Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 30
Encontrada(s) 1 página(s)

UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

SafeSU-2 L: an advanced multicore interference statistics unit for a RISC-V space SoC

  • Bas Jalón, Francisco
  • Alcaide Portet, Sergi|||0000-0001-8561-7795
  • Cabo Pitarch, Guillem
  • Lasfar, Ilham
  • Chang, Feng
  • Fuentes Díaz, Francisco Javier
  • Canal Corretger, Ramon|||0000-0003-4542-204X
  • Benedicte Illescas, Pedro|||0000-0003-1670-7783
  • Rodríguez Rivas, Juan Carlos
  • Abella Ferrer, Jaume|||0000-0001-7951-4028
Multicores for critical real-time embedded systems (CRTES) may experience interference across tasks running in different cores when accessing shared hardware resources such as shared caches and memory controllers. Precise interference diagnostics (e.g., what task interferes what other task and how much) are key for the optimization and validation of safety-related real-time applications during development, and to diagnose overruns during operation. The SafeSU statistics unit has been proposed recently for that purpose, and proven successful for systems-on-chip (SoCs) where interference can occur at a single centralized locations (e.g., a bus). However, it is unable to monitor interference in multiple-level interconnects, especially if request ownership is not available. This paper extends the SafeSU to 2-level interconnects (SafeSU-2 L), where interference can occur in the bus connecting the cores with a shared second level cache (L2C), and in the DDR4 memory controller serving L2C misses, with the latter losing track of the actual core issuing each DDR4 request. In particular, the SafeSU-2 L monitors some additional signals from the buses and caches to infer what core interferes what other core in any of the shared resources. Moreover, the SafeSU-2 L is integrated and tested on a 4-core version of the commercial CAES Gaisler NOEL-XCKU-EX space SoC., The research leading to these results has received funding from the Horizon Europe Programme under the NimbleAI Project (nimbleai.eu), grant agreement num. 101070679. Authors appreciate the support given to the Research Group SSAS (Code: 2021 SGR 00637) by the Research and University Department of the Generalitat de Catalunya., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

A qualifiable GPU sharing approach for AI workloads in critical systems

  • Solé i Bonet, Marc|||0000-0002-9634-2026
  • Wolf, Jannis
  • Álvarez Suárez, Aridane
  • Kosmidis, Leonidas
© 2025 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., Graphics Processing Units (GPUs) are increasingly considered for adoption in safety-critical systems and especially space, in order to enable high performance on-board processing, particularly needed for Artificial Intelligence (AI) acceleration. While use of GPUs has been demonstrated in qualifiable real-time operating systems, GPU sharing has not been explored so far. In this work we describe we describe the design, implementation and evaluation of such GPU sharing approach, which is portable among different operating systems. The proposed solution has been implemented and demonstrated with two real-world AI applications for space missions. The results demonstrate significant performance improvements and efficient AI utilization. By addressing the timing constraints and enabling GPU utilization, this approach opens up new possibilities for efficient and reliable AI-powered space systems., The research presented in this paper has received funding by the European Community’s Horizon Europe programme under the METASAT project (grant agreement 101082622) and the Chips Act Joint Undertaking project SMARTY (grant agreement 101140087). Moreover, it has been supported by the project PID2024-155745OB-I00 funded by Spanish MICIU/AEI/10.13039/501100011033 / FEDER, UE (IAFE)., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

SafeSU-2 L: an advanced multicore interference statistics unit for a RISC-V space SoC

  • Bas Jalón, Francisco
  • Alcaide Portet, Sergi
  • Cabo Pitarch, Guillem
  • Lasfar, Ilham
  • Chang, Feng
  • Fuentes Díaz, Francisco Javier
  • Canal Corretger, Ramon
  • Benedicte Illescas, Pedro
  • Rodríguez Rivas, Juan Carlos
  • Abella Ferrer, Jaume
Multicores for critical real-time embedded systems (CRTES) may experience interference across tasks running in different cores when accessing shared hardware resources such as shared caches and memory controllers. Precise interference diagnostics (e.g., what task interferes what other task and how much) are key for the optimization and validation of safety-related real-time applications during development, and to diagnose overruns during operation. The SafeSU statistics unit has been proposed recently for that purpose, and proven successful for systems-on-chip (SoCs) where interference can occur at a single centralized locations (e.g., a bus). However, it is unable to monitor interference in multiple-level interconnects, especially if request ownership is not available. This paper extends the SafeSU to 2-level interconnects (SafeSU-2 L), where interference can occur in the bus connecting the cores with a shared second level cache (L2C), and in the DDR4 memory controller serving L2C misses, with the latter losing track of the actual core issuing each DDR4 request. In particular, the SafeSU-2 L monitors some additional signals from the buses and caches to infer what core interferes what other core in any of the shared resources. Moreover, the SafeSU-2 L is integrated and tested on a 4-core version of the commercial CAES Gaisler NOEL-XCKU-EX space SoC., The research leading to these results has received funding from the Horizon Europe Programme under the NimbleAI Project (nimbleai.eu), grant agreement num. 101070679. Authors appreciate the support given to the Research Group SSAS (Code: 2021 SGR 00637) by the Research and University Department of the Generalitat de Catalunya., Peer Reviewed, Postprint (author's final draft)
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Speed and Conversational Large Language Models: Not All Is About Tokens per Second

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • González Saiz, Miguel
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Gao, Zhen
  • Liu, Shanshan
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Contribución de congreso (ConferenceOutput)

How Stable is Stable Diffusion under Recursive InPainting (RIP)?

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • González Saiz, Miguel
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Moral, Fernando
  • Merino Gómez, Elena
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Are Emerging Machine Learning Models Dependable at the Nanoscales?

  • Liu, Shanshan
  • Wang, Ziheng
  • Gao, Zhen
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Niknia, Farzad
  • Tang, Xiaochen
  • Zhou, Jun
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Enhanced FIWARE-Based Architecture for Cyberphysical Systems With Tiny Machine Learning and Machine Learning Operations: A Case Study on Urban Mobility Systems

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Muñoz Arcentales, Andrés|||0000-0003-1554-437X
  • Alonso González, Álvaro|||0000-0002-8456-8351
  • Salvachúa Rodríguez, Joaquín|||0000-0002-7269-8079
  • Huecas Fernández-Toribio, Gabriel|||0000-0001-5673-9312
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Establishing vocabulary tests as a benchmark for evaluating large language models

  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Merino Gómez, Elena|||0000-0003-4129-4626
  • Bermúdez Margaretto, Beatriz
  • Hernández Gutiérrez, José Alberto|||0000-0002-9551-4308
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Brysbaert, Marc|||0000-0002-3645-3189
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Operating Conversational Large Language Models (LLMs)in the Presenceof Errors

  • Gao, Zhen|||0000-0001-9887-1418
  • Deng, Jie
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Liu, Shanshan|||0000-0001-6226-2880
  • Pozo Huertas, Alejando|||0000-0002-2160-1978
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

TELEIA: A Spanish language dataset for evaluating artificial intelligence models

  • Mayor Rocher, Marina|||0000-0002-4177-7559
  • Melero, Nina
  • Merino Gómez, Elena|||0000-0003-4129-4626
  • González Saiz, Miguel|||0009-0008-6709-938X
  • Ferrando Tarazona, Raquel
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Adaptive Separately Constrained Triplet Loss (A-SCTL) for High-Performance Triplet Networks

  • Wang, Ziheng|||0000-0001-9668-7318
  • Niknia, Farzad|||0000-0002-4062-3638
  • Liu, Shanshan|||0000-0001-6226-2880
  • Jiang, Honglan|||0000-0003-3705-4240
  • Liu, Siting|||0000-0003-0505-8183
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Zhou, Jun
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Robustness against Faults in Configuration Memories of FPGA-based LLMs

  • Gao, Zhen|||0000-0001-9887-1418
  • Yuan, Lini
  • Wang, Jingya
  • Liu, Qiang
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Zeng, Shulin
  • Wang, Yu
  • Liu, Shanshan
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Energy-Efficient Stochastic Computing (SC) Neural Networks for Internet of Things Devices With Layer-Wise Adjustable Sequence Length (ASL)

  • Wang, Ziheng|||0000-0001-9668-7318
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Niknia, Farzad|||0000-0002-4062-3638
  • Gao, Zhen|||0000-0001-9887-1418
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Liu, Shanshan|||0000-0001-6226-2880
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Recursive InPainting (RIP): how much information is lost under recursive inferences?

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • González Saiz, Miguel|||0009-0008-6709-938X
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Moral, Fernando
  • Merino Gómez, Elena|||0000-0003-4129-4626
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Contribución de congreso (ConferenceOutput)

Psycholinguistic Word Features: a New Approach for the Evaluation of LLMs Alignment with Humans

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Saiz González, Miguel
  • Grandury González, María
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Brysbaert, Marc|||0000-0002-3645-3189
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Speed and Conversational Large Language Models: Not All Is About Tokens per Second

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • González Saiz, Miguel|||0009-0008-6709-938X
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Gao, Zhen|||0000-0001-9887-1418
  • Liu, Shanshan|||0000-0001-6226-2880
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Concurrent Linguistic Error Detection (CLED): a New Methodology for Error Detection in Large Language Models

  • Zhu, Jinhua|||0000-0001-8136-8111
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Gao, Zhen|||0000-0001-9887-1418
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Liu, Shanshan|||0000-0001-6226-2880
  • Lombardi, Fabrizio|||0000-0003-3152-3245
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Tree-Based Architecture for Enabling Live Interaction in Massive Online Events

  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Pozo Huertas, Alejandro|||0000-0002-2160-1978
  • Alonso González, Álvaro|||0000-0002-8456-8351
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Spanish Is Not Just One: A Dataset of Spanish Dialect Recognition for LLMs

  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Mayor Rocher, Marina|||0000-0002-4177-7559
  • Pozo Huertas, Cris
  • Melero Carrasco, María Inmaculada
  • Grandury González, María
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Capítulo o parte de un libro (BookPart)

LLM Estimates of Word Characteristics

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Trott, Sean|||0000-0002-6003-3731
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Contribución de congreso (ConferenceOutput)

Spanish and LLM Benchmarks: Is MMLU Lost in Translation?

  • Plaza Ortiz, Irene
  • Melero Carrasco, María Inmaculada
  • Pozo, Cristina del
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Mayor Rocher, Marina|||0000-0002-4177-7559
  • Grandury González, María
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Have Multimodal Large Language Models (MLLMs) Really Learned to Tell the Time on Analog Clocks?

  • Fu, Tairan|||0009-0009-9571-7250
  • González Saiz, Miguel|||0009-0008-6709-938X
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Merino Gómez, Elena|||0000-0003-4129-4626
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Updating the German psycholinguistic word toolbox with AI-generated estimates of concreteness, valence, arousal, age of acquisition, and familiarity

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Grandury González, María|||0009-0009-4703-3348
  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Haro Rodríguez, Juan|||0000-0002-3456-4731
  • Schroeder, Sascha|||0000-0001-7001-4588
  • Hintz, Florian|||0000-0002-2444-3303
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Brysbaert, Marc|||0000-0002-3645-3189
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Open Conversational LLMs do not know most Spanish words

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • González, Miguel
  • Melero, Nina
  • Ferrando Tarazona, Raquel
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Merino Gomez, Elena
  • Hernández Gutiérrez, José Alberto|||0000-0002-9551-4308
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Multiple choice questions: reasoning makes large language models (LLMs) more self-confident, especially when they are wrong

  • Fu, Tairan|||0009-0009-9571-7250
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Grandury González, María|||0009-0009-4703-3348
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Contribución de congreso (ConferenceOutput)

Analyzing recursiveness in multimodal generative artificial intelligence: stability or divergence?

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Cheung, Tobias
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Sarkar, Rik|||0000-0001-7804-4351
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

When Do Large Language Models (LLMs) Struggle to Count Letters?

  • Fu, Tairan|||0009-0009-9571-7250
  • Ferrando Tarazona, Raquel
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Using AI-powered multiple-choice question generation for self-regulated learning = Uso de generación de preguntas de opción múltiple basada en inteligencia artificial para el aprendizaje autorregulado

  • Barra Arias, Enrique|||0000-0001-9532-8962
  • Pilicita Garrido, Anabel|||0000-0002-0796-7797
  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Pozo Huertas, Alejandro|||0000-0002-2160-1978
  • López Pernas, Sonsoles|||0000-0002-9621-1392
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Adding LLMs to the psycholinguistic norming toolbox: A practical guide to getting the most out of human ratings

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Grandury González, María|||0009-0009-4703-3348
  • Fu, Tairan|||0009-0009-9571-7250
  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Clark, Thomas|||0000-0002-1720-0450
  • Trott, Sean|||0000-0002-6003-3731
  • Gerarld Green, Clarence
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Brysbaert, Marc|||0000-0002-3645-3189
Archivo Digital UPM
Publicación Artículo científico (JournalArticle)

Adding LLMs to the psycholinguistic norming toolbox: a practical guide to getting the most out of human ratings

  • Conde Díaz, Javier|||0000-0002-5304-0626
  • Grandury González, María|||0009-0009-4703-3348
  • Fu, Tairan|||0009-0009-9571-7250
  • Arriaga Prieto, Carlos|||0000-0002-0513-2550
  • Martínez Ruiz, Gonzalo|||0000-0002-9125-6225
  • Clark, Thomas|||0000-0002-1720-0450
  • Trott, Sean|||0000-0002-6003-3731
  • Green, Clarence Gerald|||0000-0003-3463-141X
  • Reviriego Vasallo, Pedro|||0000-0003-2540-5234
  • Brysbaert, Marc|||0000-0002-3645-3189