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): 165
Encontrada(s) 2 página(s)

Dipòsit Digital de Documents de la UAB
Publicación Artículo científico (JournalArticle)

BIMSA, accelerating long sequence alignment using processing-in-memory

  • Alonso-Marín, Alejandro
  • Fernandez, Ivan
  • Aguado-Puig, Quim|||0000-0003-4871-3192
  • Gómez-Luna, Juan|||0000-0002-6514-1571
  • Marco-Sola, Santiago|||0000-0001-7951-3914
  • Mutlu, Onur
  • Moreto, Miquel|||0000-0002-9848-8758
Recent advances in sequencing technologies have stressed the critical role of sequence analysis algorithms and tools in genomics and healthcare research. In particular, sequence alignment is a fundamental building block in many sequence analysis pipelines and is frequently a performance bottleneck both in terms of execution time and memory usage. Classical sequence alignment algorithms are based on dynamic programming and often require quadratic time and memory with respect to the sequence length. As a result, classical sequence alignment algorithms fail to scale with increasing sequence lengths and quickly become memory-bound due to data-movement penalties. Processing-In-Memory (PIM) is an emerging architectural paradigm that seeks to accelerate memory-bound algorithms by bringing computation closer to the data to mitigate data-movement penalties. This work presents BIMSA (idirectional n-emory equence lignment), a PIM design and implementation for the state-of-the-art sequence alignment algorithm BiWFA (Bidirectional Wavefront Alignment), incorporating new hardware-aware optimizations for a production-ready PIM architecture (UPMEM). BIMSA supports aligning sequences up to 100K bases, exceeding the limitations of state-of-the-art PIM implementations. First, BIMSA achieves speedups up to 22.24× (11.95× on average) compared to state-of-the-art PIM-enabled implementations of sequence alignment algorithms. Second, achieves speedups up to 5.84× (2.83× on average) compared to the highest-performance multicore CPU implementation of BiWFA. Third, BIMSA exhibits linear scalability with the number of compute units in memory, enabling further performance improvements with upcoming PIM architectures equipped with more compute units and achieving speedups up to 9.56× (4.7× on average).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Cache-aware optimization of matrix multiplication and matrix factorizations on multicore processors

  • Martínez-Pérez, Héctor
  • Catalán, Sandra
  • Igual, Francisco D.
  • Herrero, José R.
  • Rodríguez-Sánchez, Rafael
  • Quintana-Ortí, Enrique S.|||0000-0002-5454-165X
[EN] This paper advocates for a careful customization of the special general matrix multiplication (GEMM) kernels that are invoked from blocked routines for several relevant matrix factorizations in LAPACK, in order to improve their performance on modern multicore processors with hierarchical cache memories. To achieve this, we leverage a refined analytical model to dynamically tune the cache configuration parameters of GEMM for these kernels, taking into account the matrix operands' dimensions, in order to improve cache occupation. In addition, toward the same goal, we accommodate a flexible development of architecture-specific micro-kernels for GEMM that allows us to select the option that, depending on the operands' dimensions, ameliorates cache utilization. Our experiments for the LU and QR factorizations on two platforms, equipped with ARM (NVIDIA Carmel) and x86 (AMD EPYC) multi-core processors, demonstrate the benefits of this approach in terms of a better cache utilization and, in general, higher performance. Moreover, they also reveal the delicate balance between optimizing for multi-threaded parallelism versus cache usage as well as the positive effects of software prefetching., This work was supported by grants PID2020-113656RB-C22, PID2019-107255GB, PID2021-126576NB-I00 and PID2021-123627OB-C52 of MCIN/AEI/10.13039/501100011033, by "ERDF A way of making Europe", and 2021-SGR-01007 of the Generalitat de Catalunya. Hector Martinez is a postdoctoral fellow supported by the Consejeria de Transformacion Economica, Industria, Conocimiento y Universidades de la Junta de Andalucia. Sandra Catalan was supported by the grant RYC2021-033973- I, funded by MCIN/AEI/10.13039/501100011033 and the European Union "NextGenerationEU"/PRTR.
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.
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Enabling dynamic and intelligent workflows for HPC, data analytics, and AI convergence

  • Ejarque, Jorge
  • Badia, Rosa M.
  • Albertin, Loïc
  • Aloisio, Giovanni
  • Baglione, Enrico
  • Becerra, Yolanda
  • Boschert, Stefan
  • Berlin, Julian R.
  • D’Anca, Alessandro
  • Elia, Donatello
  • Exertier, François
  • Fiore, Sandro
  • Flich, José
  • Folch, Arnau
  • Gibbons, Steven J.
  • Koldunov, Nikolay
  • Lordan, Francesc
  • Lorito, Stefano
  • Løvholt, Finn
  • Macías, Jorge
  • Volpe, Manuela
The evolution of High-Performance Computing (HPC) platforms enables the design and execution of progressively larger and more complex workflow applications in these systems. The complexity comes not only from the number of elements that compose the workflows but also from the type of computations they perform. While traditional HPC workflows target simulations and modelling of physical phenomena, current needs require in addition data analytics (DA) and artificial intelligence (AI) tasks. However, the development of these workflows is hampered by the lack of proper programming models and environments that support the integration of HPC, DA, and AI, as well as the lack of tools to easily deploy and execute the workflows in HPC systems. To progress in this direction, this paper presents use cases where complex workflows are required and investigates the main issues to be addressed for the HPC/DA/AI convergence. Based on this study, the paper identifies the challenges of a new workflow platform to manage complex workflows. Finally, it proposes a development approach for such a workflow platform addressing these challenges in two directions: first, by defining a software stack that provides the functionalities to manage these complex workflows; and second, by proposing the HPC Workflow as a Service (HPCWaaS) paradigm, which leverages the software stack to facilitate the reusability of complex workflows in federated HPC infrastructures. Proposals presented in this work are subject to study and development as part of the EuroHPC eFlows4HPC project., This work has received funding from the European High-Performance Computing Joint Undertaking (JU) under grant agreement No 955558. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and Spain, Germany, France, Italy, Poland, Switzerland and Norway. In Spain, it has received complementary funding from MCIN/AEI/10.13039/501100011033, Spain and the European Union NextGenerationEU/PRTR (contracts PCI2021-121957, PCI2021-121931, PCI2021-121944, and PCI2021-121927). In Germany, it has received complementary funding from the German Federal Ministry of Education and Research (contracts 16HPC016K, 6GPC016K, 16HPC017 and 16HPC018). In France, it has received financial support from Caisse des dépôts et consignations (CDC) under the action PIA ADEIP (project Calculateurs). In Italy, it has been preliminary approved for complimentary funding by Ministero dello Sviluppo Economico (MiSE) (ref. project prop. 2659). In Norway, it has received complementary funding from the Norwegian Research Council, Norway under project number 323825. In Switzerland, it has been preliminary approved for complimentary funding by the State Secretariat for Education, Research, and Innovation (SERI), Norway. In Poland, it is partially supported by the National Centre for Research and Development under decision DWM/EuroHPCJU/4/2021. The authors also acknowledge financial support by MCIN/AEI /10.13039/501100011033, Spain through the “Severo Ochoa Programme for Centres of Excellence in R&D” under Grant CEX2018-000797-S, the Spanish Government, Spain (contract PID2019-107255 GB) and by Generalitat de Catalunya, Spain (contract 2017-SGR-01414). Anna Queralt is a Serra Húnter Fellow., With funding from the Spanish government through the ‘Severo Ochoa Centre of Excellence’ accreditation (CEX2018-000797-S).
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

A closer look at referring expressions for video object segmentation

  • Bellver, Miriam
  • Ventura, Carles
  • Silberer, Carina
  • Kazakos, Ioannis
  • Torres, Jordi
  • Giró i Nieto, Xavier
The task of Language-guided Video Object Segmentation (LVOS) aims at generating binary masks for an object referred by a linguistic expression. When this expression unambiguously describes an object in the scene, it is named referring expression (RE). Our work argues that existing benchmarks used for LVOS are mainly composed of trivial cases, in which referents can be identified with simple phrases. Our analysis relies on a new categorization of the referring expressions in the DAVIS-2017 and Actor-Action datasets into trivial and non-trivial REs, where the non-trivial REs are further annotated with seven RE semantic categories. We leverage these data to analyze the performance of RefVOS, a novel neural network that obtains competitive results for the task of language-guided image segmentation and state of the art results for LVOS. Our study indicates that the major challenges for the task are related to understanding motion and static actions., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work was partially supported by the projects PID2019-107255GB-C22 and PID2020-117142GB-I00 funded by MCIN/ AEI /10.13039/501100011033 Spanish Ministry of Science, and the grant 2017-SGR-1414 of the Government of Catalonia. This work was also partially supported by the project RTI2018-095232-B-C22 funded by the Spanish Ministry of Science, Innovation and Universities.
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 Artículo científico (JournalArticle)

OmpSs@FPGA framework for high performance FPGA computing

  • De Haro Ruiz, Juan Miguel|||0000-0002-7427-9118
  • Bosch Pons, Jaume|||0000-0002-4040-3416
  • 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
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
This paper presents the new features of the OmpSs@FPGA framework. OmpSs is a data-flow programming model that supports task nesting and dependencies to target asynchronous parallelism and heterogeneity. OmpSs@FPGA is the extension of the programming model addressed specifically to FPGAs. OmpSs environment is built on top of Mercurium source to source compiler and Nanos++ runtime system. To address FPGA specifics Mercurium compiler implements several FPGA related features as local variable caching, wide memory accesses or accelerator replication. In addition, part of the Nanos++ runtime has been ported to hardware. Driven by the compiler this new hardware runtime adds new features to FPGA codes, such as task creation and dependence management, providing both performance increases and ease of programming. To demonstrate these new capabilities, different high performance benchmarks have been evaluated over different FPGA platforms using the OmpSs programming model. The results demonstrate that programs that use the OmpSs programming model achieve very competitive performance with low to moderate porting effort compared to other FPGA implementations., This work has received funding from EuroEXA project (European Union’s Horizon 2020 Research and Innovation Programme, under grant agreement No 754337), from Spanish Government (projects PID2019-107255GB and SEV-2015- 0493, grant BES-2016-078046), and 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)

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 Artículo científico (JournalArticle)

Performance characterization of containerization for HPC workloads on InfiniBand clusters: an empirical study

  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
Containerization technology offers an appealing alternative for encapsulating and operating applications (and all their dependencies) without being constrained by the performance penalties of using Virtual Machines and, as a result, has got the interest of the High-Performance Computing (HPC) community to obtain fast, customized, portable, flexible, and reproducible deployments of their workloads. Previous work on this area has demonstrated that containerized HPC applications can exploit InfiniBand networks, but has ignored the potential of multi-container deployments which partition the processes that belong to each application into multiple containers in each host. Partitioning HPC applications has demonstrated to be useful when using virtual machines by constraining them to a single NUMA (Non-Uniform Memory Access) domain. This paper conducts a systematical study on the performance of multi-container deployments with different network fabrics and protocols, focusing especially on Infiniband networks. We analyze the impact of container granularity and its potential to exploit processor and memory affinity to improve applications’ performance. Our results show that default Singularity can achieve near bare-metal performance but does not support fine-grain multi-container deployments. Docker and Singularity-instance have similar behavior in terms of the performance of deployment schemes with different container granularity and affinity. This behavior differs for the several network fabrics and protocols, and depends as well on the application communication patterns and the message size. Moreover, deployments on Infiniband are also more impacted by the computation and memory allocation, and because of that, they can exploit the affinity better., We thank Lenovo for providing the testbed to run the experiments in this paper. 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 No. 2020 FI-B 00257., 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 Artículo científico (JournalArticle)

NagareDB: A resource-efficient document-oriented time-series database

  • García Calatrava, Carlos
  • Becerra Fontal, Yolanda|||0000-0003-2357-7796
  • Cucchietti, Fernando|||0000-0002-9027-1263
  • Diví Cuesta, Carla
The recent great technological advance has led to a broad proliferation of Monitoring Infrastructures, which typically keep track of specific assets along time, ranging from factory machinery, device location, or even people. Gathering this data has become crucial for a wide number of applications, like exploration dashboards or Machine Learning techniques, such as Anomaly Detection. Time-Series Databases, designed to handle these data, grew in popularity, becoming the fastest-growing database type from 2019. In consequence, keeping track and mastering those rapidly evolving technologies became increasingly difficult. This paper introduces the holistic design approach followed for building NagareDB, a Time-Series database built on top of MongoDB—the most popular NoSQL Database, typically discouraged in the Time-Series scenario. The goal of NagareDB is to ease the access to three of the essential resources needed to building time-dependent systems: Hardware, since it is able to work in commodity machines; Software, as it is built on top of an open-source solution; and Expert Personnel, as its foundation database is considered the most popular NoSQL DB, lowering its learning curve. Concretely, NagareDB is able to outperform MongoDB recommended implementation up to 4.7 times, when retrieving data, while also offering a stream-ingestion up to 35% faster than InfluxDB, the most popular Time-Series database. Moreover, by relaxing some requirements, NagareDB is able to reduce the disk space usage up to 40%., This research was partly supported by the H2020 IoTwins project (Distributed Digital Twins for industrial SMEs: a big-data platform) funded by the EU under the call ICT-11-2018-2019, Grant Agreement No 857191, by the Spanish Ministry of Science and Innovation (contract PID2019- 107255GB) and by the Generalitat de Catalunya (contract 2017-SGR-1414)., 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 Artículo científico (JournalArticle)

Compute units in OpenMP: extensions for heterogeneous parallel programming

  • González Tallada, Marc|||0000-0002-3780-1106
  • Morancho Llena, Enrique|||0000-0003-2403-8145
This article evaluates the current support for heterogeneous OpenMP 5.2 applications regarding the simultaneous activation of host and device computing units (e.g., CPUs, GPUs, or FPGAs). The article identifies limitations in the current OpenMP specification and describes the design and implementation of novel OpenMP extensions and runtime support for heterogeneous parallel programming. The Compute Unit (CUs) abstraction is introduced in the OpenMP programming model. The Compute Unit abstraction is defined in terms of an aggregation of computing elements (e.g., CPUs, GPUs, FPGAs). On top of CUs, the article describes dynamic work sharing constructs and schedulers that address the inherent differences in compute power of host and device CUs. New constructs and the corresponding runtime support are described for the new abstractions. The article evaluates the case of a hybrid multilevel parallelization of the NPB-MZ benchmark suite. The implementation exploits both coarse-grain and fine-grain parallelism, mapped to CUs of different nature (GPUs and CPUs). All CUs are activated using the new extensions and runtime support. We compare hybrid and nonhybrid executions under two state-of-the-art work-distribution schemes (Static and Dynamic Task schedulers). On a computing node composed of one AMD EPYC 7742 @ 2.250GHz (64 cores and 2 threads/core, totalling 128 threads per node) and 2 GPU AMD Radeon Instinct MI50 with 32GB, hybrid executions present speedups from 1.08 up to 3.18 with respect to a nonhybrid GPU implementation, depending on the number of activated CUs., This work was supported by the Spanish Ministry of Science and Technology (PID2019-107255GB)., 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 Artículo científico (JournalArticle)

A closer look at referring expressions for video object segmentation

  • Bellver Bueno, Míriam
  • Ventura Royo, Carles|||0000-0002-3055-339X
  • Silberer, Carina
  • Kazakos, Ioannis
  • Torres Viñals, Jordi|||0000-0003-1963-7418
  • Giró Nieto, Xavier|||0000-0002-9935-5332
The task of Language-guided Video Object Segmentation (LVOS) aims at generating binary masks for an object referred by a linguistic expression. When this expression unambiguously describes an object in the scene, it is named referring expression (RE). Our work argues that existing benchmarks used for LVOS are mainly composed of trivial cases, in which referents can be identified with simple phrases. Our analysis relies on a new categorization of the referring expressions in the DAVIS-2017 and Actor-Action datasets into trivial and non-trivial REs, where the non-trivial REs are further annotated with seven RE semantic categories. We leverage these data to analyze the performance of RefVOS, a novel neural network that obtains competitive results for the task of language-guided image segmentation and state of the art results for LVOS. Our study indicates that the major challenges for the task are related to understanding motion and static actions., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work was partially supported by the projects PID2019-107255GB-C22 and PID2020-117142GB-I00 funded by MCIN/ AEI /10.13039/501100011033 Spanish Ministry of Science, and the grant 2017-SGR-1414 of the Government of Catalonia. This work was also partially supported by the project RTI2018-095232-B-C22 funded by the Spanish Ministry of Science, Innovation and Universities., 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 Artículo científico (JournalArticle)

Boosting HPC data analysis performance with the ParSoDA-Py library

  • Belcastro, Loris
  • Giampà, Salvatore
  • Marozzo, Fabrizio
  • Talia, Domenico
  • Trunfio, Paolo
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Ejarque, Jorge|||0000-0003-4725-5097
  • Mammadli, Nihad
Developing and executing large-scale data analysis applications in parallel and distributed environments can be a complex and time-consuming task. Developers often find themselves diverted from their application logic to handle technical details about the underlying runtime and related issues. To simplify this process, ParSoDA, a Java library, has been proposed to facilitate the development of parallel data mining applications executed on HPC systems. It simplifies the process by providing built-in scalability mechanisms relying on the Hadoop and Spark frameworks. This paper presents ParSoDA-Py, the Python version of the ParSoDA library, which allows for further support of commonly used runtimes and libraries for big data analysis. After a complete library redesign, ParSoDA can be now easily integrated with other Python-based distributed runtimes for HPC systems, such as COMPSs and Apache Spark, and with the large ecosystem of Python-based data processing libraries. The paper discusses the adaptation process, which takes into consideration the new technical requirements, and evaluates both usability and scalability through some case study applications., This work has been partially funded by the European Commission’s Horizon 2020 Framework program and the European High-Performance Computing Joint Undertaking (JU) under Grant agreement No 955558 and by MCIN/AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR (PCI2021-121957), project eFlows4HPC. It has also been supported by the Spanish Government (PID2019-107255GB) and by the Departament de Recerca i Universitats de la Generalitat de Catalunya to the Research Group MPiEDist (2021 SGR 00412) We also acknowledge financial support from “National Centre for HPC, Big Data and Quantum Computing," CN00000013 - CUP H23C22000360005, and from “FAIR - Future Artificial Intelligence Research" Project - CUP H23C22000860006., 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 Artículo científico (JournalArticle)

Low-cost training of image-to-image diffusion models with incremental learning and task/domain adaptation

  • Antona Pizà, Héctor
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Tous Liesa, Rubén|||0000-0002-1409-5843
Diffusion models specialized in image-to-image translation tasks, like inpainting and colorization, have outperformed the state of the art, yet their computational requirements are exceptionally demanding. This study analyzes different strategies to train image-to-image diffusion models in a low-resource setting. The studied strategies include incremental learning and task/domain transfer learning. First, a base model for human face inpainting is trained from scratch with an incremental learning strategy. The resulting model achieves an FID score almost equivalent to that of its batch learning equivalent while significantly reducing the training time. Second, the base model is fine-tuned to perform a different task, image colorization, and, in a different domain, landscape images. The resulting colorization models showcase exceptional performances with a minimal number of training epochs. We examine the impact of different configurations and provide insights into the ability of image-to-image diffusion models for transfer learning across tasks and domains., This work is partially supported by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, by the Generalitat de Catalunya under grants 2017-SGR-962, 2021-SGR-00326, 2021-SGR-00478 and DRAC (IU16-011591). The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation program under the HORIZON-AG PHOENI2X (101070586) and HORIZON-EU VITAMIN-V (101093062) projects., 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 Artículo científico (JournalArticle)

Near-optimal replacement policies for shared caches in multicore processors

  • Díaz Maag, Javier
  • Ibáñez Marín, Pablo
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Viñals Yúfera, Victor
  • Llaberia Griñó, José M.|||0000-0002-3753-4108
An optimal replacement policy that minimizes the miss rate in a private cache was proposed several decades ago. It requires knowing the future access sequence the cache will receive. There is no equivalent for shared caches because replacement decisions alter this future sequence. We present a novel near-optimal policy for minimizing the miss rate in a shared cache that approaches the optimal execution iteratively. During each iteration, the future access sequence is reconstructed on every miss interleaving the future per-core sequences, taken from the previous iteration. This single sequence feeds a classical private-cache optimum replacement policy. Our evaluation on a shared last-level cache shows that our proposal iteratively converges to a near-optimal miss rate that is independent of the initial conditions, within a margin of 0.1%. The best state-of-the-art online policies achieve around 65% of the miss rate reduction obtained by our near-optimal proposal. In a shared cache, miss rate optimization does not imply the optimization of other metrics. Therefore, we also propose a new near-optimal policy to maximize fairness between cores. The best state-of-the-art online policy achieves 60% of the improvement in fairness seen with our near-optimal policy. Our proposals are useful both for setting upper performance bounds and inspiring implementable mechanisms for shared caches., The authors acknowledge support from grants (1) PID2019-105660RB-C21 and PID2019-107255GB-C22 from Agencia Estatal de Investigación (AEI) from Spain and European Regional Development Fund (ERDF); (2) gaZ: T58_20R research group from Aragón Government and European Social Fund (ESF); and (3) 2014-2020 "Construyendo Europa desde Aragón" from European Regional Development Fund (ERDF)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Embedded scaffolding for teaching and assessing inquiry-based hands-on laboratory on distributed systems

  • Guitart Fernández, Jordi|||0000-0003-0751-3100
Context Information Technology education must cultivate proficiency on distributed systems, including strong hands-on laboratory skills, to meet the needs of the society and the industry. Given the complexity of distributed systems, any successful methodology to teach them to novice students must be scaffolded appropriately to ensure that the students acquire the required degree of expertise. Objective We propose a comprehensive scaffolding approach for inquiry-based hands-on laboratory on a distributed systems course, which guides not only the learning process, but also its assessment. The approach is based mainly on embedded scaffolds, namely explicit coding and experimental milestones and open questions with predefined grades, but also features contingent scaffolds provided by the teacher when additional assistance is needed. Method We apply the methodology in the context of the subject ‘Distributed Network Systems’ offered by our university. We compare the students' performance during three academic courses using the proposed methodology with respect to the three previous courses that were still using the former methodology. We use both visual representations and planned Analysis of Variance (ANOVA) tests to verify our hypothesis defined as a complex contrast. Findings We find that there is a statistically significant improvement in the students' performance when using the new methodology, both in their grades of the assignments (F(1, 75.364) = 17.770, p = 6.85*10^-5) and, more importantly, also in their grades of the exam questions about the practicals (F(1, 123.186) = 13.285, p = 3.93*10^-4). Implications Our results encourage other instructors to incorporate embedded scaffolds for teaching and assessing their hands-on laboratories on distributed systems., This research was partially supported by the Agencia Estatal de Investigación under contracts PID2019-107255GB-C22 and PID2021-126248OB-I00 and by the Generalitat de Catalunya under contract 2021-SGR-00478., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Optimizing MPI collective operations for intra-node shared memory exploitation and oversubscription

  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Bull, J. Mark
Node sizes in multicore clusters are becoming larger, so applications should exploit the shared memory inside a node, to potentially reduce communication latencies compared to network communications. The Message Passing Interface library (MPI) serves as the de facto standard for parallel applications on distributed memory environments. The appearance of the MPI-3 shared memory extension has made the idea of optimizing the MPI collective operations at the intra-node level both attractive and portable. Taking advantage of this facility, we present a hierarchical design of the algorithms for MPI_Allreduce and MPI_Bcast collective operations, which we name Fullpar. The proposal is based on partitioning the messages and exploiting concurrency between network communication and shared-memory operations at intra-node level in the message dissemination process. Previously proposed hierarchical collective algorithms do not exploit all the available parallelism and/or the direct implementation of shared-memory at intra-node level. Furthermore, we explore the application of oversubscription, to enhance resource utilization. Oversubscribing CPUs offers opportunities to optimize resource allocation, as well as making CPUs available to other applications. We implement our proposal on top of the Intel MPI and OpenMPI libraries using the MPI profiling (PMPI) mechanism, and carry out evaluations on platforms with different architectures characteristics such as the node size and interconnection network. Experimental results show that the proposed approach often achieves lower execution times than the evaluated alternatives for medium and large message sizes. Furthermore, the introduction of oversubscription is shown to have almost no overhead, especially for the broadcast of large messages., This work was supported by the Spanish Ministry of Education (PID2019-107255GB-C22); the Generalitat de Catalunya (2021-SGR-01007); and Project HPC-EUROPA3 (INFRAIA-2016-1-730897), with the support of the EC Research Innovation Action under the H2020 Programme., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Human pose completion in partial body camera shots

  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Nin Guerrero, Jordi|||0000-0002-9659-2762
  • Igual Muñoz, Laura
Many actual images contain partial body camera shots, in which a significant part of the body is not visible. This issue is especially prevalent in film images, where less than 10% are full-body shots. Most 2D human pose estimation methods return incomplete poses when applied to partial body images. This lack of completeness becomes a problem in some situations, for example, when the 2D pose is converted to a 3D pose by a two-stage 3D human pose estimation method since most of these methods require a complete pose to work. This article proposes a new technique, called CompletePose, consisting of completing the missing keypoints when 2D human pose estimation methods are applied to images with partial body camera shots. A Conditional Generative Adversarial Network is used to obtain a complete and plausible pose, realistic enough to predict a 3D pose with a two-stage 3D human pose estimation method. A complete empirical validation has been carried out with the Human3.6 M dataset and a new dataset, called CHARADE, specially built and made public for reproducibility and benchmarking for this research. Quantitative evaluation employing the Fréchet distance shows that the approach manages to approximate the actual data distribution. The qualitative evaluation shows that the completed poses enable obtaining plausible 3D poses from images previously intractable., This work is partially supported by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB, PLEC2021-007850 and RTI2018-095232-B-C21, and by the SGR programme of the Catalan Government under contracts 2021-SGR00478, 2017-SGR-1414 and 2017-SGR-1742., 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 Artículo científico (JournalArticle)

Programming parallel dense matrix factorizations and inversion for new-generation NUMA architectures

  • Catalán Pallarés, Sandra
  • Igual Peña, Francisco D.
  • Herrero Zaragoza, José Ramón|||0000-0002-4060-367X
  • Rodríguez Sánchez, Rafael
  • Quintana Ortí, Enrique Salvador
We propose a methodology to address the programmability issues derived from the emergence of new-generation shared-memory NUMA architectures. For this purpose, we employ dense matrix factorizations and matrix inversion (DMFI) as a use case, and we target two modern architectures (AMD Rome and Huawei Kunpeng 920) that exhibit configurable NUMA topologies. Our methodology pursues performance portability across different NUMA configurations by proposing multi-domain implementations for DMFI plus a hybrid task- and loop-level parallelization that configures multi-threaded executions to fix core-to-data binding, exploiting locality at the expense of minor code modifications. In addition, we introduce a generalization of the multi-domain implementations for DMFI that offers support for virtually any NUMA topology in present and future architectures. Our experimentation on the two target architectures for three representative dense linear algebra operations validates the proposal, reveals insights on the necessity of adapting both the codes and their execution to improve data access locality, and reports performance across architectures and inter- and intra-socket NUMA configurations competitive with state-of-the-art message-passing implementations, maintaining the ease of development usually associated with shared-memory programming., This research was sponsored by project PID2019-107255GB of Ministerio de Ciencia, Innovación y Universidades; project S2018/TCS-4423 of Comunidad de Madrid; project 2017-SGR-1414 of the Generalitat de Catalunya and the Madrid Government under the Multiannual Agreement with UCM in the line Program to Stimulate Research for Young Doctors in the context of the V PRICIT, project PR65/19-22445. This project has also received funding from the European High-Performance Computing Joint Undertaking (JU) under grant agreement No 955558. The JU receives support from the European Union’s Horizon 2020 research and innovation programme, and Spain, Germany, France, Italy, Poland, Switzerland, Norway. The work is also supported by grants PID2020-113656RB-C22 and PID2021-126576NB-I00 of MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Mixed-precision pre-pivoting strategy for the LU factorization

  • Sahraneshinsamani, Nima
  • Catalán Pallarés, Sandra
  • Herrero Zaragoza, José Ramón|||0000-0002-4060-367X
This paper investigates the efficient application of half-precision floating-point (FP16) arithmetic on GPUs for boosting LU decompositions in double (FP64) precision. Addressing the motivation to enhance computational efficiency, we introduce two novel algorithms: Pre-Pivoted LU (PRP) and Mixed-precision Panel Factorization (MPF). Deployed in both hybrid CPU-GPU setups and native GPU-only configurations, PRP identifies pivot lists through LU decomposition computed in reduced precision and subsequently reorders matrix rows in FP64 precision before executing LU decomposition without pivoting. Two variants of PRP, namely hPRP and xPRP, are introduced, differing in their computation of pivot lists in full half-precision or mixed half-single precision. The MPF algorithm generates FP64 LU factorization while internally utilizing hPRP for panel factorization, showcasing accuracy on par with standard DGETRF but with superior speed. The study further explores auxiliary functions required for the native mode implementation of PRP variants and MPF., The authors acknowledge the funding received from project PID2019-107255GB of the Spanish Ministerio de Economía Competitividad, and project 2021-SGR-01007 from the Generalitat de Catalunya, Spain. Sandra Catalán was supported by the grant RYC2021-033973-I, funded by MCIN/AEI/10.13039/501100011033 and the European Union “NextGenerationEU”/PRTR, and UJI-2023-04, funded by Universitat Jaume I., Peer Reviewed
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 Artículo científico (JournalArticle)

Evaluation of work distribution schedulers for heterogeneous architectures and scientific applications

  • González Tallada, Marc|||0000-0002-3780-1106
  • Morancho Llena, Enrique|||0000-0003-2403-8145
This article explores and evaluates variants of state-of-the-art work distribution schemes adapted for scientific applications running on hybrid systems. A hybrid implementation (multi-GPU and multi-CPU) of the NASA Parallel Benchmarks - MutiZone (NPB-MZ) benchmarks is described to study the different elements that condition the execution of this suite of applications when parallelism is spread over a set of computing units (CUs) of different computational power (e.g., GPUs and CPUs). This article studies the influence of the work distribution schemes on the data placement across the devices and the host, which in turn determine the communications between the CUs, and evaluates how the schedulers are affected by the relationship between data placement and communications. We show that only the schedulers aware of the different computational power of the CUs and minimize communications are able to achieve an appropriate work balance and high performance levels. Only then does the combination of GPUs and CPUs result in an effective parallel implementation that boosts the performance of a non-hybrid multi-GPU implementation. The article describes and evaluates the schedulers static-pcf, Guided, and Clustered Guided to solve the previously mentioned limitations that appear in hybrid systems. We compare them against state-of-the-art static and memorizing dynamic schedulers. Finally, on a system with an AMD EPYC 7742 at 2.250GHz (64 cores, 2 threads per core, 128 threads) and two AMD Radeon Instinct MI50 GPUs with 32GB, we have observed that hybrid executions speed up from 1.1× up to 3.5× with respect to a non-hybrid GPU implementation., This work was supported by the Spanish Ministry of Science and Technology under the grant no. PID2019-107255GB., 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 Artículo científico (JournalArticle)

HRM: merging hardware event monitors for improved timing analysis of complex MPSoCs

  • Vilardell Moreno, Sergi
  • Serra Mochales, Isabel
  • Santalla, Roberto
  • Mezzetti, Enrico|||0000-0002-1886-2931
  • Abella Ferrer, Jaume|||0000-0001-7951-4028
  • Cazorla Almeida, Francisco Javier
The Performance Monitoring Unit (PMU) in MPSoCs is at the heart of the latest measurement-based timing analysis techniques in Critical Embedded Systems. In particular, hardware event monitors (HEMs) in the PMU are used as building blocks in the process of budgeting and verifying software timing by tracking and controlling access counts to shared resources. While the number of HEMs in current MPSoCs reaches hundreds, they are read via Performance Monitoring Counters whose number is usually limited to 4-8, thus requiring multiple runs of each experiment in order to collect all desired HEMs. Despite the effort of engineers in controlling the execution conditions of each experiment, the complexity of current MPSoCs makes it arguably impossible to completely remove the noise affecting each run. As a result, HEMs read in different runs are subject to different variability, and hence, those HEMs captured in different runs cannot be ‘blindly’ merged. In this work, we focus on the NXP T2080 platform where we observed up to 59% variability across different runs of the same experiment for some relevant HEMs (e.g. processor cycles). We develop a HEM reading and merging (HRM) approach to join reliably HEMs across different runs as a fundamental element of any measurement-based timing budgeting and verification technique. Our method builds on order statistics and the selection of an anchor HEM read in all runs to derive the most plausible combination of HEM readings that keep the distribution of each HEM and their relationship with the anchor HEM intact., 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)

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 Artículo científico (JournalArticle)

Performance comparison of multi-container deployment schemes for HPC workloads: an empirical study

  • Liu, Peini|||0000-0003-0058-8732
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
The high-performance computing (HPC) community has recently started to use containerization to obtain fast, customized, portable, flexible, and reproducible deployments of their workloads. Previous work showed that deploying an HPC workload into a single container can keep bare-metal performance. However, there is a lack of research on multi-container deployments that partition the processes belonging to each application into different containers. Partitioning HPC applications has shown to improve their performance on virtual machines by allowing to set affinity to a non-uniform memory access (NUMA) domain for each of them. Consequently, it is essential to understand the performance implications of distinct multi-container deployment schemes for HPC workloads, focusing on the impact of the container granularity and its combination with processor and memory affinity. This paper presents a systematic performance comparison and analysis of multi-container deployment schemes for HPC workloads on a single-node platform, which considers different containerization technologies (including Docker and Singularity), two different platform architectures (UMA and NUMA), and two application subscription modes (exact subscription and over-subscription). Our results indicate that finer-grained multi-container deployments, on the one side, can benefit the performance of some applications with low interprocess communication, especially in over-subscribed scenarios and when combined with affinity, but, on the other side, they can incur some performance degradation for communication-intensive applications when using containerization technologies that deploy isolated network namespaces., 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 Artículo científico (JournalArticle)

Automatic distributed deep learning using resource-constrained edge devices

  • Gutiérrez Torre, Alberto|||0000-0002-5548-3359
  • Bahadori, Kiyana
  • Baig, Shuja-ur-Rehman
  • Iqbal, Waheed
  • Vardanega, Tullio
  • Berral García, Josep Lluís|||0000-0003-3037-3580
  • Carrera Pérez, David|||0000-0003-4898-3424
Processing data generated at high volume and speed from the Internet of Things, smart cities, domotic, intelligent surveillance, and e-healthcare systems require efficient data processing and analytics services at the Edge to reduce the latency and response time of the applications. The Fog Computing Edge infrastructure consists of devices with limited computing, memory, and bandwidth resources, which challenge the construction of predictive analytics solutions that require resource-intensive tasks for training machine learning models. In this work, we focus on the development of predictive analytics for urban traffic. Our solution is based on deep learning techniques localized in the Edge, where computing devices have very limited computational resources. We present an innovative method for efficiently training of Gated Recurrent-Units (GRUs) across available resource-constrained CPU and GPU Edge devices. Our solution employs distributed GRU model learning and dynamically stops the training process to utilize the low-power and resource-constrained Edge devices while ensuring good estimation accuracy effectively. The proposed solution was extensively evaluated using low-powered ARM-based devices, including Raspberry Pi v3 and the low-powered GPU-enabled device NVIDIA Jetson Nano, and also compared them with Single-CPU Intel Xeon machines. For the evaluation experiments, we used real-world Floating Car Data. The experiments show that the proposed solution delivers excellent prediction accuracy and computational performance on the Edge when compared with the baseline methods., This work has been partially supported by the Spanish Government (contract PID2019-107255GB), the Generalitat de Catalunya (contract 2014-SGR-1051), the University of Padova, and Severo Ochoa CoE (SEV-2015-0493-16-5). The statements made herein are solely the responsibility of the authors., 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 Artículo científico (JournalArticle)

Pictonaut: movie cartoonization using 3D human pose estimation and GANs

  • Tous Liesa, Rubén|||0000-0002-1409-5843
This article describes Pictonaut, a novel method to automatically synthetise animated shots from motion picture footage. Its results are editable (backgrounds, characters, lighting, etc.) with conventional 3D software, and they have the finish of professional 2D animation. Rather than addressing the challenge solely as an image translation problem, a hybrid approach combining multi-person 3D human pose estimation and GANs is taken. Sub-sampled video frames are processed with OpenPose and SMPLify-X to obtain the 3D parameters of the pose (body, hands and face expression) of all depicted characters. The captured parameters are retargeted into manually selected 3D models, cel shaded to mimic the style of a 2D cartoon. The results of sub-sampled frames are interpolated to generate a complete and smooth motion for all the characters. The background is cartoonized with a GAN. Qualitative evaluation shows that the approach is feasible, and a small dataset of synthetised shots obtained from real movie scenes is provided., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB, and by the SGR programme 2017-SGR-1414 of the Catalan Government., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Adaptive power shifting for power-constrained heterogeneous systems

  • Ortega Carrasco, Cristobal
  • Alvarez Martí, Lluc
  • Buyuktosunoglu, Alper
  • Bertran Monfort, Ramon
  • Rosedahl, Todd Jon
  • Bose, Pradip
  • Moretó Planas, Miquel|||0000-0002-9848-8758
The number and heterogeneity of compute devices, even within a single compute node, has been steadily on the rise. Since all systems must operate under a power cap, the number of discrete devices that can run simultaneously at their highest frequency is limited by the globally-imposed power cap. Current systems incorporate a centralized power management unit that statically controls the distribution of power among the devices within the node. However, such static distribution policies are unaware of the dynamic utilization profile across the devices, which leads to unfair power allocations that end up degrading system throughput performance. The problem is particularly acute in the presence of heterogeneity since type-specific performance-boost capabilities cannot be leveraged via utilization-agnostic static power allocations. This paper proposes Adaptive Power Shifting for multi-accelerator heterogeneous systems (APS), a technique that leverages system utilization information to dynamically allocate and re-distribute power budgets across multiple discrete devices. Democratizing the power allocation based on dynamic needs results in dramatic speedup over a need-agnostic static allocation. We use APS in a real OpenPOWER compute node with 2 CPUs and 4 GPUs to demonstrate the value of on-demand, equitable power allocations. Overall, the proposed solution increases performance with respect to two state-of-the-art techniques by up to 14.9% and 13.8%., This work has been partially supported by the European Union’s Horizon 2020 research and innovation program under the Mont-Blanc 2020 project (grant agreement 779877), by the Spanish Ministry of Science and Innovation (contract PID2019-107255GB-C22), by Generalitat de Catalunya (contracts 2014-SGR-1051 and 2014-SGR-1272) and by the IBM/BSC Deep Learning Center initiative. Ll. Alvarez has been supported in part 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 supported in part 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)

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 Artículo científico (JournalArticle)

Temporally coherent video cartoonization for animation scenery generation

  • Rayo Hernández, Gustavo Enrique
  • Tous Liesa, Rubén|||0000-0002-1409-5843
The automatic transformation of short background videos from real scenarios into other forms with a visually pleasing style, like those used in cartoons, holds application in various domains. These include animated films, video games, advertisements, and many other areas that involve visual content creation. A method or tool that can perform this task would inspire, facilitate, and streamline the work of artists and people who produce this type of content. This work proposes a method that integrates multiple components to translate short background videos into other forms that contain a particular style. We apply a fine-tuned latent diffusion model with an image-to-image setting, conditioned with the image edges (computed with holistically nested edge detection) and CLIP-generated prompts to translate the keyframes from a source video, ensuring content preservation. To maintain temporal coherence, the keyframes are translated into grids and the style is interpolated with an example-based style propagation algorithm. We quantitatively assess the content preservation and temporal coherence using CLIP-based metrics over a new dataset of 20 videos translated into three distinct styles., This work is partially supported by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB, and by the SGR programme 2021-SGR-00478 of the Catalan Government., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Cache-aware optimization of matrix multiplication and matrix factorizations on multicore processors

  • Martínez Pérez, Héctor
  • Catalán Pallarés, Sandra
  • Igual Peña, Francisco D.
  • Herrero Zaragoza, José Ramón|||0000-0002-4060-367X
  • Rodríguez Sánchez, Rafael
  • Quintana Ortí, Enrique Salvador
This paper advocates for a careful customization of the special general matrix multiplication (GEMM) kernels that are invoked from blocked routines for several relevant matrix factorizations in LAPACK, in order to improve their performance on modern multicore processors with hierarchical cache memories. To achieve this, we leverage a refined analytical model to dynamically tune the cache configuration parameters of GEMM for these kernels, taking into account the matrix operands’ dimensions, in order to improve cache occupation. In addition, toward the same goal, we accommodate a flexible development of architecture-specific micro-kernels for GEMM that allows us to select the option that, depending on the operands’ dimensions, ameliorates cache utilization. Our experiments for the LU and QR factorizations on two platforms, equipped with ARM (NVIDIA Carmel) and x86 (AMD EPYC) multi-core processors, demonstrate the benefits of this approach in terms of a better cache utilization and, in general, higher performance. Moreover, they also reveal the delicate balance between optimizing for multi-threaded parallelism versus cache usage as well as the positive effects of software prefetching., This work was supported by grants PID2020- 113656RB-C22, PID2019-107255GB, PID2021-126576NB-I00 and PID2021-123627OB-C52 of MCIN/AEI/10.13039/501100011033, by ‘‘ERDF A way of making Europe’’, and 2021-SGR-01007 of the Generalitat de Catalunya. Héctor Martínez is a postdoctoral fellow supported by the Consejería de Transformación Económica, Industria, Conocimiento y Universidades de la Junta de Andalucía. Sandra Catalán was supported by the grant RYC2021-033973- I, funded by MCIN/AEI/10.13039/501100011033 and the European Union ‘‘NextGenerationEU’’/PRTR. Funding Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature., 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 Artículo científico (JournalArticle)

Open project-based learning for dynamic adaptability of IT education

  • Tous Liesa, Rubén|||0000-0002-1409-5843
  • Freitag, Fèlix|||0000-0001-5438-479X
The pace of technological change is accelerating, the duration of the trends is becoming shorter and the number of technological alternatives to solve IT problems grows quickly. Keeping IT higher education course curricula up-to-date becomes more and more challenging. This paper presents an empirical study that aims to assess whether an open-statement and open-solution methodology in an IT-related project-based learning (PBL) course led to the adaptation of the skills learned by students to the trends of the main IT technology domains. The study analyses data from more than 90 projects by students from eight academic years of the course” Project on Information Technologies”, carried out by students in the last year of the Bachelor of Science in Informatics Engineering at the Barcelona School of Informatics. Our results identify a high correlation between the technologies chosen in student projects and technological trends, which shows that the methodology encourages students to incorporate the latest technological innovations in their project work plan., 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 contracts PID2019-107255GB and PID2019-106774RB-C21, and by the SGR programmes (2017-SGR-962 and 2017-SGR-990) of the Catalan Government., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

SpChar: characterizing the sparse puzzle via decision trees

  • Sgherzi, Francesco
  • Siracusa, Marco
  • Fernández Vega, Ivan
  • Armejach Sanosa, Adrià|||0000-0003-2869-668X
  • Moretó Planas, Miquel|||0000-0002-9848-8758
Sparse matrix computation is crucial in various modern applications, including large-scale graph analytics, deep learning, and recommender systems. The performance of sparse kernels varies greatly depending on the structure of the input matrix, making it difficult to gain a comprehensive understanding of sparse computation and its relationship to inputs, algorithms, and target machine architecture. Despite extensive research on certain sparse kernels, such as Sparse Matrix-Vector Multiplication (SpMV), the overall family of sparse algorithms has yet to be investigated as a whole. This paper introduces SpChar, a workload characterization methodology for general sparse computation. SpChar employs tree-based models to identify the most relevant hardware and input characteristics, starting from hardware and input-related metrics gathered from Performance Monitoring Counters (PMCs) and matrices. Our analysis enables the creation of a characterization loop that facilitates the optimization of sparse computation by mapping the impact of architectural features to inputs and algorithmic choices. We apply SpChar to more than 600 matrices from the SuiteSparse Matrix collection and three state-of-the-art Arm Central Processing Units (CPUs) to determine the critical hardware and software characteristics that affect sparse computation. In our analysis, we determine that the biggest limiting factors for high-performance sparse computation are (1) the latency of the memory system, (2) the pipeline flush overhead resulting from branch misprediction, and (3) the poor reuse of cached elements. Additionally, we propose software and hardware optimizations that designers can implement to create a platform suitable for sparse computation. We then investigate these optimizations using the gem5 simulator to achieve a significant speedup of up to 2.63× compared to a CPU where the optimizations are not applied., This research was supported by: the Spanish Ministry of Science and Innovation (MCIN) through contracts PID2019-107255GB-C21, PID2019-107255GBC21 (MCIN/AEI/10.13039/501100011033) and the European Processor Initiative (EPI) under grant agreement No. 101036168. M. Siracusa has been partially supported by an FI fellowship (2022FI B 00969). Finally, A. Armejach is a Serra Hunter Fellow, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A differential privacy protection-based federated deep learning framework to fog-embedded architectures

  • Gutiérrez Escobar, Norma
  • Otero Calviño, Beatriz|||0000-0002-9194-559X
  • Rodríguez Luna, Eva|||0000-0001-5904-7039
  • Utrera Iglesias, Gladys Miriam|||0000-0002-0637-1297
  • Mus León, Sergi
  • Canal Corretger, Ramon|||0000-0003-4542-204X
Nowadays, companies collect massive quantities of data to enhance their operations, often at the expense of sharing user sensible information. This data is widely used to train Deep Learning (DL) neural networks to model, classify, or recognize complex data. These activities enable companies to offer an array of services to users, such as precise advertising and optimal location services. This study explores potential solutions for preserving privacy while utilizing DL applications. To address the privacy issue, we develop a privacy-preserving framework specifically designed for fog computing environments. Unlike traditional cloud computing architectures, fog embedded architectures only share a small portion of user data with a nearby fog node, ensuring that the majority of sensitive data remains secure. Within these fog nodes, we incorporate two additional algorithms, namely Generalization and Threshold, to enhance the privacy-preserving capabilities of the framework. The first algorithm, Generalization, introduces a validation dataset within the fog nodes which not only increases the accuracy of the fog-embedded framework but also ensures that user data is preserved. The second algorithm, Threshold, is responsible for protecting user data samples and reducing the amount of information sent to the server. By combining these two algorithms, we are able to provide an additional layer of protection for user privacy while still maintaining the accuracy of the model. We conduct an evaluation to test its effectiveness using two separate datasets. In addition, we analyze them through a Feed Forward Neural Network (FFNN) and compare the results with a traditional centralized architecture to validate the effectiveness of the proposed framework. The results of our evaluation demonstrate that the proposed privacy-preserving framework, when combined with the Generalization and Threshold algorithms, can preserve up to 38.44% of user data. Additionally, we were able to extend the framework to multiple fog nodes without compromising the network’s accuracy, as we only observed a 0.1% decrease in accuracy when using the proposed architecture. This study emphasizes the importance of preserving user information while using DL applications and provides a solution that trains the desired network without violating user privacy, hence preserving their anonymity. Overall, the study highlights the potential of Federated Deep Learning to improve the accuracy and privacy of DL applications in fog computing environments., This work is partially supported by the Spanish Ministry of Science and Innovation, Spain under contracts PID2021-124463OB-IOO and PID2019-107255GB-C22; by the Generalitat de Catalunya, Spain under grant 2021SGR00326; and by the DRAC (IU16-011591), the HORIZON Vitamin-V, Spain(101093062), the HORIZON-AG PHOENI2X, Spain (101070586) and the HORIZON HORSE, Spain (101096342) projects., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

L2C2: Last-level compressed-contents non-volatile cache and a procedure to forecast performance and lifetime

  • Escuín Blasco, Carlos
  • Ibáñez Marín, Pablo
  • Navarro, Denis
  • Monreal Arnal, Teresa|||0000-0002-0458-2234
  • Llaberia Griñó, José M.|||0000-0002-3753-4108
  • Viñals Yúfera, Victor
Several emerging non-volatile (NV) memory technologies are rising as interesting alternatives to build the Last-Level Cache (LLC). Their advantages, compared to SRAM memory, are higher density and lower static power, but write operations wear out the bitcells to the point of eventually losing their storage capacity. In this context, this paper presents a novel LLC organization designed to extend the lifetime of the NV data array and a procedure to forecast in detail the capacity and performance of such an NV-LLC over its lifetime. From a methodological point of view, although different approaches are used in the literature to analyze the degradation of an NV-LLC, none of them allows to study in detail its temporal evolution. In this sense, this work proposes a forecasting procedure that combines detailed simulation and prediction, allowing an accurate analysis of the impact of different cache control policies and mechanisms (replacement, wear-leveling, compression, etc.) on the temporal evolution of the indices of interest, such as the effective capacity of the NV-LLC or the system IPC. We also introduce L2C2, a LLC design intended for implementation in NV memory technology that combines fault tolerance, compression, and internal write wear leveling for the first time. Compression is not used to store more blocks and increase the hit rate, but to reduce the write rate and increase the lifetime during which the cache supports near-peak performance. In addition, to support byte loss without performance drop, L2C2 inherently allows N redundant bytes to be added to each cache entry. Thus, L2C2+N, the endurance-scaled version of L2C2, allows balancing the cost of redundant capacity with the benefit of longer lifetime. For instance, as a use case, we have implemented the L2C2 cache with STT-RAM technology. It has affordable hardware overheads compared to that of a baseline NV-LLC without compression in terms of area, latency and energy consumption, and increases up to 6-37 times the time in which 50% of the effective capacity is degraded, depending on the variability in the manufacturing process. Compared to L2C2, L2C2+6 which adds 6 bytes of redundant capacity per entry, that means 9.1% of storage overhead, can increase up to 1.4-4.3 times the time in which the system gets its initial peak performance degraded., All authors acknowledge support in part from grants (1) PID2019-105660RB-C21, PID2019-107255GB-C22, PID2019-103939RB-I00, or PDC2021-120898-I00 from Agencia Estatal de Investigación (AEI, https://www.aei.gob.es/) and Ministerio de Ciencia e Innovación (MCIN, https://www.ciencia.gob.es), (2) gaZ: T5820R research group from Aragón Goverment (http://www.aragon.es), (3) EU through FEDER and NextGenerationEU/PRTR programs by the DGA-FSE, and (4) by ISCIII under Project PI21/00440., 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 Artículo científico (JournalArticle)

DDS: integrating data analytics transformations in task-based workflows [version 1, peer review: 1 approved, 2 approved with reservations]

  • Mammadli, Nihad
  • Ejarque Artigas, Jorge
  • Álvarez Cid-Fuentes, Javier
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
High-performance data analytics (HPDA) is a current trend in e-science research that aims to integrate traditional HPC with recent data analytic frameworks. Most of the work done in this field has focused on improving data analytic frameworks by implementing their engines on top of HPC technologies such as Message Passing Interface. However, there is a lack of integration from an application development perspective. HPC workflows have their own parallel programming models, while data analytic (DA) algorithms are mainly implemented using data transformations and executed with frameworks like Spark. Task-based programming models (TBPMs) are a very efficient approach for implementing HPC workflows. Data analytic transformations can also be decomposed as a set of tasks and implemented with a task-based programming model. In this paper, we present a methodology to develop HPDA applications on top of TBPMs that allow developers to combine HPC workflows and data analytic transformations seamlessly. A prototype of this approach has been implemented on top of the PyCOMPSs task- based programming model to validate two aspects: HPDA applications can be seamlessly developed and have better performance than Spark. We compare our results using different programs. Finally, we conclude with the idea of integrating DA into HPC applications and evaluation of our method against Spark., This research was financially supported by the European Union’s Horizon 2020 research and innovation programme under the grant agreement No 780622; and the Spanish Government (PID2019-107255GB), Generalitat de Catalunya (2014-SGR-1051)., 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 Artículo científico (JournalArticle)

Exploiting mixed-precision redundancy for soft-error detection in LU decomposition

  • Sahraneshinsamani, Nima
  • Catalán Pallarés, Sandra
  • Herrero Zaragoza, José Ramón|||0000-0002-4060-367X
Mixed-precision arithmetic, widely adopted for performance, creates an overlooked opportunity for fault tolerance. Our mixed-precision panel factorization (MPF) algorithm already computes each LU panel in both FP16 and FP64; we show that this inherent redundancy enables soft-error detection at negligible additional cost. We propose four complementary detectors that compare checksum sketches of the two factorizations: Relative (thresholded checksum discrepancy), Ratio (outlier analysis of per-row differences), Hybrid (a randomized-probe combination of both with separate L- and U-factor checks), and CrossCheck (a lightweight cross-check inspired by algorithm-based fault tolerance). Experimental evaluation shows that Relative and CrossCheck achieve high sensitivity with low false-positive rates, while Hybrid maintains near-zero false positives with moderate sensitivity and complementary coverage for challenging significand-bit faults. Bit-level analysis confirms reliable detection of exponent- and sign-bit errors, with sensitivity decreasing gracefully for low-order significand bits whose perturbations approach the FP16 rounding floor. A lightweight panel-fingerprint mechanism extends protection beyond the factorization loop, closing the temporal gap before the triangular solve with guaranteed detection of any single-element corruption and zero false positives. The approach requires only linear work and constant storage per panel, preserving the cubic scaling of standard LU decomposition., The authors would like to thank the Generalitat Valenciana for its support through project CIGE/2023/87. The researchers from UJI were additionally funded by project PID2023-146569NB-C22, supported by MICIU/AEI/10.13039/501100011033 and ERDF/EU. Sandra Catalán was supported by grant RYC2021-033973-I, funded by MCIN/AEI/10.13039/501100011033 and the European Union “NextGenerationEU”/PRTR. José R. Herrero was supported by project PID2019-107255GB of the Spanish Ministerio de Economía y Competitividad, and project 2021-SGR-01007 from the Generalitat de Catalunya., 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 Artículo científico (JournalArticle)

Heterogeneous programming using OpenMP and CUDA/HIP for hybrid CPU-GPU scientific applications

  • González Tallada, Marc|||0000-0002-3780-1106
  • Morancho Llena, Enrique|||0000-0003-2403-8145
Hybrid computer systems combine compute units (CUs) of different nature like CPUs, GPUs and FPGAs. Simultaneously exploiting the computing power of these CUs requires a careful decomposition of the applications into balanced parallel tasks according to both the performance of each CU type and the communication costs among them. This paper describes the design and implementation of runtime support for OpenMP hybrid GPU-CPU applications, when mixed with GPU-oriented programming models (e.g. CUDA/HIP). The paper describes the case for a hybrid multi-level parallelization of the NPB-MZ benchmark suite. The implementation exploits both coarse-grain and fine-grain parallelism, mapped to compute units of different nature (GPUs and CPUs). The paper describes the implementation of runtime support to bridge OpenMP and HIP, introducing the abstractions of Computing Unit and Data Placement. We compare hybrid and non-hybrid executions under state-of-the-art schedulers for OpenMP: static and dynamic task schedulings. Then, we improve the set of schedulers with two additional variants: a memorizing-dynamic task scheduling and a profile-based static task scheduling. On a computing node composed of one AMD EPYC 7742 @ 2.250 GHz (64 cores and 2 threads/core, totalling 128 threads per node) and 2 × GPU AMD Radeon Instinct MI50 with 32 GB, hybrid executions present speedups from 1.10× up to 3.5× with respect to a non-hybrid GPU implementation, depending on the number of activated CUs., This work was supported by the Spanish Ministry of Science and Technology (PID2019-107255GB)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Acceleration strategies for large-scale sequential simulations using parallel neighbour search: Non-LVA and LVA scenarios

  • Peredo Andrade, Oscar Francisco|||0000-0002-1148-6312
  • Herrero Zaragoza, José Ramón|||0000-0002-4060-367X
This paper describes the application of acceleration techniques into existing implementations of Sequential Gaussian Simulation and Sequential Indicator Simulation. These implementations might incorporate Locally Varying Anisotropy (LVA) to capture non-linear features of the underlying physical phenomena. The imple- mentation focuses on a novel parallel neighbour search algorithm, which can be used on both non-LVA and LVA codes. Additionally, parallel shortest path executions and optimized linear algebra libraries are applied with focus on LVA codes. Execution time, speedup and accuracy results are presented. Non-LVA codes are benchmarked using two scenarios with approximately 50 million domain points each. Speedup results of 2× and 4× were obtained on SGS and SISIM respectively, where each scenario is compared against a baseline code published in Peredo et al. (2018). The aggregated contribution to speedup of both works results in 12× and 50× respectively. LVA codes are benchmarked using two scenarios with approximately 1.7 million domain points each. Speedup results of 56× and 1822× were obtained on SGS and SISIM respectively, where each scenario is compared against the original baseline sequential codes., The authors acknowledge the donated resources from project PID2019-107255GB of the Spanish Ministerio de Economía y Competitividad, and project 2017-SGR-1414 from the Generalitat de Catalunya, Spain., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Practicable live container migrations in high performance computing clouds: Diskless, iterative, and connection-persistent

  • Guitart Fernández, Jordi|||0000-0003-0751-3100
Checkpoint/Restore techniques had been thoroughly used by the High Performance Computing (HPC) community in the context of failure recovery. Given the current trend in HPC to use containerization to obtain fast, customized, portable, flexible, and reproducible deployments of their workloads, as well as efficient and reliable sharing and management of HPC Cloud infrastructures, there is a need to integrate Checkpoint/Restore with containerization in such a way that the freeze time of the application is minimal and live migrations are practicable. Whereas current Checkpoint/Restore tools (such as CRIU) support several options to accomplish this, most of them are rarely exploited in HPC Clouds and, consequently, their potential impact on the performance is barely known. Therefore, this paper explores the use of CRIU’s advanced features to implement diskless, iterative (pre-copy and post-copy) migrations of containers with external network namespaces and established TCP connections, so that memory-intensive and connection-persistent HPC applications can live-migrate. Our extensive experiments to characterize the performance impact of those features demonstrate that properly-configured live migrations incur low application downtime and memory/disk usage and are indeed feasible in containerized HPC Clouds., This research was partially supported by the Spanish Government under contract PID2019-107255GB-C22, by the Generalitat de Catalunya, Spain under contract 2021-SGR-00478, and by the EU-HORIZON programme under grant agreement 101092646., 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 Artículo científico (JournalArticle)

Multi-GPU systems and Unified Virtual Memory for scientific applications: The case of the NAS multi-zone parallel benchmarks

  • González Tallada, Marc|||0000-0002-3780-1106
  • Morancho Llena, Enrique|||0000-0003-2403-8145
GPU-based computing systems have become a widely accepted solution for the high-performance-computing (HPC) domain. GPUs have shown highly competitive performance-per-watt ratios and can exploit an astonishing level of parallelism. However, exploiting the peak performance of such devices is a challenge, mainly due to the combination of two essential aspects of multi-GPU execution: memory allocation and work distribution. Memory allocation determines the data mapping to GPUs, and therefore conditions all work distribution schemes and communication phases in the application. Unified Virtual Memory simplifies the codification of memory allocations, but its effects on performance depend on how data is used by the devices and how the devices' driver is going to orchestrate the data transfers across the system. In this paper we present a multi-GPU and Unified Virtual Memory (UM) implementation of the NAS Multi-Zone Parallel Benchmarks which alternate communication and computation phases offering opportunities to overlap these phases. We analyse the programmability and performance effects of the introduction of the UM support. Our experience shows that the programming efforts for introducing UM are similar to those of having a memory allocation per GPU. On an evaluation environment composed of 2 x IBM Power9 8335-GTH and 4 x GPU NVIDIA V100 (Volta), our UM-based parallelization outperforms the manual memory allocation versions by 1.10x to 1.85x. However, these improvements are highly sensitive to the information forwarded to the devices' driver describing the most convenient location for specific memory regions. We analyse these improvements in terms of the relationship between the computational and communication phases of the applications., This work was supported by the Spanish Ministry of Science and Technology (PID2019-107255GB)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Exploring the potential of BCI in education: an experiment in musical training

  • Folgieri, Raffaella
  • Lucchiari, Claudio
  • Gricar, Sergej
  • Baldigara, Tea
  • Gil, Marisa|||0000-0001-9755-3311
Brain–computer interfaces (BCIs) have gained significant attention in recent years for various applications, including education and skill development: studies have shown that BCIs can boost memory, concentration, and even creativity and can improve learning and memory retention in healthy people. In our current study, we investigated the effectiveness of real-time feedback provided by a BCI system for improving performance on a specific task. A total of 20 participants completed a pre-training assessment, followed by a training period with the BCI system and a post-training assessment. The BCI system provided real-time feedback based on the participants’ level of accuracy, with positive feedback given for scores above 70%. Results showed a significant improvement in accuracy scores from pre- to post-training, with an average improvement of 15%. Participants also reported high levels of satisfaction with the feedback provided by the BCI system. These findings suggest that real-time feedback provided by a BCI system can be an effective tool for skill development and education, particularly when tailored to the specific needs of individual learners. Further research is needed to explore the potential of BCIs for a wide range of educational applications., 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 (PID2019-107255GB-C22) and partially supported by the Generalitat de Catalunya (contract 2021-SGR-01007). Certain sections of this article were written using tools funded by the Slovenian Research and Innovation Agency under grant number CRP2023 V5-2331. The same institutions funded the APC., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Automated generation of high-performance computational fluid dynamics codes

  • Macià Sorrosal, Sandra
  • Martínez Ferrer, Pedro J.
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Beltran Querol, Vicenç
Domain-Specific Languages (DSLs) improve programmers productivity by decoupling problem descriptions from algorithmic implementations. However, DSLs for High-Performance Computing (HPC) have two additional critical requirements: performance and scalability. This paper presents the automated process of generating, from abstract mathematical specifications of Computational Fluid Dynamics (CFD) problems, optimised parallel codes that perform and scale as manually optimised ones. We consciously combine within Saiph, a DSL for solving CFD problems, low-level optimisations and parallelisation strategies, enabling high-performance single-core executions which effectively scale to multi-core and distributed environments. Our results demonstrate how high-level DSLs can offer competitive performance by transparently leveraging state-of-the-art HPC techniques., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement N. 955606 (DEEP-SEA), and is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255 GB), and by the Generalitat de Catalunya (2017-SGR1414). This work is also supported by the Ministry of Economy of Spain through Severo Ochoa Center of Excellence Program (SEV-2015-0493)., 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 Artículo científico (JournalArticle)

Lester: rotoscope animation through video object segmentation and tracking

  • Tous Liesa, Rubén|||0000-0002-1409-5843
This article introduces Lester, a novel method to automatically synthesize retro-style 2D animations from videos. The method approaches the challenge mainly as an object segmentation and tracking problem. Video frames are processed with the Segment Anything Model (SAM) and the resulting masks are tracked through subsequent frames with DeAOT, a method of hierarchical propagation for semi-supervised video object segmentation. The geometry of the masks’ contours is simplified with the Douglas–Peucker algorithm. Finally, facial traits, pixelation and a basic rim light effect can be optionally added. The results show that the method exhibits an excellent temporal consistency and can correctly process videos with different poses and appearances, dynamic shots, partial shots and diverse backgrounds. The proposed method provides a more simple and deterministic approach than diffusion models based video-to-video translation pipelines, which suffer from temporal consistency problems and do not cope well with pixelated and schematic outputs. The method is also more feasible than techniques based on 3D human pose estimation, which require custom handcrafted 3D models and are very limited with respect to the type of scenes they can process., This research is partially funded by the Spanish Ministry of Science and Innovation under contract PID2019-107255GB, and by the SGR programme 2021-SGR-00478 of the Catalan Government., 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 Artículo científico (JournalArticle)

Role-shifting threads: Increasing OpenMP malleability to address load imbalance at MPI and OpenMP

  • Criado Ledesma, Joel
  • López Herrero, Víctor
  • Vinyals Ylla Català, Joan
  • Ramirez Miranda, Guillem
  • Teruel García, Xavier
  • Garcia Gasulla, Marta|||0000-0003-3682-9905
This paper presents the evolution of the free agent threads for OpenMP to the new role-shifting threads model and their integration with the Dynamic Load Balancing (DLB) library. We demonstrate how free agent threads can improve resource utilization in OpenMP applications with load imbalance in their nested parallel regions. We also demonstrate how DLB efficiently manages the malleability exposed by the role-shifting threads to address load imbalance issues. We use three real-world scientific applications, one of them to demonstrate that free agents alone can improve the OpenMP model without external tools, and two other MPI+OpenMP applications, one of them with a coupling case, to illustrate the potential of the free agent threads’ malleability with an external resource manager to increase the efficiency of the system. In addition, we demonstrate that the new implementation is more usable than the former one, letting the runtime system automatically make decisions that were made by the programmer previously. All software is released open-source., This work has received funding from the DEEP Projects, at the European Commission’s FP7, H2020, and EuroHPC Programmes, under Grant Agreements 287530, 610476, 754304, and 955606. The PCI2021-121958 financed by the Spanish State Research Agency - Ministry of Science and Innovation. And it also has the support of the Spanish Ministry of Science and Innovation (Computacion de Altas Prestaciones VIII: PID2019-107255GB)., 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 Artículo científico (JournalArticle)

Role-shifting threads: Increasing OpenMP malleability to address load imbalance at MPI and OpenMP

  • Criado Ledesma, Joel
  • López Herrero, Víctor
  • Vinyals Ylla Català, Joan
  • Ramirez Miranda, Guillem
  • Teruel García, Xavier
  • Garcia Gasulla, Marta
This paper presents the evolution of the free agent threads for OpenMP to the new role-shifting threads model and their integration with the Dynamic Load Balancing (DLB) library. We demonstrate how free agent threads can improve resource utilization in OpenMP applications with load imbalance in their nested parallel regions. We also demonstrate how DLB efficiently manages the malleability exposed by the role-shifting threads to address load imbalance issues. We use three real-world scientific applications, one of them to demonstrate that free agents alone can improve the OpenMP model without external tools, and two other MPI+OpenMP applications, one of them with a coupling case, to illustrate the potential of the free agent threads’ malleability with an external resource manager to increase the efficiency of the system. In addition, we demonstrate that the new implementation is more usable than the former one, letting the runtime system automatically make decisions that were made by the programmer previously. All software is released open-source., This work has received funding from the DEEP Projects, at the European Commission’s FP7, H2020, and EuroHPC Programmes, under Grant Agreements 287530, 610476, 754304, and 955606. The PCI2021-121958 financed by the Spanish State Research Agency - Ministry of Science and Innovation. And it also has the support of the Spanish Ministry of Science and Innovation (Computacion de Altas Prestaciones VIII: PID2019-107255GB)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Boosting HPC data analysis performance with the ParSoDA-Py library

  • Belcastro, Loris
  • Giampà, Salvatore
  • Marozzo, Fabrizio
  • Talia, Domenico
  • Trunfio, Paolo
  • Badia Sala, Rosa Maria
  • Ejarque, Jorge
  • Mammadli, Nihad
Developing and executing large-scale data analysis applications in parallel and distributed environments can be a complex and time-consuming task. Developers often find themselves diverted from their application logic to handle technical details about the underlying runtime and related issues. To simplify this process, ParSoDA, a Java library, has been proposed to facilitate the development of parallel data mining applications executed on HPC systems. It simplifies the process by providing built-in scalability mechanisms relying on the Hadoop and Spark frameworks. This paper presents ParSoDA-Py, the Python version of the ParSoDA library, which allows for further support of commonly used runtimes and libraries for big data analysis. After a complete library redesign, ParSoDA can be now easily integrated with other Python-based distributed runtimes for HPC systems, such as COMPSs and Apache Spark, and with the large ecosystem of Python-based data processing libraries. The paper discusses the adaptation process, which takes into consideration the new technical requirements, and evaluates both usability and scalability through some case study applications., This work has been partially funded by the European Commission’s Horizon 2020 Framework program and the European High-Performance Computing Joint Undertaking (JU) under Grant agreement No 955558 and by MCIN/AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR (PCI2021-121957), project eFlows4HPC. It has also been supported by the Spanish Government (PID2019-107255GB) and by the Departament de Recerca i Universitats de la Generalitat de Catalunya to the Research Group MPiEDist (2021 SGR 00412) We also acknowledge financial support from “National Centre for HPC, Big Data and Quantum Computing," CN00000013 - CUP H23C22000360005, and from “FAIR - Future Artificial Intelligence Research" Project - CUP H23C22000860006., Peer Reviewed, Postprint (published version)
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 Artículo científico (JournalArticle)

Programming parallel dense matrix factorizations and inversion for new-generation NUMA architectures

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

Automated generation of high-performance computational fluid dynamics codes

  • Macià Sorrosal, Sandra
  • Martínez Ferrer, Pedro J.
  • Ayguadé Parra, Eduard
  • Beltran Querol, Vicenç
Domain-Specific Languages (DSLs) improve programmers productivity by decoupling problem descriptions from algorithmic implementations. However, DSLs for High-Performance Computing (HPC) have two additional critical requirements: performance and scalability. This paper presents the automated process of generating, from abstract mathematical specifications of Computational Fluid Dynamics (CFD) problems, optimised parallel codes that perform and scale as manually optimised ones. We consciously combine within Saiph, a DSL for solving CFD problems, low-level optimisations and parallelisation strategies, enabling high-performance single-core executions which effectively scale to multi-core and distributed environments. Our results demonstrate how high-level DSLs can offer competitive performance by transparently leveraging state-of-the-art HPC techniques., This research has received funding from the European Union’s Horizon 2020/EuroHPC research and innovation programme under grant agreement N. 955606 (DEEP-SEA), and is supported by the Spanish State Research Agency - Ministry of Science and Innovation (contract PID2019-107255 GB), and by the Generalitat de Catalunya (2017-SGR1414). This work is also supported by the Ministry of Economy of Spain through Severo Ochoa Center of Excellence Program (SEV-2015-0493)., 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 Artículo científico (JournalArticle)

Low-cost training of image-to-image diffusion models with incremental learning and task/domain adaptation

  • Antona Pizà, Héctor
  • Otero Calviño, Beatriz
  • Tous Liesa, Rubén
Diffusion models specialized in image-to-image translation tasks, like inpainting and colorization, have outperformed the state of the art, yet their computational requirements are exceptionally demanding. This study analyzes different strategies to train image-to-image diffusion models in a low-resource setting. The studied strategies include incremental learning and task/domain transfer learning. First, a base model for human face inpainting is trained from scratch with an incremental learning strategy. The resulting model achieves an FID score almost equivalent to that of its batch learning equivalent while significantly reducing the training time. Second, the base model is fine-tuned to perform a different task, image colorization, and, in a different domain, landscape images. The resulting colorization models showcase exceptional performances with a minimal number of training epochs. We examine the impact of different configurations and provide insights into the ability of image-to-image diffusion models for transfer learning across tasks and domains., This work is partially supported by the Spanish Ministry of Science and Innovation under contracts PID2019-107255GB and PID2021-124463OB-IOO, by the Generalitat de Catalunya under grants 2017-SGR-962, 2021-SGR-00326, 2021-SGR-00478 and DRAC (IU16-011591). The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation program under the HORIZON-AG PHOENI2X (101070586) and HORIZON-EU VITAMIN-V (101093062) projects., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

SpChar: characterizing the sparse puzzle via decision trees

  • Sgherzi, Francesco
  • Siracusa, Marco
  • Fernández Vega, Ivan
  • Armejach Sanosa, Adrià
  • Moretó Planas, Miquel
Sparse matrix computation is crucial in various modern applications, including large-scale graph analytics, deep learning, and recommender systems. The performance of sparse kernels varies greatly depending on the structure of the input matrix, making it difficult to gain a comprehensive understanding of sparse computation and its relationship to inputs, algorithms, and target machine architecture. Despite extensive research on certain sparse kernels, such as Sparse Matrix-Vector Multiplication (SpMV), the overall family of sparse algorithms has yet to be investigated as a whole. This paper introduces SpChar, a workload characterization methodology for general sparse computation. SpChar employs tree-based models to identify the most relevant hardware and input characteristics, starting from hardware and input-related metrics gathered from Performance Monitoring Counters (PMCs) and matrices. Our analysis enables the creation of a characterization loop that facilitates the optimization of sparse computation by mapping the impact of architectural features to inputs and algorithmic choices. We apply SpChar to more than 600 matrices from the SuiteSparse Matrix collection and three state-of-the-art Arm Central Processing Units (CPUs) to determine the critical hardware and software characteristics that affect sparse computation. In our analysis, we determine that the biggest limiting factors for high-performance sparse computation are (1) the latency of the memory system, (2) the pipeline flush overhead resulting from branch misprediction, and (3) the poor reuse of cached elements. Additionally, we propose software and hardware optimizations that designers can implement to create a platform suitable for sparse computation. We then investigate these optimizations using the gem5 simulator to achieve a significant speedup of up to 2.63× compared to a CPU where the optimizations are not applied., This research was supported by: the Spanish Ministry of Science and Innovation (MCIN) through contracts PID2019-107255GB-C21, PID2019-107255GBC21 (MCIN/AEI/10.13039/501100011033) and the European Processor Initiative (EPI) under grant agreement No. 101036168. M. Siracusa has been partially supported by an FI fellowship (2022FI B 00969). Finally, A. Armejach is a Serra Hunter Fellow, Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Automatic distributed deep learning using resource-constrained edge devices

  • Gutiérrez Torre, Alberto
  • Bahadori, Kiyana
  • Baig, Shuja-ur-Rehman
  • Iqbal, Waheed
  • Vardanega, Tullio
  • Berral García, Josep Lluís
  • Carrera Pérez, David
Processing data generated at high volume and speed from the Internet of Things, smart cities, domotic, intelligent surveillance, and e-healthcare systems require efficient data processing and analytics services at the Edge to reduce the latency and response time of the applications. The Fog Computing Edge infrastructure consists of devices with limited computing, memory, and bandwidth resources, which challenge the construction of predictive analytics solutions that require resource-intensive tasks for training machine learning models. In this work, we focus on the development of predictive analytics for urban traffic. Our solution is based on deep learning techniques localized in the Edge, where computing devices have very limited computational resources. We present an innovative method for efficiently training of Gated Recurrent-Units (GRUs) across available resource-constrained CPU and GPU Edge devices. Our solution employs distributed GRU model learning and dynamically stops the training process to utilize the low-power and resource-constrained Edge devices while ensuring good estimation accuracy effectively. The proposed solution was extensively evaluated using low-powered ARM-based devices, including Raspberry Pi v3 and the low-powered GPU-enabled device NVIDIA Jetson Nano, and also compared them with Single-CPU Intel Xeon machines. For the evaluation experiments, we used real-world Floating Car Data. The experiments show that the proposed solution delivers excellent prediction accuracy and computational performance on the Edge when compared with the baseline methods., This work has been partially supported by the Spanish Government (contract PID2019-107255GB), the Generalitat de Catalunya (contract 2014-SGR-1051), the University of Padova, and Severo Ochoa CoE (SEV-2015-0493-16-5). The statements made herein are solely the responsibility of the authors., Peer Reviewed, Postprint (author's final draft)