Automata models on a multi-GPU architecture. Although Cellular Automata models can be easily parallelized on a single GPU, the domain size and transition function complexity may require the use of multiple GPUs. Our goal is to allow modellers to be completely unaware of the parallel execution context, i.e., the code implementing the Cellular Automata model remains the same regardless if the execution is performed on CPU, single GPU, or multi-GPU systems. This paper supplies meaningful technical insights on how to ensure both transparency and efficiency in multi-GPU execution of Cellular Automata models. In particular, an object-oriented approach is exploited in which a transparent layer is devised that abstracts the parallelization details and allows a strong “separation of concerns” between the execution parallelism issues and the model implementation. Preliminary experiments have been carried out on the multi-GPU cluster CTE-POWER available at the Barcelona Supercomputing Center (BSC), witnessing good speedups notwithstanding the transparency feature supplied by our approach., This research was partially funded by the Italian “ICSC National Center for HPC, Big Data and Quantum Computing” Project, CN00000013 (approved under the Call M42C – Investment 1.4 – Avviso “Centri Nazionali” - D.D. n. 3138 of 16.12.2021, admitted to financing with MUR Decree n. 1031 of 06.17.2022). This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ’Innovation Ecosystems’, building ’Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work was supported by Italian Ministry of Research (MUR) under PNRR projects FAIR “Future AI ResearcH” - Spoke 9 - CUP H23C22000860006. This research was partially funded by the Spanish Ministry of Education (PID2019-107255GB-C22) and the Generalitat de Catalunya (2021-SGR-01007)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

TADIL: task-agnostic domain-incremental learning through task-ID inference using transformer nearest-centroid embeddings

  • Bravo Rocca, Gusseppe Jesus
  • Liu, Peini
  • Guitart Fernández, Jordi
  • Dholakia, Ajay
  • Ellison, David
Classical Machine Learning models have problems when faced with learning from data that changes over time or across domains due to factors such as noise, occlusion, illumination, or frequency variation, which humans can adapt to without being given independent and identically distributed data. Therefore, a Continual Learning (CL) approach is essential, particularly in this case, the Domain-Incremental Learning. This paper presents a novel pipeline for identifying tasks in domain-incremental learning scenarios without supervision. The pipeline consists of four steps. First, we obtain base embeddings from the raw data through a transformer-based existing model. Second, we group the embedding densities based on similarity and extract few nearest points to each cluster centroid. Third, we train an incremental task classifier using only the previous few points. Finally, we make the pipeline lightweight in terms of computational requirements and build an algorithm that decides when to learn a new task in an online way using the task classifier and drift detector. We experiment with the real-world driving dataset SODA10M and several CL strategies. We conclude that the performance of CL strategies with our pipeline is better not only when the task boundaries are given, but also in the more general practical case of task-agnostic strategies that demand identifying new tasks on-the-fly., We thank Lenovo for providing the technical infrastructure to run the experiments in this paper. This work was partially supported by Lenovo and Intel as part of the Lenovo AI Innovators University Research program, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2021-SGR-00478, Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Performance characterization of multi-container deployment schemes for online learning inference

  • Liu, Peini
  • Guitart Fernández, Jordi
  • Taherkordi, Amir
Online machine learning (ML) inference services provide users with an interactive way to request for predictions in realtime. To meet the notable computational requirements of such services, they are increasingly being deployed in the Cloud. In this context, the efficient provisioning and optimization of ML inference services in the Cloud is critical to achieve the required performance and meet the dynamic queries by end-users. Existing provisioning solutions focus on framework parameter tuning and infrastructure resources scaling, without considering deployments based on containerization technologies. The latter promises reproducibility and portability features for ML inferences services. There is limited knowledge about the impact of distinct deployment schemes at the container-level on the performance of online ML inference services, particularly on how to exploit multi-container deployments and its relation with processor and memory affinity. In light of this, in this paper we investigate experimentally the containerization of ML inference services and analyze the performance of multi-container deployments that partition the threads belonging to an online learning application into multiple containers in each node. This paper shares the findings and lessons learned from conducting realistic client patterns on an image classification model across numerous deployment configurations, especially including the impact of container granularity and its potential to exploit processor and memory affinity. Our results indicate that fine-grained multi-container deployments and affinity are useful for improving performance (both throughput and latency). In particular, our experiments on single-node and four-node clusters show up to 69% and 87% performance improvement compared to the single-container deployment, respectively., This work was partially supported by Lenovo as part of Lenovo-BSC collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2021-SGR-00478 and under grant 2020 FI-B 00257., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Efficiently running SpMV on long vector architectures

  • Gómez Crespo, Constantino
  • Mantovani, Filippo
  • Focht, Erich
  • Casas, Marc
Sparse Matrix-Vector multiplication (SpMV) is an essential kernel for parallel numerical applications. SpMV displays sparse and irregular data accesses, which complicate its vectorization. Such difficulties make SpMV to frequently experiment non-optimal results when run on long vector ISAs exploiting SIMD parallelism. In this context, the development of new optimizations becomes fundamental to enable high performance SpMV executions on emerging long vector architectures. In this paper, we improve the state-of-the-art SELL-C-s sparse matrix format by proposing several new optimizations for SpMV. We target aggressive long vector architectures like the NEC Vector Engine. By combining several optimizations, we obtain an average 12% improvement over SELL-C-s considering a heterogeneous set of 24 matrices. Our optimizations boost performance in long vector architectures since they expose a high degree of SIMD parallelism., The authors would like to acknowledge the support of NEC Corporation. This work is partially supported by the Spanish Ministry of Science and Technology through PID2019-107255GB project and by the Generalitat de Catalunya (contract 2017-SGR-1414). Marc Casas has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramon y Cajal fellowship number RYC-2017-23269., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Improving the discovery and clustering of three-dimensional protein patterns with OpenMP

  • Valdés Jiménez, Alejandro Mauricio
  • Reyes Parada, Miguel
  • Núñez Vivanco, Gabriel
  • Durán-Verdugo, Fabio
  • Jiménez González, Daniel
The discovery of conserved three-dimensional (3D) amino-acid patterns among a set of protein structures can be useful, for instance, to predict the functions of unknown proteins or for the rational design of multi-target drugs. There are several applications that perform a three-dimensional search of patterns in the structures of proteins. However, discovering conserved 3D patterns in a set of proteins with no other baseline patterns is a challenge. In this paper, we analyze and improve a state-of-the-art algorithm, 3D-PP, that implements this discovery. In this algorithm, the 3D patterns are detected and clustered using the root mean square deviation value, measured among each pair of 3D patterns (topological variability indicator). Even when 3D-PP deals with this task, the simultaneous processing of high amounts of proteins becomes a computational challenge with the size and the number of proteins to be evaluated. In this work, we present and analyze different shared memory parallel strategies of 3D-PP, using OpenMP. Those strategies improve the overall performance of the original implementation by reducing parallel load unbalance among threads and overall increasing parallelism. The results show significant performance improvements compared to the original version, achieving up to 13x speedup for a small number of proteins and 17.7× for a larger set., This research was funded by DICYT-USACH grant 5392102RP-AC and the Fondo Nacional de Desarrollo Cientifico y Tecnologico (FONDECYT) grants 1191133, Proyecto Fondecyt (Chile) 1220656 (MR-P), and by the Spanish Government (Grants PCI2021-121964 - TEXTAROSSA, PID2019-107255GB-C21/AEI/10.13039/501100011033, PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148-S), and by Generalitat de Catalunya (2021 SGR 01007). Additional tests were performed in the cluster obtained with the grant CONICYT-FONDEQUIP-EQM160063., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Hypervisor extension for a RISC-V processor

  • Gauchola Vilardell, Jaume
  • Costa Prats, Juan José
  • Morancho Llena, Enrique
  • Canal Corretger, Ramon
  • Carril Gil, Xavier
  • Doblas Font, Max
  • Otero Calviño, Beatriz
  • Pajuelo González, Manuel Alejandro
  • Rodríguez Luna, Eva
  • Salamero Sanz, Javier
  • Verdú Mulà, Javier
This paper describes our experience implementing a Hypervisor extension for a 64-bit RISC-V processor. We describe the design process and the main required parts with a brief explanation of each one., This work is partially supported by the DRAC (IU16-011591), the HORIZON Vitamin-V (101093062) and the Computación de Altas Prestaciones VIII (PID2019-107255GB) projects., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

AI-powered personalised learning platforms for EFL learning: preliminary results

  • Folgieri, Raffaella
  • Gil, Marisa
  • Bait, Miriam
  • Lucchiari, Claudio
Artificial intelligence (AI) has been increasingly integrated into the field of education, including personalised learning platforms. However, concerns have been raised about the potential of AI to replace human teachers and the impact on student agency and autonomy. In this research, we discuss the development of an AI-powered platform as a helper, not a substitution, for self-directed personal and professional growth. The present study investigates the effectiveness of an AI-powered personalised learning platform in enhancing self-directed learning and personal and professional growth. We also explored the role of human teaching and the ethical considerations of AI in education. A mixed-methods approach was used, including surveys, interviews, and qualitative analysis of participant feedback. The participants were randomly assigned to either an AI group or the traditional learning group. Findings suggest that the AI-powered personalised learning platform that we used is a promising approach for enhancing self-directed learning and personal and professional growth. However, it is important to note that these are just preliminary findings, and further research is needed to confirm our results and to understand the mechanisms by which a specific use of AI in education may lead to positive effects., This research was funded by the Department of Philosophy “Piero Martinetti” of the University of Milan under the Project “Departments of Excellence 2023-2027” awarded by the Ministry of University and Research (MUR). Marisa Gil's work has been supported by the Spanish Ministry of Education (PID2019-107255GBC22) and the Generalitat de Catalunya (2021-SGR01007)., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Transparent load balancing of MPI programs using OmpSs-2@Cluster and DLB

  • Aguilar Mena, Jimmy
  • Ali, Omar Shaaban Ibrahim
  • López Herrero, Víctor
  • Garcia Casulla, Marta
  • Carpenter, Paul Matthew
  • Ayguadé Parra, Eduard
  • Labarta Mancho, Jesús José
Load imbalance is a long-standing source of inefficiency in high performance computing. The situation has only got worse as applications and systems increase in complexity, e.g., adaptive mesh refinement, DVFS, memory hierarchies, power and thermal management, and manufacturing processes. Load balancing is often implemented in the application, but it obscures application logic and may need extensive code refactoring. This paper presents an automated and transparent dynamic load balancing approach for MPI applications with OmpSs-2 tasks, which relieves applications from this burden. Only local and trivial changes are required to the application. Our approach exploits the ability of OmpSs-2@Cluster to offload tasks for execution on other nodes, and it reallocates compute resources among ranks using the Dynamic Load Balancing~(DLB) library. It employs LeWI to react to fine-grained load imbalances and DROM to address coarse-grained load imbalances by reserving cores on other nodes that can be reclaimed on demand. We use an expander graph to limit the amount of point-to-point communication and state. The results show 46% reduction in time-to-solution for micro-scale solid mechanics on 32 nodes and a 20% reduction beyond DLB for $n$-body on 16 nodes, when one node is running slow. A synthetic benchmark shows that performance is within 10% of optimal for an imbalance of up to 2.0 on 8 nodes. All software is released open source., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement No 955606 (DEEP-SEA) and 754337 (EuroEXA). It is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255GB and Ramon y Cajal fellowship RYC2018-025628-I) and by the Generalitat de Catalunya (2017-SGR-1414)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Fine-grained scheduling for containerized HPC workloads in Kubernetes clusters

  • Liu, Peini
  • Guitart Fernández, Jordi
Containerization technology offers lightweight OS-level virtualization, and enables portability, reproducibility, and flexibility by packing applications with low performance overhead and low effort to maintain and scale them. Moreover, container orchestrators (e.g., Kubernetes) are widely used in the Cloud to manage large clusters running many containerized applications. However, scheduling policies that consider the performance nuances of containerized High Performance Computing (HPC) workloads have not been well-explored yet. This paper conducts fine-grained scheduling policies for containerized HPC workloads in Kubernetes clusters, focusing especially on partitioning each job into a suitable multi-container deployment according to the application profile. We implement our scheduling schemes on different layers of management (application and infrastructure), so that each component has its own focus and algorithms but still collaborates with others. Our results show that our fine-grained scheduling policies outperform baseline and baseline with CPU/memory affinity enabled policies, reducing the overall response time by 35% and 19%, respectively, and also improving the makespan by 34% and 11%, respectively. They also provide better usability and flexibility to specify HPC workloads than other comparable HPC Cloud frameworks, while providing better scheduling efficiency thanks to their multi-layered approach., This work was partially supported by Lenovo as part of Lenovo-BSC collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2017-SGR-1414 and under grant 2020 FI-B 00257., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Towards on-demand I/O forwarding in HPC platforms

  • Bez, Jean Luca
  • Zanon Boito, Francieli
  • Miranda Bueno, Alberto
  • Nou Castell, Ramon
  • Cortés, Toni
  • Navaux, Philippe O.A.
I/O forwarding is an established and widely-adopted technique in HPC to reduce contention and improve I/O performance in the access to shared storage infrastructure. On such machines, this layer is often physically deployed on dedicated nodes, and their connection to the clients is static. Furthermore, the increasingly heterogeneous workloads entering HPC installations stress the I/O stack, requiring tuning and reconfiguration based on the applications' characteristics. Nonetheless, it is not always feasible in a production system to explore the potential benefits of this layer under different configurations without impacting clients. In this paper, we investigate the effects of I/O forwarding on performance by considering the application's I/O access patterns and system characteristics. We aim to explore when forwarding is the best choice for an application, how many I/O nodes it would benefit from, and whether not using forwarding at all might be the correct decision. To gather performance metrics, explore, and understand the impact of forwarding I/O requests of different access patterns, we implemented FORGE, a lightweight I/O forwarding layer in user-space. Using FORGE, we evaluated the optimal forwarding configurations for several access patterns on MareNostrum 4 (Spain) and Santos Dumont (Brazil) supercomputers. Our results demonstrate that shifting the focus from a static system-wide deployment to an on-demand reconfigurable I/O forwarding layer dictated by application demands can improve I/O performance on future machines., This study was financed by the Coordenaçao de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001. It has also received support from the Conselho Nacional de Desenvolvimento Científico e Tecnologico (CNPq), Brazil; It is also partially supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under grants PID2019-107255GB; and the Generalitat de Catalunya under contract 2014–SGR–1051., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Distributing deep learning hyperparameter tuning for 3D medical image segmentation

  • Berral García, Josep Lluís
  • Aranda Llorens, Oriol
  • Domínguez Bermúdez, Juan Luis
  • Torres Viñals, Jordi
Most research on novel techniques for 3D Medical Image Segmentation (MIS) is currently done using Deep Learning with GPU accelerators. The principal challenge of such technique is that a single input can easily cope computing resources, and require prohibitive amounts of time to be processed. Distribution of deep learning and scalability over computing devices is an actual need for progressing on such research field. Conventional distribution of neural networks consist in “data parallelism”, where data is scattered over resources (e.g., GPUs) to parallelize the training of the model. However, “experiment parallelism” is also an option, where different training processes (i.e., on a hyper-parameter search) are parallelized across resources. While the first option is much more common on 3D image segmentation, the second provides a pipeline design with less dependence among parallelized processes, allowing overhead reduction and more potential scalability. In this work we present a design for distributed deep learning training pipelines, focusing on multi-node and multi-GPU environments, where the two different distribution approaches are deployed and benchmarked. We take as proof of concept the 3D U-Net architecture, using the MSD Brain Tumor Segmentation dataset, a state-of-art problem in medical image segmentation with high computing and space requirements. Using the BSC MareNostrum supercomputer as benchmarking environment, we use TensorFlow and Ray as neural network training and experiment distribution platforms. We evaluate the experiment speed-up when parallelizing, showing the potential for scaling out on GPUs and nodes. Also comparing the different parallelism techniques, showing how experiment distribution leverages better such resources through scaling, e.g. by a speed-up factor from x12 to x14 using 32 GPUs. Finally, we provide the implementation of the design open to the community, and the non-trivial steps and methodology for adapting and deploying a MIS case as the here presented., This work has been partially financed by the European Commission (EU-H2020 INCISIVE GA.952179, and CALLISTO GA.101004152). Also the Spanish Ministry of Science (PID2019- 107255GB-C22/ AEI / 10.13039/501100011033), and Generalitat de Catalunya through the 2017-SGR-1414 project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Deploying vision foundation AI models on the edge. The SAM2 experience

  • Lin, Zheshuo
  • Tous Liesa, Rubén
  • Otero Calviño, Beatriz
As AI-driven applications expand across industries, the need for efficient edge computing solutions becomes increasingly critical. Traditional AI models are designed for high-performance cloud infrastructures, but emerging constraints—such as privacy requirements, network limitations, and real-time processing needs—necessitate optimized deployment on resource-constrained edge devices. This study presents a practical experience in adapting Segment Anything Model 2 (SAM2), a vision foundation model, for edge AI environments. The adaptation process involved translating the model to C++ using ONNX Runtime, enabling efficient execution on heterogeneous hardware. Experimental evaluations demonstrate that deploying SAM2 at the edge enhances processing efficiency, reduces reliance on network stability, and improves real-time responsiveness. This research provides valuable insights into AI in pervasive computing environments, contributing to the sustainable and scalable deployment of foundation models on edge devices., This work is partially funded by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, as well as by the SGR programs 2021-SGR-00478 and 2021-SGR-00326 of the Catalan government. Finally, the research that led to these results has received funding from the Collaboration Grant for the 2024/2025 period from the Spanish Ministry of Education., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Pronóstico de capacidad efectiva y prestaciones en una cache no volátil de último nivel

  • Escuín Blasco, Carlos
  • Monreal Arnal, Teresa
  • Llaberia Griñó, José M.
  • Ibáñez Marín, Pablo Enrique
  • Viñals Yúfera, Victor
La degradación debida a las escrituras que sufren las bitcells implementadas con tecnologi´as de memoria no volátil (NVM) es uno de los principales escollos que se presentan a la hora de construir la cache de último nivel (LLC) con estas tecnologi´as. Aunque en la literatura se recogen diferentes propuestas para hacer frente a esta degradación, la metodologi´a usada en los trabajos previos no permite estudiar en detalle la evolución de la capacidad efectiva ni de las prestaciones en una cache no volátil de último nivel (NV-LLC). Por ello, en este trabajo se propone un procedimiento de pronóstico que combina simulación y predicción con el objetivo de estudiar dicha evolución. Por otra parte, la compresión es una de las técnicas propuestas en la literatura para lidiar con la degradación de las memorias no volátiles. En primer lugar, la compresión diminuye la cantidad de información escrita en una NV-LLC. En segundo lugar, cuando los contenedores ven mermada su capacidad debido a la degradación, la compresión permite mantener su funcionalidad albergando bloques de tamaño reducido. El procedimiento de pronóstico desarrollado en este trabajo permite evaluar el impacto de diferentes técnicas y mecanismos de gestión de contenidos en la esperanza de vida y las prestaciones de una NV-LLC de manera detallada. El mecanismo de compresión adoptado en este trabajo multiplica hasta por 5 veces la esperanza de vida de una NV-LLC., Este trabajo ha sido financiado por MINECO/AEI/FEDER (UE) (proyectos PID2019-105660RB-C21 y PID2019-107255GB-C22 / AEI / 10.13039/501100011033), el Gobierno de Aragon (Grupo T58 20R) y FEDER 2014-2020 “Construyendo Europa desde Aragón”., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Improving the discovering and comparing of three-dimensional protein patterns

  • Valdés Jiménez, Alejandro Mauricio
  • Núñez Vivanco, Gabriel
  • Jiménez González, Daniel
The discovery and comparison of three-dimensional (3D) patterns in protein structures are critical tasks in structural bioinformatics, particularly for identifying functional sites and potential drug targets. In this work, we present significant enhancements of Geomfinder, a ligand-independent algorithm for identifying and comparing 3D protein patterns. Our contribu- tions include the following key improvements: i) a mechanism that reduces the number of unnecessary site comparisons; ii) an optimization that lowers the computational complexity of partial score calculations from O(N×M) to O(N+M); and iii) an updated OpenMP-based parallel version to accelerate the site discovery and comparison processes further. Our results show significant performance improvements for a base dataset of proteins. Results show that the optimized sequential Geomfinder achieves up to 26.3x speedup, compared to the original sequential Geomfinder, for relatively large proteins (reducing from approximately 1 hour to approximately 2 and a half minutes). And our OpenMP Geomfinder proposal, combined with the sequential optimization, reaches a speedup of more than 1000x compared to the sequential original Geomfinder. This novel optimized and parallel Geomfinder version allows for processing large protein structures that were not reasonable to compare before., This research was funded by the Spanish Government (Grants PID2019-107255GB-C22 - UPC-COMPUTACION DE ALTAS PRESTACIONES VIII and CEX2021-001148- S), by Generalitat de Catalunya (2021 SGR 01007), and by Vicerrectoría de Investigación y Postgrado de la Universidad del Bío-Bío (Proyecto Iniciación 2025 IN2537025)., Peer Reviewed, Postprint (author's final draft)