Pasar al contenido principal

TECNICAS PARA LA MEJORA DE LAS PRESTACIONES, COSTE Y CONSUMO DE ENERGIA DE LOS SERVIDORES

TIN2015-66972-C5-1-R

•
Nombre agencia financiadora Ministerio de Economía y Competitividad
Acrónimo agencia financiadora MINECO
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos I+D+i Retos de Investigación
Año convocatoria 2015
Plan Estatal 2013-2016 Plan Estatal de Investigación Científica y Técnica y de Innovación
Unidad de gestión Dirección General de Investigación Científica y Técnica
Centro beneficiario UNIVERSITAT POLITÈCNICA DE VALÈNCIA
Identificador persistente http://dx.doi.org/10.13039/501100003329

Resultados relacionados

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

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A Hardware Approach to Fairly Balance the Inter-Thread Interference in Shared Caches

  • Selfa-Oliver, Vicent
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[EN] Shared caches have become the common design choice in the vast majority of modern multi-core and many-core processors, since cache sharing improves throughput for a given silicon area. Sharing the cache, however, has a downside: the requests from multiple applications compete among them for cache resources, so the execution time of each application increases over isolated execution. The degree in which the performance of each application is affected by the interference becomes unpredictable yielding the system to unfairness situations. This paper proposes Fair-Progress Cache Partitioning (FPCP), a low-overhead hardware-based cache partitioning approach that addresses system fairness. FPCP reduces the interference by allocating to each application a cache partition and adjusting the partition sizes at runtime. To adjust partitions, our approach estimates during multicore execution the time each application would have taken in isolation, which is challenging. The proposed approach has two main differences over existing approaches. First, FPCP distributes cache ways incrementally, which makes the proposal less prone to estimation errors. Second, the proposed algorithm is much less costly than the state-of-the-art ASM-Cache approach. Experimental results show that, compared to ASM-Cache, FPCP reduces unfairness by 48 percent in four-application workloads and by 28 percent in eight-application workloads, without harming the performance., This work was supported in part by the Spanish Ministerio de Economia y Competitividad (MINECO) and Plan E funds, under grants TIN2014-62246-EXP and TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Accurately modeling the on-chip and off-chip GPU memory subsystem

  • Candel-Margaix, Francisco
  • Duato Marín, José Francisco
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
[EN] Research on GPU architecture is becoming pervasive in both the academia and the industry because these architectures offer much more performance per watt than typical CPU architectures. This is the main reason why massive deployment of GPU multiprocessors is considered one of the most feasible solutions to attain exascale computing capabilities. The memory hierarchy of the GPU is a critical research topic, since its design goals widely differ from those of conventional CPU memory hierarchies. Researchers typically use detailed microarchitectural simulators to explore novel designs to better support GPGPU computing as well as to improve the performance of GPU and CPU-GPU systems. In this context, the memory hierarchy is a critical and continuously evolving subsystem. Unfortunately, the fast evolution of current memory subsystems deteriorates the accuracy of existing state-of-the-art simulators. This paper focuses on accurately modeling the entire (both on-chip and off-chip) GPU memory subsystem. For this purpose, we identify four main memory related components that impact on the overall performance accuracy. Three of them belong to the on-chip memory hierarchy: (i) memory request coalescing mechanisms, (ii) miss status holding registers, and (iii) cache coherence protocol; while the fourth component refers to the memory controller and GDDR memory working activity. To evaluate and quantify our claims, we accurately modeled the aforementioned memory components in an extended version of the state-of-the-art Multi2Sim heterogeneous CPUGPU processor simulator. Experimental results show important deviations, which can vary the final system performance provided by the simulation framework up to a factor of three. The proposed GPU model has been compared and validated against the original framework and the results from a real AMD Southern-Islands 7870HD GPU. (C) 2017 Elsevier B.V. All rights reserved., This work was supported in part by Generalitat Valenciana under grant AICO/2016/059, by the Spanish Ministerio de Economía y Competitividad (MINECO) and Plan E funds under Grant TIN2015-66972-C5-1-R, and by Programa de Ayudas de Investigación y Desarrollo (PAID) de la Universitat Politècnica de València .
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Accurately Modeling a Photonic NoC in a Detailed CMP Simulation Framework

  • Puche Lara, José
  • Lechago Buendía, Sergio
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
Photonic interconnects are a promising solution for the so-called communication bottleneck in current Chip Multiprocessor (CMPs) architectures. This technology presents an inherent low-latency and power consumption almost independent of communication distance, which are really desirable features in future Networks on Chip for next CMPs generations. However, since nanophotonic technology is still growing and therefore in an immature state, current simulators of detailed systems may not provide accurate models of photonic components. In this context, non-representative results are obtained when unaccurate photonic models are assumed. This paper summarizes all of the components that conform a fully operative photonic NoC and presents their current state of the art. Moreover, we evaluate a realistic photonic network that consists of two photonic rings and a token-based arbitration mechanism and compare it against a non-realistic model. In addition, both realistic and non-realistic schemes are valuated under different configurations varying the number of wavelengths that photonic waveguides employ. The experimental results show that the non-realistic NoC presents up 6× network latency deviation with respect to the accurate model. This deviation is translated into a performance deviation higher than 10% in several applications studied, which demonstrates the importance of accurate models when simulating current technologies under development like nanophotonics. Finally, a power consumption model of the realistic photonic network is presented. The results show that the overall photonic network power consumption grows with the number of wavelengths per waveguide since the number of required modulators and receivers becomes higher. In this way, the proposed realistic photonic network, which employs only two wavelengths for arbitration and destination selection tasks, increases its power consumption up to 3%, so network designs with more complex arbitration mechanisms must take into account the impact of the number of wavelengths on the power consumption., This work was supported by the Spanish Ministerio de Economía y Competitividad (MINECO) and by Plan E funds under Grant TIN2015-66972-C5-1-R and the ExaNest project, funded by the European Union’s Horizon 2020 research andinnovation programme under grant agreement No 671553.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Energy efficient torus networks with on/off links

  • Andújar, Francisco J.
  • Sánchez, José L.
  • Alfaro, Francisco J.
  • Martínez, Raúl
  • Coll, Salvador|||0000-0002-7795-7252
  • Alonso Díaz, Marina|||0000-0001-7079-5898
  • Martínez-Rubio, Juan-Miguel|||0000-0002-9748-0332
  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
[EN] Future exascale computing systems will require energy and performance efficient interconnection networks to respond to the high data movement demands of new applications, such as those coming from big-data and artificial intelligence areas. The network structure plays a major role in the overall interconnect performance, for this reason torus is a common topology used in the current largest supercomputers. There are several proposals to improve energy efficiency of interconnection networks. However, few works combine both energy and performance, and sometimes they are treated as opposed issues. In this paper, we try to determine which torus network configuration offers the best performance/energy ratio when high-radix switches are used to build the interconnect system. The performance/energy evaluation has been performed by trace-driven simulation under realistic scenarios, where several mixes of scientific applications share a supercomputer system and are scheduled to be executed with the available resources at each moment., This work has been supported by the Spanish MINECO and European Commission (FEDER funds) under project TIN2015-66972-05-1-R and project TIN2015-66972-05-2-R. Francisco J. Andujar is also funded by the Spanish MINECO under a Juan de la Cierva grant FJCI-2015-26080.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

On Microarchitectural Mechanisms for Cache Wearout Reduction

  • Valero Bresó, Alejandro
  • Miralaei, Negar
  • Jones, Timothy M.
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
[EN] Hot carrier injection (HCI) and bias temperature instability (BTI) are two of the main deleterious effects that increase a transistor's threshold voltage over the lifetime of a microprocessor. This voltage degradation causes slower transistor switching and eventually can result in faulty operation. HCI manifests itself when transistors switch from logic "0" to "1" and vice versa, whereas BTI is the result of a transistor maintaining the same logic value for an extended period of time. These failure mechanisms are especially acute in those transistors used to implement the SRAM cells of first-level (L1) caches, which are frequently accessed, so they are critical to performance, and they are continuously aging. This paper focuses on microarchitectural solutions to reduce transistor aging effects induced by both HCI and BTI in the data array of L1 data caches. First, we show that the majority of cell flips are concentrated in a small number of specific bits within each data word. In addition, we also build upon the previous studies, showing that logic "0" is the most frequently written value in a cache by identifying which cells hold a given logic value for a significant amount of time. Based on these observations, this paper introduces a number of architectural techniques that spread the number of flips evenly across memory cells and reduce the amount of time that logic "0" values are stored in the cells by switching OFF specific data bytes. Experimental results show that the threshold voltage degradation savings range from 21.8% to 44.3% depending on the application., This work was supported in part by the Spanish Ministerio de Economia y Competitividad within the Plan E Funds under Grant TIN2015-66972-C5-1-R, in part by the HiPEAC Collaboration Grant funded by the FP7 HiPEAC Network of Excellence under Grant 287759, and in part by the Engineering and Physical Sciences Research Council under Grant EP/K026399/1 and Grant EP/J016284/1.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Workload Characterization for Exascale Computing Networks

  • Duro-Gómez, José
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[EN] Exascale computing is the next step in high performance computing provided by systems composed of millions of interconnected processing cores. In order to guide the design and implementation of such systems, multiple workload characterization studies and system performance evaluations are required. This paper provides a workload characterization study in the context of the European Project ExaNeSt, which focuses, among others, on developing the network technology required to implement future exascale systems. In this work, we characterize different ExaNeSt applications from the computer network perspective by analyzing the distribution of messages, the dynamic bandwidth consumption, and the spatial communication patterns among cores. The analysis highlights three main observations; i) message sizes are, in general, below 50 kB; ii) communication patterns are usually bursty; and iii) spatial communication among cores concentrate in hot spots for most applications. Taking into account these observations, one can conclude that in order to unclog congested network links, an exascale network must be designed to briefly support higher-than-average bandwidths in the vicinity of key network cores., This work was supported by the ExaNest project, funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 671553, and by the Spanish Ministerio de Econom´ıa y Competitividad (MINECO) and Plan E funds under Grant TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Enforcing Predictability of Many-cores with DCFNoC

  • Picornell-Sanjuan, Tomás
  • Duato Marín, José Francisco
  • Flich Cardo, José
  • Hernández Luz, Carles|||0000-0003-1192-6223
[EN] The ever need for higher performance forces industry to include technology based on multi-processors system on chip (MPSoCs) in their safety-critical embedded systems. MPSoCs include a network-on-chip (NoC) to interconnect the cores between them and with memory and the rest of shared resources. Unfortunately, the inclusion of NoCs compromises guaranteeing time predictability as network-level conflicts may occur. To overcome this problem, in this paper we propose DCFNoC, a new time-predictable NoC design paradigm where conflicts within the network are eliminated by design. This new paradigm builds on top of the Channel Dependency Graph (CDG) in order to deterministically avoid network conflicts. The network guarantees predictability to applications and is able to naturally inject messages using a TDM period equal to the optimal theoretical bound without the need of using a computationally demanding offline process. DCFNoC is integrated in a tile-based many-core system and adapted to its memory hierarchy. Our results show that DCFNoC guarantees time predictability avoiding network interference among multiple running applications. DCFNoC always guarantees performance and also improves wormhole performance in a 4 × 4 setting by a factor of 3.7× when interference traffic is injected. For a 8 × 8 network differences are even larger. In addition, DCFNoC obtains a total area saving of 10.79% over a standard wormhole implementation., This work has been supported by MINECO under Grant BES-2016-076885, by MINECO and funds from the European ERDF under Grant TIN2015-66972-C05-1-R and Grant RTI2018-098156-B-C51, and by the EC H2020 RECIPE project under Grant 801137.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A research-oriented course on Advanced Multicore Architecture: Contents and active learning methodologies

  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
  • Selfa-Oliver, Vicent
[EN] The fast evolution of multicore processors makes it difficult for professors to offer computer architecture courses with updated contents. To deal with this shortcoming that could discourage students, the most appropriate solution is a research-oriented course based on current microprocessor industry trends. Additionally, we also seek to improve the students' skills by applying active learning methodologies, where teachers act as guiders and resource providers while students take the responsibility for their learning. In this paper, we present the Advanced Multicore Architecture (AMA) course, which follows a research-oriented approach to introduce students in architectural breakthroughs and uses active learning methodologies to enable students to develop practical research skills such as critical analysis of research papers or communication abilities. To this end five main activities are used: (i) lectures dealing with key theoretical concepts, (ii) paper review & discussion, (iii) research-oriented practical exercises, (iv) lab sessions with a state-of-the-art multicore simulator, and (v) paper presentation. An important part of all these activities is driven by active learning methodologies. Special emphasis is put on the practical side by allocating 40% of the time to labs and exercises. This work also includes an assessment study that analyzes both the course contents and the used methodology (both of them compared to other courses)., This work was supported in part by the Spanish Ministerio de Economia y Competitividad (MINECO) and by Plan E funds under Grant TIN2014-62246-EXP and Grant TIN2015-66972-C5-1-R, and by Generalitat Valenciana under grant AICO/2016/059. Authors also would like to thank Onur Mutlu for making available online his valuable teaching material.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Enhancing large-scale docking simulation on heterogeneous systems: An MPI vs rCUDA study

  • Imbernón, Baldomero
  • Gimenez Canovas, Domingo
  • Cecilia, J. M.
  • Prades, Javier|||0000-0003-3349-2200
  • Silla, Federico|||0000-0002-6435-1200
[EN] Virtual Screening (VS) methods can considerably aid clinical research by predicting how ligands interact with pharmacological targets, thus accelerating the slow and critical process of finding new drugs. VS methods screen large databases of chemical compounds to find a candidate that interacts with a given target. The computational requirements of VS models, along with the size of the databases, containing up to millions of biological macromolecular structures, means computer clusters are a must. However, programming current clusters of computers is no easy task, as they have become heterogeneous and distributed systems where various programming models need to be used together to fully leverage their resources. This paper evaluates several strategies to provide peak performance to a GPU-based molecular docking application called METADOCK in heterogeneous clusters of computers based on CPU and NVIDIA Graphics Processing Units (GPUs). Our developments start with an OpenMP, MPI and CUDA METADOCK version as a baseline case of cluster utilization. Next, we explore the virtualized GPUs provided by the rCUDA framework in order to facilitate the programming process. rCUDA allows us to use remote GPUs, i.e. installed in other nodes of the cluster, as if they were installed in the local node, so enabling access to them using only OpenMP and CUDA. Finally, several load balancing strategies are analyzed in a search to enhance performance. Our results reveal that the use of middleware like rCUDA is a convincing alternative to leveraging heterogeneous clusters, as it offers even better performance than traditional approaches and also makes it easier to program these emerging clusters., This work is jointly supported by the Fundacion Seneca (Agencia Regional de Ciencia y Tecnologia, Region de Murcia) under grant 18946/JLI/13, and by the Spanish MEC and European Commission FEDER under grants TIN2015-66972-C5-3-R and TIN2016-78799-P (AEI/FEDER, UE). We also thank NVIDIA for hardware donation under GPU Educational Center 2014-2016 and Research Center 2015-2016. Furthermore, researchers from Universitat Politecnica de Valencia are supported by the Generalitat Valenciana under Grant PROMETEO/2017/077. Authors are also grateful for the generous support provided by Mellanox Technologies Inc.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

HP-DCFNoC: High Performance Distributed Dynamic TDM Scheduler Based on DCFNoC Theory

  • Picornell-Sanjuan, Tomás
  • Duato Marín, José Francisco
  • Flich Cardo, José
  • Hernández Luz, Carles|||0000-0003-1192-6223
[EN] The need for increasing the performance of critical real-time embedded systems pushes the industry to adopt complex multi-core processor designs with embedded networks-on-chip. In this paper we present hp-DCFNoC, a distributed dynamic scheduler design that by relying on the key properties of a delayed confict-free NoC (DCFNoC) is able to achieve peak performance numbers very close to a wormhole-based NoC design without compromising its real-time guarantees. In particular, our results show that the proposed scheduler achieves an overall throughput improvement of 6.9x and 14.4x over a baseline DCFNoC for 16 and 64-node meshes, respectively. When compared against a standard wormhole router 95% of its network throughput is preserved while strict timing predictability as property is kept. This achievement opens the door to new high performance time predictable NoC designs., This work was supported in part by the Secretara de Estado de Investigacin Desarrollo e Innovacin (MINECO) under Grant BES-2016-076885, in part by the European Regional Development Fund (ERDF) under Grant TIN2015-66972-C05-1-R and Grant RTI2018-098156-B-C51, and in part by the EC H2020 European Institute of Innovation and Technology (SELENE) Project under Grant 871467.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

The Tag Filter Architecture: An energy-efficient cache and directory design

  • Valls, Joan
  • Ros Bardisa, Alberto
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
[EN] Power consumption in current high-performance chip multiprocessors (CMPs) has become a major design concern that aggravates with the current trend of increasing the core count. A significant fraction of the total power budget is consumed by on-chip caches which are usually deployed with a high associativity degree (even L1 caches are being implemented with eight ways) to enhance the system performance. On a cache access, each way in the corresponding set is accessed in parallel, which is costly in terms of energy. On the other hand, coherence protocols also must implement efficient directory caches that scale in terms of power consumption. Most of the state-of-the-art techniques that reduce the energy consumption of directories are at the cost of performance, which may become unacceptable for high-performance CMPs. In this paper, we propose an energy-efficient architectural design that can be effectively applied to any kind of cache memory. The proposed approach, called the Tag Filter (TF) Architecture, filters the ways accessed in the target cache set, and just a few ways are searched in the tag and data arrays. This allows the approach to reduce the dynamic energy consumption of caches without hurting their access time. For this purpose, the proposed architecture holds the XX least significant bits of each tag in a small auxiliary X-bit-wide array. These bits are used to filter the ways where the least significant bits of the tag do not match with the bits in the X-bit array. Experimental results show that, on average, the TF Architecture reduces the dynamic power consumption across the studied applications up to 74.9%74.9%, 85.9%85.9%, and 84.5%84.5% when applied to L1 caches, L2 caches, and directory caches, respectively., This work has been jointly supported by MINECO and European Commission (FEDER funds) under the project TIN2015-66972-C5-1-R/3-R and by Fundación Séneca, Agencia de Ciencia y Tecnología de la Región de Murcia under the project Jóvenes Líderes en Investigación 18956/JLI/13.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

POWAR: Power-Aware Routing in HPC Networks with On/Off Links

  • Andújar-Muñoz, Francisco José
  • Coll, Salvador|||0000-0002-7795-7252
  • Alonso Díaz, Marina|||0000-0001-7079-5898
  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
  • Martínez-Rubio, Juan-Miguel|||0000-0002-9748-0332
[EN] In order to save energy in HPC interconnection networks, one usual proposal is to switch idle links into a low-power mode after a certain time without any transmission, as IEEE Energy Efficient Ethernet standard proposes. Extending the low-power mode mechanism, we propose POWer-Aware Routing (POWAR), a simple power-aware routing and selection function for fat-tree and torus networks. POWAR adapts the amount of network links that can be used, taking into account the network load, and obtaining great energy savings in the network (55%-65%) and the entire system (9%-10%) with negligible performance overhead., This work has been supported by the Spanish MINECO and European Commission (FEDER funds) under project TIN2015-66972-C5-1-R. Francisco J. Andujar has been partially funded by the Spanish MICINN and by the ERDF program of the European Union: PCAS Project (TIN2017-88614-R), CAPAP-H6 (TIN2016-81840-REDT), and Junta de Castilla y Leon FEDER Grant VA082P17 (PROPHET Project).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Modeling and Analysis of the Performance of Exascale Photonic Networks

  • Duro-Gómez, José
  • Pascual Pérez, José Antonio
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[EN] Photonics technology has become a promising and viable alternative for both on-chip and off-chip interconnection networks of future Exascale systems. Nevertheless, this technology is not mature enough yet in this context, so research efforts focusing on photonic networks are still required to achieve realistic suitable network implementations. In this regard, system-level photonic network simulators can help guide designers to assess the multiple design choices. Most current research is done on electrical network simulators, whose components work widely different from photonics components. In this work, we summarize and compare the working behavior of both technologies which includes the use of optical routers, wavelength-division multiplexing and circuit switching among others. After implementing them into a well-known simulation framework, an extensive simulation study has been carried out using realistic photonic network configurations with synthetic and realistic traffic. Experimental results show that, compared to electrical networks, optical networks can reduce the execution time of the studied real workloads in almost one order of magnitude. Our study also reveals that the photonic configuration highly impacts on the network performance, being the bandwidth per channel and the message length the most important parameters., This work was supported by the ExaNeSt project, funded by the European Union's Horizon 2020 Research and Innovation Program under grant 671553, and by the Spanish Ministerio de Economía y Competitividad (MINECO) and Plan E funds under grant TIN2015-66972-C5-1-R. Pascual was supported by a HiPEAC Collaboration Grant.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Improving System Turnaround Time with Intel CAT by Identifying LLC Critical Applications

  • Pons-Escat, Lucía|||0000-0002-4582-7744
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Pons Terol, Julio|||0000-0002-5654-6753
  • Selfa, Vicent
[EN] Resource sharing is a major concern in current multicore processors. Among the shared system resources, the Last Level Cache (LLC) is one of the most critical, since destructive interference between applications accessing it implies more off-chip accesses to main memory, which incur long latencies that can severely impact the overall system performance. To help alleviate this issue, current processors implement huge LLCs, but even so, inter-application interference can harm the performance of a subset of the running applications when executing multiprogram workloads. For this reason, recent Intel processors feature Cache Allocation Technologies (CAT) to partition the cache and assign subsets of cache ways to groups of applications. This paper proposes the Critical-Aware (CA) LLC partitioning approach, which leverages CAT and improves the performance of multiprogram workloads, by identifying and protecting the applications whose performance is more damaged by LLC sharing. Experimental results show that CA improves turnaround time on average by 15%, and up to 40% compared to a baseline system without partitioning., This work was supported by the Spanish Ministerio de Economia y Competitividad (MINECO) and Plan E funds, under grants TIN2015-66972-C5-1-R and TIN2017-92139-EXP. It was also supported by the ExaNest project, with funds from the European Union Horizon 2020 project, with grant agreement No 671553.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

TokenTLB+CUP: A Token-Based Page Classification with Cooperative Usage Prediction

  • Esteve Garcia, Albert
  • Ros Bardisa, Alberto
  • Robles Martínez, Antonio
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[EN] Discerning the private or shared condition of the data accessed by the applications is an increasingly decisive approach to achieving efficiency and scalability in multi- and many-core systems. Since most memory accesses in both sequential and parallel applications are either private (accessed only by one core) or read-only (not written) data, devoting the full cost of coherence to every memory access results in sub-optimal performance and limits the scalability and efficiency of the multiprocessor. This paper introduces TokenTLB, a TLB-based page classification approach based on exchange and count of tokens. Token counting on TLBs is a natural and efficient way for classifying memory pages, and it does not require the use of complex and undesirable persistent requests or arbitration. In addition, classification is extended with Cooperative Usage Predictor (CUP), a token-based system-wide page usage predictor retrieved through TLB cooperation, in order to perform a classification unaffected by TLB size. Through cycle-accurate simulation we observed that TokenTLB spends 43.6% of cycles as private per page on average, and CUP further increases the time spent as private by 22.0%. CUP avoids 4 out of 5 TLB invalidations when compared to state-of-the-art predictors, thus proving far better prediction accuracy and making usage prediction an attractive mechanism for the first time., This work has been jointly supported by the MINECO and European Commission (FEDER funds) under the project TIN2015-66972-C5-1-R and TIN2015-66972-C5-3-R and the Fundacion Seneca-Agencia de Ciencia y Tecnologia de la Region de Murcia under the project Jovenes Lideres en Investigacion 18956/JLI/13.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

On a course on computer cluster configuration and administration

  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
  • Baydal Cardona, María Elvira
[EN] Computer clusters are today a cost-effective way of providing either high-performance and/or high-availability. The flexibility of their configuration aims to fit the needs of multiple environments, from small servers to SME and large Internet servers. For these reasons, their usage has expanded not only in academia but also in many companies. However, each environment needs a different ¿cluster flavour¿. High-performance and high-throughput computing are required in universities and research centres while high-performance service and high-availability are usually reserved to use in companies. Despite this fact, most university cluster computing courses continue to cover only high-performance computing, usually ignoring other possibilities. In this paper, a master-level course which attempts to fill this gap is discussed. It explores the different types of cluster computing as well as their functional basis, from a very practical point of view. As part of the teaching methodology, each student builds from scratch a computer cluster based on a virtualization tool. The entire process is designed to be scalable. The goal is to be able to apply it to an actual computer cluster with a larger number of nodes, such as those the students may subsequently encounter in their professional life., This work was supported in part by the Spanish Ministerio de Economia y Competitividad (MINECO) and by FEDER funds under Grant TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A fault-tolerant routing strategy for k-ary n-direct s-indirect topologies based on intermediate nodes

  • Peñaranda Cebrián, Roberto
  • Gran, Ernst Gunnar
  • Skeie, Tor
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
[EN] Exascale computing systems are being built with thousands of nodes. The high number of components of these systems significantly increases the probability of failure. A key component for them is the interconnection network. If failures occur in the interconnection network, they may isolate a large fraction of the machine. For this reason, an efficient fault-tolerant mechanism is needed to keep the system interconnected, even in the presence of faults. A recently proposed topology for these large systems is the hybrid k-ary n-direct s-indirect family that provides optimal performance and connectivity at a reduced hardware cost. This paper presents a fault-tolerant routing methodology for the k-ary n-direct s-indirect topology that degrades performance gracefully in presence of faults and tolerates a large number of faults without disabling any healthy computing node. In order to tolerate network failures, the methodology uses a simple mechanism. For any source-destination pair, if necessary, packets are forwarded to the destination node through a set of intermediate nodes (without being ejected from the network) with the aim of circumventing faults. The evaluation results shows that the proposed methodology tolerates a large number of faults. For instance, it is able to tolerate more than 99.5% of fault combinations when there are 10 faults in a 3-D network with 1000 nodes using only 1 intermediate node and more than 99.98% if 2 intermediate nodes are used. Furthermore, the methodology offers a gracious performance degradation. As an example, performance degrades only by 1% for a 2-D network with 1024 nodes and 1% faulty links., This work was supported by the Spanish Ministerio de Economía y Competitividad (MINECO), by FEDER funds under Grant TIN2015-66972-C5-1-R, by Programa de Ayudas de Investigación y Desarrollo (PAID) from Universitat Politècnica de alència and by the financial support of the FP7 HiPEAC Network of Excellence under grant agreement 287759
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Improving IBM POWER8 Performance Through Symbiotic Job Scheduling

  • Feliu-Pérez, Josué|||0000-0003-3017-4266
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Eyerman, Stijn
  • Eeckhout, Lieven
[EN] Symbiotic job scheduling, i.e., scheduling applications that co-run well together on a core, can have a considerable impact on the performance of processors with simultaneous multithreading (SMT) cores. SMT cores share most of their microarchitectural components among the co-running applications, which causes performance interference between them. Therefore, scheduling applications with complementary resource requirements on the same core can greatly improve the throughput of the system. This paper enhances symbiotic job scheduling for the IBM POWER8 processor. We leverage the existing cycle accounting mechanism to build an interference model that predicts symbiosis between applications. The proposed models achieve higher accuracy than previous models by predicting job symbiosis from throttled CPI stacks, i.e., CPI stacks of the applications when running in the same SMT mode to consider the statically partitioned resources, but without interference from other applications. The symbiotic scheduler uses these interference models to decide, at run-time, which applications should run on the same core or on separate cores. We prototype the symbiotic scheduler as a user-level scheduler in the Linux operating system and evaluate it on an IBM POWER8 server running multiprogram workloads. The symbiotic job scheduler significantly improves performance compared to both an agnostic random scheduler and the default Linux scheduler. Across all evaluated workloads in SMT4 mode, throughput improves by 12.4 and 5.1 percent on average over the random and Linux schedulers, respectively., This work was supported in part by the Spanish Ministerio de Econom ıa y Competitividad (MINECO) and Plan E funds, under grants TIN2015-66972- C5-1-R and TIN2014-62246-EXP, as well as by the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement No. 259295.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Caracterización de Cargas para Redes de Computación Exascale

  • Duro-Gómez, José
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[ES] La computación exascale es el siguiente paso en la computación de alto rendimiento proporcionada por sistemas compuestos por millones de núcleos de procesamiento interconectados. Para guiar el diseño e implementación de dichos sistemas, se requieren múltiples estudios de caracterización de carga de trabajo y evaluaciones del sistema. Este documento proporciona un estudio de caracterización de la carga de trabajo en el contexto del proyecto europeo ExaNeSt, que se centra, entre otros, en el desarrollo de la tecnologı́a de red necesaria para implementar futuros sistemas de exascale. En este trabajo, caracterizamos diferentes aplicaciones ExaNeSt desde la perspectiva de la red informática mediante el análisis de la distribución de mensajes por tamaño, el consumo de ancho de banda a lo largo de la ejecución y los patrones espaciales de comunicación entre los núcleos. El análisis destaca tres observaciones principales; i) los tamaños de los mensajes son, en general, inferiores a 50 kB; ii) los patrones de comunicación suelen ser de ráfaga; y iii) la comunicación espacial entre los núcleos se concentra en los puntos crı́ticos para la mayorı́a de las aplicaciones. Tomando en cuenta estas observaciones, uno puede concluir que para desbloquear enlaces de red congestionados, una red exascale debe diseñarse para soportar brevemente anchos de banda superiores a la media en las proximidaes de los núcleos claves de la red., Este trabajo fue apoyado por el proyecto ExaNest, financiado por el programa de investigación e innovación Horizon 2020 de la Unión Europea bajo el acuerdo de subvención No. 671553, y por el Ministerio de Economía y Competitividad (MINECO) y fondos del Plan E bajo Grant TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Teaching high-performance service in a cluster computing course

  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
  • Baydal Cardona, María Elvira
[EN] Most courses on cluster computing in graduate and postgraduate studies are focused on parallel programming and high-performance/high-throughput computing. This is the typical usage of clusters in academia and research centres. However, nowadays, many companies are providing web, mail and, in general, Internet services using computer clusters. These services require a different ``cluster flavour'': high-performance service and high availability. Despite the fact that computer clusters for each environment demand a different configuration, most university cluster computing courses keep focusing only on high-performance computing, ignoring other possibilities. In this paper, we propose several teaching strategies for a course on cluster computing that could fill this gap. The content developed here would be taught as a part of the course. The subject shows several strategies about how to configure, test and evaluate a high-availability/load-balanced Internet server. A virtualization-based platform is used to build a cluster prototype, using Linux as its operating system. Evaluation of the course shows that students knowledge and skills on the subject are improved at the end of the course. On the other hand, regarding the teaching methodology, the results obtained in the yearly survey of the University confirm student satisfaction., This work was supported in part by the Spanish Ministerio de Economia y Competitividad (MINECO) and by FEDER funds under Grant TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

TLB-Based Temporality-Aware Classification in CMPs with Multilevel TLBs

  • Esteve Garcia, Albert
  • Ros Bardisa, Alberto
  • Robles Martínez, Antonio
  • Duato Marín, José Francisco
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
[EN] Recent proposals are based on classifying memory accesses into private or shared in order to process private accesses more efficiently and reduce coherence overhead. The classification mechanisms previously proposed are either not able to adapt to the dynamic sharing behavior of the applications or require frequent broadcast messages. Additionally, most of these classification approaches assume single-level translation lookaside buffers (TLBs). However, deeper and more efficient TLB hierarchies, such as the ones implemented in current commodity processors, have not been appropriately explored. This paper analyzes accurate classification mechanisms in multilevel TLB hierarchies. In particular, we propose an efficient data classification strategy for systems with distributed shared last-level TLBs. Our approach classifies data accounting for temporal private accesses and constrains TLB-related traffic by issuing unicast messages on first-level TLB misses. When our classification is employed to deactivate coherence for private data in directory-based protocols, it improves the directory efficiency and, consequently, reduces coherence traffic to merely 53.0%, on average. Additionally, it avoids some of the overheads of previous classification approaches for purely private TLBs, improving average execution time by nearly 9% for large-scale systems., This work has been jointly supported by the MINECO and European Commission (FEDER funds) under the project TIN2015-66972-C5-1-R and TIN2015-66972-C5-3-R and the Fundacion Seneca-Agencia de Ciencia y Tecnologia de la Region de Murcia under the project Jovenes Lideres en Investigacion 18956/JLI/13.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Designing lab sessions focusing on real processors for computer architecture courses: A practical perspective

  • Feliu-Pérez, Josué|||0000-0003-3017-4266
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
[EN] Computer architecture courses typically include lab sessions to reinforce, from a practical perspective, concepts and architectural mechanisms studied in lectures. Lab sessions are mainly based on simulation frameworks because they benefit learning. Reading the source code that models certain processor mechanisms allows students to acquire a sound knowledge of how hardware works. Unfortunately, simulators that model current multicore processors are getting more and more complex, which lengthens the learning phase and complicates their use in time-bounded lab sessions. In this paper, we propose a new approach that complements the use of simulation frameworks in lab sessions of computer architecture courses. This approach is based on performing experiments on current commercial processors, where multiple hardware events related to the performance of the computer components under study are monitored. Then, students analyze the measured events and how they impact the overall performance. Such analysis motivates students and, not only helps reinforcing the theoretical concepts, but also increases their analysis skills. In this paper we present the methodology and scheduling framework that support the proposed approach and discuss five lab sessions, which can be applied in different courses, covering multiple computer architecture topics. (C) 2018 Elsevier Inc. All rights reserved., Josue Feliu has been partially supported through a postdoctoral fellowship by the Generalitat Valenciana (APOSTD/2017/052). Additional support has been provided by the Spanish Ministerio de Economia y Competitividad (MINECO) and Plan E funds, under grants TIN2015-66972-C5-1-R and TIN2014-62246-EXP.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

An Aging-Aware GPU Register File Design Based on Data Redundancy

  • Valero Bresó, Alejandro
  • Candel-Margaix, Francisco
  • Suárez-Gracia, Darío
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
[EN] Nowadays, GPUs sit at the forefront of high-performance computing thanks to their massive computational capabilities. Internally, thousands of functional units, architected to be fed by large register files, fuel such a performance. At deep nanometer technologies, the SRAM memory cells that implement GPU register files are very sensitive to the Negative Bias Temperature Instability (NBTI) effect. NBTI ages cell transistors by degrading their threshold voltage Vth over the lifetime of the GPU. This degradation, which manifests when a cell keeps the same logic value for a relatively long period of time, compromises the cell read stability and increases the transistor switching delay, which can lead to wrong read values and eventually exceed the processor cycle time, respectively, so resulting in faulty operation. Thiswork proposes architectural mechanisms leveraging the redundancy of the data stored in GPU register files to attack NBTI aging. The proposed mechanisms are based on data compression, power gating, and register address rotation techniques. All these mechanismsworking together balance the distribution of logic values stored in the cells along the execution time, reducing both the overall Vth degradation and the increase in the transistor switching delays. Experimental results show that a conventional GPU register file suffers the worst case for NBTI, since a significant fraction of the cells maintain the same logic value during the entire application execution (i.e., a 100 percent '0' and '1' duty cycle distributions). On average, the proposal reduces these distributions by 58 and 68 percent, respectively, which translates into Vth degradation savings by 54 and 62 percent, respectively., This work was supported by the Gobierno de Aragon and the European ESF (gaZ: T58_17R research group), and by the Ministerio de Economia y Competitividad (MINECO) and AEI/FEDER (EU) funds under Grants TIN2016-76635-C2-1-R and TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

XOR-based HoL-blocking Reduction Routing Mechanisms for Direct Networks

  • Peñaranda Cebrián, Roberto
  • Gómez Requena, Crispín
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
  • López Rodríguez, Pedro Juan|||0000-0003-4544-955X
[EN] Routing is a key design parameter in the interconnection network of large parallel computers. Routing algorithms are classified into two different categories depending on the number of routing options available for each source-destination pair: deterministic (there is one path available) and adaptive (there are several ones). Adaptive routing has two opposed effects on network performance. On one hand, it provides routing flexibility that may help on avoiding a congested network area, thus improving network performance. On the other hand, it also may increase the Head-of-Line blocking effect due to more destination nodes sharing the port queues. Usually, adaptive routing uses virtual channels to provide routing flexibility and to guarantee deadlock freedom. Deterministic routing is simpler, which implies lower routing delay and it introduces less Head-of-Line blocking effect. In this paper, we propose an adaptive and HoL-blocking reduction routing algorithm for direct topologies that tries to combine the good properties of both worlds: It provides routing flexibility but also reduces the Head-of-Line blocking effect. To do that, this paper proposes several functions which use the XOR operation to efficiently distribute the packets among virtual channels based on their destination node. The resulting routing mechanisms have different properties depending on whether they enforce routing flexibility or Head-of-Line blocking reduction., This work was supported by the Spanish Ministerio de Economia y Competitividad (MINECO) and by FEDER funds under Grant TIN2015-66972-05-1-R and by Programa de Ayudas de Investigacion y Desarrollo (PAID) from Universitat Politecnica de Valencia.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Constructing virtual 5-dimensional tori out of lower-dimensional network cards

  • Andújar-Muñoz, Francisco José
  • Villar, Juan Antonio
  • Sanchez Garcia, Jose Luis
  • Alfaro Cortes, Francisco Jose
  • Duato Marín, José Francisco
  • Fröning, Holger
[EN] In the Top500 and Graph500 lists of the last years, some of the most powerful systems implement a torus topology to interconnect themillions of computing nodes they include. Some of these torus networks are of five or six dimensions, which implies an additional difficulty as the node degree increases. In previous works, we proposed and evaluated the nD Twin (nDT) torus topology to virtually increase the dimensions a torus is able to implement. We showed that this new topology reduces the distances between nodes, increasing, therefore, global network performance. In this work, we present how to build a 5DT torus network using a specific commercial 6-port network card (EXTOLL card) to interconnect those nodes. We show, using the same number of cards, that the performance of the 5DT torus network we are able to implement using our proposal is higher than the performance of the 3D torus network for the same number of compute nodes., Spanish MINECO; European Commission, Grant/Award Number: TIN2015-66972-C5-1-R and TIN2015-66972-C5-2-R; JCCM, Grant/Award Number: PEII-2014-028-P; Spanish MICINN, Grant/Award Number: FJCI-2015-26080
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Modelado de una Red Fotónica para Computación Exascale

  • Duro Gómez, José
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
La mayoría de las investigaciones actuales se realizan en simuladores desarrollados para redes eléctricas, cuyos componentes difieren mucho de los componentes fotónicos. Además, la tecnología fotónica añade nuevos componentes que no están presentes en las redes eléctricas. Este artículo discute cómo se puede construir una herramienta de simulación de redes fotónicas a partir de un simulador de redes eléctricas. Se resume y compara el comportamiento de ambas tecnologías –eléctrica y fotónica– y se discuten las extensiones propuestas. Entre otros aspectos, las extensiones propuestas modelan routers ópticos, wavelength-division multiplexing, conmutación de circuitos y algoritmos de encaminamiento., Este trabajo ha sido apoyado por el proyecto ExaNest, financiado por el programa de investigación e innovación Horizon 2020 de la Unión Europea bajo el acuerdo de subvención no 671553, y los fondos del Ministerio de Economía y Competitividad (MINECO) y el Plan E bajo la subvención TIN2015-66972-C5-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Perf&amp, Fair: A Progress-Aware Scheduler to Enhance Performance and Fairness in SMT Multicores

  • Feliu-Pérez, Josué|||0000-0003-3017-4266
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Duato Marín, José Francisco
[EN] Nowadays, high performance multicore processors implement multithreading capabilities. The processes running concurrently on these processors are continuously competing for the shared resources, not only among cores, but also within the core. While resource sharing increases the resource utilization, the interference among processes accessing the shared resources can strongly affect the performance of individual processes and its predictability. In this scenario, process scheduling plays a key role to deal with performance and fairness. In this work we present a process scheduler for SMT multicores that simultaneously addresses both performance and fairness. This is a major design issue since scheduling for only one of the two targets tends to damage the other. To address performance, the scheduler tackles bandwidth contention at the L1 cache and main memory. To deal with fairness, the scheduler estimates the progress experienced by the processes, and gives priority to the processes with lower accumulated progress. Experimental results on an Intel Xeon E5645 featuring six dual-threaded SMT cores show that the proposed scheduler improves both performance and fairness over two state-of-the-art schedulers and the Linux OS scheduler. Compared to Linux, unfairness is reduced to a half while still improving performance by 5.6 percent., We thank the anonymous reviewers for their constructive and insightful feedback. This work was supported in part by the Spanish Ministerio de Economia y Competitividad (MINECO) and Plan E funds, under grants TIN2015-66972-C5-1-R and TIN2014-62246EXP, and by the Intel Early Career Faculty Honor Program Award.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Modeling a Photonic Network for Exascale Computing

  • Duro Gómez, José
  • Petit Martí, Salvador Vicente|||0000-0003-2426-4134
  • Sahuquillo Borrás, Julio|||0000-0001-8630-4846
  • Gómez Requena, María Engracia|||0000-0003-1466-4118
Photonics technology has become a promising and viable alternative for both on-chip and off-chip computer networks of future Exascale systems. Nevertheless, this technology is not mature enough yet in this context, so research efforts focusing on photonic networks are still required to achieve realistic suitable network implementations. In this context, system-level photonic network simulators can help to guide designers to assess the multiple design choices. Most current research is done on electrical network simulators, whose components work widely different from photonics components. Moreover, photonics technology adds new components that are not present in electrical networks. This paper discusses how a photonics simulation tool can be built by extending an electrical simulation framework. We summarize and compare the working behavior of both technologies -electrical and photonics, and discuss the rationale behind the proposed extensions. Among others, the devised extensions model optical routers, wavelength-division multiplexing, circuit switching, and specific routing algorithms. This work is aimed to provide support to investigate off- chip optical networks in the context of the European Exascale System Interconnect and Storage project (ExaNeSt) project. The experiments presented in this paper study multiple realistic photonic networks configurations and have been performed with excerpts of real traces. Experimental results show that, compared to electrical networks, optical networks can reduce the execution time of the workload by several orders of magnitude. Our study reveals that future optical technologies presenting a 3.2 Tbps aggregate link bandwidth will not provide additional performance benefits over state-of-the-art 1.6 Tbps optical links across the studied workloads, but 1.6 Tbps network links are enough to achieve the highest optical performance on computer networks. Regarding the link configuration, the bandwidth per optical channel is the parameter with highest impact on the network delay and so on the execution time, while for a given optical bandwidth per channel the better strategy is to reduce the phit size., This work was supported by the ExaNest project, funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 671553, and by the Spanish Ministerio de Economía y Competitividad (MINECO) and Plan E funds under Grant TIN2015-66972-C5-1-R.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Artículo científico (JournalArticle)

An aging-aware GPU register file design based on data redundancy

  • Valero Bresó, Alejandro
  • Candel Margaix, Francisco
  • Suárez Gracia, Darío
  • Petit Martí, Salvador Vicente
  • Sahuquillo Borrás, Julio
Nowadays, GPUs sit at the forefront of high-performance computing thanks to their massive computational capabilities. Internally, thousands of functional units, architected to be fed by large register files, fuel such a performance. At deep nanometer technologies, the SRAM memory cells that implement GPU register files are very sensitive to the Negative Bias Temperature Instability (NBTI) effect. NBTI ages cell transistors by degrading their threshold voltage Vth over the lifetime of the GPU. This degradation, which manifests when a cell keeps the same logic value for a relatively long period of time, compromises the cell read stability and increases the transistor switching delay, which can lead to wrong read values and eventually exceed the processor cycle time, respectively, so resulting in faulty operation. This work proposes architectural mechanisms leveraging the redundancy of the data stored in GPU register files to attack NBTI aging. The proposed mechanisms are based on data compression, power gating, and register address rotation techniques. All these mechanisms working together balance the distribution of logic values stored in the cells along the execution time, reducing both the overall Vth degradation and the increase in the transistor switching delays. Experimental results show that a conventional GPU register file suffers the worst case for NBTI, since a significant fraction of the cells maintain the same logic value during the entire application execution (i.e., a 100% ‘0’ and ‘1’ duty cycle distributions). On average, the proposal reduces these distributions by 58% and 68%, respectively, which translates into Vth degradation savings by 54% and 62%, respectively.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Artículo científico (JournalArticle)

Improving GPU cache hierarchy performance with a fetch and replacement cache

  • Candel, Francisco
  • Petit, Salvador
  • Valero, Alejandro
  • Sahuquillo, Julio
In the last few years, GPGPU computing has become one of the most popular computing paradigms in high-performance computers due to its excellent performance to power ratio. The memory requirements of GPGPU applications widely differ from the requirements of CPU counterparts. The amount of memory accesses is several orders of magnitude higher in GPU applications than in CPU applications, and they present disparate access patterns. Because of this fact, large and highly associative Last-Level Caches (LLCs) bring much lower performance gains in GPUs than in CPUs. This paper presents a novel approach to manage LLC misses that efficiently improves LLC hit ratio, memory-level parallelism, and miss latencies in GPU systems. The proposed approach leverages a small additional Fetch and Replacement Cache (FRC) that stores control and coherence information of incoming blocks until they are fetched from main memory. Then, fetched blocks are swapped with victim blocks to be replaced in the LLC. After that, the eviction of victim blocks is performed from the FRC. This management approach improves performance due to three main reasons: (i) the lifetime of blocks being replaced is increased, (ii) the main memory path is unclogged on long bursts of LLC misses, and (iii) the average L2 miss delaying latency is reduced. Experimental results show that our proposal increases the performance (OPC) over 25% in most of the studied applications, reaching improvements up to 150% in some applications.