Pasar al contenido principal

ARQUITECTURAS DE DOMINIO ESPECIFICO PARA SISTEMAS DE COMPUTACION ENERGETICAMENTE EFICIENTES II

PID2024-155476OB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para la Investigación y el Desarrollo Experimental
Subprograma Subprograma Estatal de Generación de Conocimiento Científico-Técnico y Desarrollo Experimental
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2024
Plan Estatal 2024-2027 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITECNICA DE CATALUNYA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Dissecting and modeling the architecture of modern GPU cores

  • Huerta Gañán, Rodrigo
  • Abaie Shoushtary, Mojtaba
  • Cruz Díaz, Josep Llorenç
  • González Colás, Antonio María
GPUs are the most popular platform for accelerating HPC workloads, such as artificial intelligence and science simulations. However, most microarchitectural research in academia relies on simulators that model GPU core architectures based on designs that are more than 15 years old, and differ significantly from modern core architectures. This work reverse engineers the architecture of modern NVIDIA GPU cores, unveiling key aspects of its design and the important role of the compiler in some of its main components. In particular, it reveals how the issue logic works, the structure of the register file and its associated cache, multiple features of the instruction and data memory pipelines. When modeling all these discovered microarchitectural details in a state-of-the-art simulation framework, we show that its accuracy is significantly improved, achieving a 20.58% reduction in mean absolute percentage error (MAPE) on average, which results in a 13.45% MAPE on average with respect to real modern hardware. In addition, we show that the software-based dependence management mechanism included in modern NVIDIA GPUs outperforms a hardware mechanism based on scoreboards in terms of performance and area., This work has been supported by the CoCoUnit ERC Advanced Grant of the EU’s Horizon 2020 program (grant No 833057), the Spanish State Research Agency (MCIU/AEI/ 10.13039/501100011033) and FEDER/UE under grants PID2020-113172RB-I00 and PID2024-155476OB-I00, the Catalan Agency for University and Research (AGAUR) under grant 2021SGR00383, the ICREA Academia program and Beques Santander - Ayudas predoctorales 2025., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación

WaRM: Warp reorder to reduce memory latency in graphics workloads

  • Joseph, Diya
  • Aragón Alcaraz, Juan Luis
  • Parcerisa Bundó, Joan Manuel
  • González Colás, Antonio María
GPUs are designed to handle long-latency operations effectively; however, challenges such as increasing memory latency and core occupancy (number of warps in a core) can impede their latency management. Increasing the number of warps in a core can help manage higher memory latency, but this has a high cost in Register File size, Instruction Buffer size, and other resources in the core. Moreover, in Tile-Based Rendering (TBR) GPUs, core occupancy might remain low for extended durations. To address this challenge, we introduce WaRM, a lightweight warp reordering technique tailored for GPUs in graphics applications. WaRM strategically initiates a select subset of warps, termed priority warps, early in execution to reduce memory latency for subsequent warps. This optimization taps into the inherent but underutilized memory parallelism within the GPUcore. This underutilization is a consequence of a baseline warp order that evenly spaces misses throughout execution to exploit the inherent spatial locality in graphics workloads. WaRM improves on this by reducing average memory latency while maintaining locality for the majority of warps. While maximizing memory parallelism utilization, WaRM prevents saturating the caches with misses to avoid filling up the MSHRs (Miss Status Holding Registers). This approach reduces cache stalls that halt further accesses to the cache. Overall, WaRM yields a significant 3.9% performance speedup. Importantly, WaRM accomplishes these enhancements with negligible overhead, positioning it as a promising solution for enhancing the efficiency of GPUs in managing the increasingly relevant latency challenges., This work has been supported by the CoCoUnit ERC Advanced Grant of the EU’s Horizon 2020 program (grant No 833057), the Spanish State Research Agency (MCIU/AEI/ 10.13039/501100011033) and FEDER/UE under grants PID2020-113172RB-I00 and PID2024-155476OB-I00, the Catalan Agency for University and Research (AGAUR) under grant 2021 SGR 00383, the ICREA Academia program and the AGAUR grant 2020-FISDU-00287., Preprint
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

ARAS: Adaptive low-cost ReRAM-based accelerator for DNNs

  • Sabri Abrebekoh, Mohammad
  • Riera Villanueva, Marc
  • González Colás, Antonio María
Processing Using Memory (PUM) accelerators have the potential to perform Deep Neural Network (DNN) inference by using arrays of memory cells as computation engines. Among various memory technologies, ReRAM crossbars show promising performance in computing dot-product operations in the analog domain. Nevertheless, the expensive writing procedure of ReRAM cells has led researchers to design accelerators whose crossbars have enough capacity to store the full DNN. Given the tremendous and continuous increase in DNN model sizes, this approach is unfeasible for some networks, or inefficient due to the huge hardware requirements. Those accelerators lack the flexibility to adapt to any given DNN model, facing an adaptability challenge. To address this issue we introduce ARAS, a cost-effective ReRAM-based accelerator that employs an offline scheduler to adapt different DNNs to the resource-limited hardware. ARAS also overlaps the computation of a layer with the weight writing of several layers to mitigate the high writing latency of ReRAM. Furthermore, ARAS introduces three optimizations aimed at reducing the energy overheads of writing in ReRAM. Our key optimization capitalizes on the observation that DNN weights can be re-encoded to augment their similarity between layers, increasing the amount of bitwise values that are equal or similar when overwriting ReRAM cells and, hence, reducing the amount of energy required to update the cells. Overall, ARAS greatly reduces the ReRAM writing activity. We evaluate ARAS on a popular set of DNNs. ARAS provides up to 2. 2x speedup and 45% energy savings over a baseline PUM accelerator without any optimization. Compared to a TPU-like accelerator, ARAS provides up to 1. 5x speedup and 62% energy savings., This work has been supported by the CoCoUnit ERC Advanced Grant of the EU’s Horizon 2020 program (grant No 833057), the Spanish State Research Agency (MCIN/AEI) under grants PID2020-113172RBI00 and PID2024-155476OB-I00, the Catalan Agency for University and Research (AGAUR) under grant 2021SGR00383, the ICREA Academia program, and the FPI-UPC 2020 grant., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Towards efficient LUT-based PIM: A scalable and low-power approach for modern workloads

  • Khabbazan, Bahareh
  • Riera Villanueva, Marc
  • González Colás, Antonio María
Data movement in memory-intensive workloads is several orders of magnitude more energy-consuming than computation. Frequent memory-processor transfers make this a critical bottleneck. Processing-using-memory (PuM) alleviates this by enabling in-memory computation and reducing data movement. In this paper, we propose Lama, a LUT-based PuM architecture that efficiently executes SIMD operations by supporting independent column accesses within each mat of a DRAM subarray. Lama minimizes the number of energy-intensive memory activation commands, which are the primary source of overhead in most PuM architectures. Besides, Lama supports up to 8-bit operand precision without decomposing computations, unlike prior PuM solutions, while incurring only a 2.47% area overhead. Our evaluation shows that Lama achieves an average performance improvement of 8.5× over state-of-the-art PuM architectures and a 3.8× improvement over CPU, along with energy efficiency gains of 6.9×/8×, respectively, for bulk 8-bit multiplication. We introduce LamaAccel, an HBM-based PuM accelerator that leverages Lama to accelerate LLM inference. By transforming MAC computations into energy-efficient addition and counting operations, LamaAccel achieves up to 9.3×/19.2× energy reduction and 4.8×/9.8× speedup over a TPU/GPU, as well as up to 5.8× lower energy consumption and 2.1× higher performance compared to a state-of-the-art PuM baseline., This work has been supported by the CoCoUnit ERC Advanced Grant of the EU’s Horizon 2020 program (grant No 833057), the Spanish State Research Agency (MCIN/AEI) under grants PID2020-113172RB-I00 and PID2024-155476OBI00, the Catalan Agency for University and Research (AGAUR) under grant 2021SGR00383, the ICREA Academia program, and the FPI-UPC 2020 grant., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

BGP anomaly detection using the raw internet topology

  • Latif Martínez, Hamid
  • Paillissé Vilanova, Jordi
  • Barlet Ros, Pere
  • Cabellos Aparicio, Alberto
The Border Gateway Protocol (BGP) is central to the global connectivity of the Internet, enabling fast and efficient dissemination of routing information. Hence, detecting any anomaly concerning BGP announcements is of critical importance to ensure the continuous operation of Internet services. Typically, BGP anomaly detection algorithms have relied on features of the BGP messages, such as the average length of the AS_PATH attribute, the volume of messages, or the type of message (announcement or withdrawal). Even though these algorithms provide good performance, they do not take into account the Internet topology, that is, the graph of Autonomous Systems (AS) created by the BGP announcements. In addition, some of the existing algorithms can detect only specific types of anomalies, while others require retraining them to support new scenarios. In this paper we propose detecting BGP anomalies by leveraging the raw BGP topology graph, instead of manually curated features of the BGP messages. We implement a Machine Learning algorithm to process the entire BGP topology and evaluate it with real-world data from 4 well-known incidents. We compare our proposal against two state-of-the-art solutions and a classical method that use BGP features and features of the BGP topology, not the topology itself. Our results show that our solution obtains remarkable performance identifying the incidents. Finally, we test our model with regular data (non-anomalous) to prove that it can be used in a production scenario, with samples processed on the fly and guaranteeing a low false alarm rate., This work is part of the I+D+i project titled BLOSSOMS, grant PID2024-158530OB-I00, funded by MICIU/AEI/10.13039/501100011033/ and by ERDF/EU. This work is also partially funded by the Catalan Institution for Research and Advanced Studies (ICREA). This work was also supported by Grant PCI2023-145974-2 funded by MICIU/AEI/10.13039/501100011033 and co-funded by the European Union (GRAPHS4SEC project). Jordi Paillissé is funded by European Union-Next Generation EU, Ministry of Universities and Recovery, Transformation and Resilience Plan, through a call from Universitat Politècnica de Catalunya (Grant Ref. 2022UPC-MSC-93871)., Peer Reviewed, Postprint (published version)