Explore Microsoft's innovative approach to query optimization in production environments, addressing challenges with general-purpose optimization and introducing specialized cloud-based optimizers. Learn about the implementation details, experiments conducted, and the solution proposed. Discover how

2 views • 27 slides

Optimization in process engineering involves obtaining the best possible solution for a given process by minimizing or maximizing a specific performance criterion while considering various constraints. This process is crucial for achieving improved yields, reducing pollutants, energy consumption, an

10 views • 52 slides

Explore the intricate structure of the DRAM subsystem, including memory channels, DIMMs, ranks, chips, banks, and rows/columns. Delve into the breakdown of DIMMs, ranks, chips, and banks to comprehend the design and functioning of DRAM memory systems. Gain insights into address decoding, row/column

0 views • 16 slides

Swarm Intelligence (SI) is an artificial intelligence technique inspired by collective behavior in nature, where decentralized agents interact to achieve goals. Swarms are loosely structured groups of interacting agents that exhibit collective behavior. Examples include ant colonies, flocking birds,

2 views • 88 slides

The DNN Inference Optimization Challenge, organized by Liya Yuan from ZTE, focuses on optimizing deep neural network (DNN) models for efficient inference on-device, at the edge, and in the cloud. The challenge addresses the need for high accuracy while minimizing data center consumption and inferenc

0 views • 13 slides

In the field of computer architecture, SRAM and DRAM are two prevalent memory technologies with distinct characteristics. SRAM retains data as long as power is present, while DRAM is dynamic and requires data refreshing. SRAM is built with high-speed CMOS technology, whereas DRAM is more dense and b

3 views • 38 slides

DRAM-based QUAC-TRNG provides high-throughput and low-latency true random number generation by utilizing commodity DRAM devices. By employing Quadruple Row Activation (QUAC), this method outperforms existing TRNGs, achieving a 15.08x improvement in throughput and passing all 15 NIST randomness tests

0 views • 10 slides

SIMDRAM introduces a novel framework for efficient computation in DRAM, aiming to overcome data movement bottlenecks. It emphasizes Processing-in-Memory (PIM) and Processing-using-Memory (PuM) paradigms to enhance processing capabilities within DRAM while minimizing architectural changes. The motiva

2 views • 14 slides

Discrete Optimization is a field of applied mathematics that uses techniques from combinatorics, graph theory, linear programming, and algorithms to solve optimization problems over discrete structures. This involves creating mathematical models, defining objective functions, decision variables, and

0 views • 12 slides

This study delves into the generalization of Empirical Risk Minimization (ERM) in stochastic convex optimization, focusing on minimizing true objective functions while considering generalization errors. It explores the application of ERM in machine learning and statistics, particularly in supervised

0 views • 11 slides

Explore the world of optimization through convex and general problems, understanding the concepts, constraints, and the difference between convex and non-convex optimization. Discover the significance of local and global optima in solving complex optimization challenges.

0 views • 24 slides

Detailed experimental study reveals that DRAM power models may not provide accurate insights into power consumption. The increasing importance of managing DRAM power in system design is emphasized. The study delves into DRAM organization, operation, and power consumption patterns, highlighting the n

0 views • 43 slides

Understanding the liabilities and responsibilities of bar and tavern owners under the Dram Shop Act based on the case of Build It and They Will Drink, Inc. v. Strauch. The act outlines exceptions where licensees can be held civilly liable for selling alcohol to minors or visibly intoxicated individu

0 views • 12 slides

Limited DRAM bandwidth poses a critical bottleneck in GPU performance, necessitating a comprehensive scheduling policy to reduce cache miss rates, enhance DRAM bandwidth, and improve latency hiding for GPUs. The CTA-aware scheduling techniques presented address these challenges by optimizing resourc

0 views • 33 slides

Exploring the integration of Coherent DRAM Caches in multi-node systems to improve memory performance. Discusses the benefits, challenges, and potential performance improvements compared to existing memory-side cache solutions.

0 views • 28 slides

Explore the challenges faced by Moore's Law in relation to bandwidth limitations and the innovative solutions such as 3D-DRAM caches and compressed memory systems. Discover how compressing DRAM caches can improve bandwidth and capacity, leading to enhanced performance in memory-intensive application

0 views • 48 slides

Recent research highlights advancements in solving sampling problems in convex optimization, exemplified by works by Yin Tat Lee and Santosh Vempala. The complexity of convex problems, such as the Minimum Cost Flow Problem and Submodular Minimization, are being unraveled through innovative formulas

1 views • 47 slides

Addressing fundamental latency trade-offs in designing DRAM caches involves considerations such as memory stacking for improved latency and bandwidth, organizing large caches at cache-line granularity to minimize wasted space, and optimizing cache designs to reduce access latency. Challenges include

0 views • 32 slides

Demonstrating and analyzing RowPress, a novel read disturbance phenomenon causing bitflips in DRAM chips. Different from RowHammer vulnerability, RowPress showcases effective solutions on real Intel systems with DRAM chips.

0 views • 46 slides

Addressing memory latency challenges in irregular GPGPU applications, this study explores techniques like warp-aware memory scheduling and GPU memory controller optimization to reduce DRAM latency divergence. The research delves into the impact of SIMD lanes, coalescers, and warp-aware scheduling on

0 views • 33 slides

Despite extensive research, DRAM remains vulnerable to Rowhammer attacks. The Panopticon project proposes a novel in-DRAM mitigation technique using counter mats within DRAM devices. This approach does not require costly changes at multiple layers and leverages existing DRAM logic for efficient miti

0 views • 17 slides

This piece discusses approximation algorithms for stochastic optimization problems, focusing on modeling uncertainty in inputs, adapting to stochastic predictions, and exploring different optimization themes. It covers topics such as weakening the adversary in online stochastic optimization, two-sta

0 views • 33 slides

Exploring the nature of DRAM errors, this study delves into the causes, types, and implications for system design. From soft errors caused by cosmic rays to hard errors due to permanent hardware issues, the research examines error protection mechanisms and open questions surrounding DRAM errors. Pre

0 views • 31 slides

Explore the innovative use of stacked DRAM as Part of Memory (PoM) to increase overall memory capacity and eliminate duplication. The system involves OS-managed PoM, challenges, and the potential of hardware-managed PoM to reduce OS-related overhead. Learn about the practical implications and evalua

0 views • 24 slides

The evolution of memory scaling poses challenges for system performance growth due to limitations in memory hierarchy, capacity gaps, and DRAM scaling obstacles. The need for alternative technologies and architectural support to address these challenges is highlighted, focusing on reducing latency,

0 views • 23 slides

This research delves into the complexities of latency variation in modern DRAM chips, highlighting factors such as imperfect manufacturing processes and high standard latencies chosen to boost yield. The study aims to characterize latency variation, optimize DRAM performance, and develop mechanisms

0 views • 37 slides

This collection of research delves into the power challenges faced by memory-intensive databases (MMDBs) and explores strategies for reducing DRAM power draw. Topics covered include the impact of hardware features on power consumption, experimental setups for analyzing power breakdown, and the effec

0 views • 13 slides

Memory requests from different threads can cause interferences in DRAM banks, impacting performance. The solution proposed involves partitioning DRAM banks between threads to eliminate interferences, leading to improved performance and energy savings.

0 views • 32 slides

Reduce average DRAM access latency by leveraging row access locality with ChargeCache, a cost-effective solution requiring no modifications to existing DRAM chips. By tracking recently accessed rows and adjusting timing parameters, ChargeCache achieves higher performance and lower DRAM energy consum

0 views • 33 slides

Real-world design problems often require multi-objective optimization, and the Flower Pollination Algorithm (FPA) developed by Xin-She Yang in 2012 mimics the pollination process of flowering plants to efficiently solve such optimization tasks. FPA has shown promising results in extending to multi-o

0 views • 15 slides

Efficiently managing DRAM caches is crucial due to increasing memory demands and bandwidth limitations. Strategies like using DRAM as a cache, architectural considerations for large DRAM caches, and understanding replacement policies are explored in this study to enhance memory bandwidth and capacit

0 views • 23 slides

This research focuses on improving bulk data movement efficiency within DRAM by introducing Low-Cost Inter-Linked Subarrays (LISA). By providing wide connectivity between subarrays, LISA enables fast inter-subarray data transfers, reducing latency and energy consumption. Key applications include fas

0 views • 49 slides

This research presents a hybrid optimization heuristic approach for efficient instruction scheduling in programmable accelerator codesign. It discusses Google's TPU architecture, problem-solving strategies, and computation graph mapping, routing, and timing optimizations. The technique overview high

0 views • 33 slides

This presentation discusses the application of machine learning for optimizing EBIS beam intensity and RHIC luminosity. It covers topics such as motivation, EBIS beam intensity optimization, luminosity optimization, and outlines the plan and summary of the project. Collaborators from MSU, LBNL, and

0 views • 23 slides

Bayesian optimization is being used at LCLS to tune the Free Electron Laser (FEL) pulse energy efficiently. The current approach involves a tradeoff between human optimization and numerical optimization methods, with Gaussian processes providing a probabilistic model for tuning strategies. Prior mea

0 views • 16 slides

The evolution of metalearning in the machine learning community is traced from the initial workshop in 1998 to recent developments in hyper-parameter optimization. Challenges in classifier selection and the validity of hyper-parameter optimization claims are discussed, urging the exploration of spec

1 views • 32 slides

CLR-DRAM introduces a low-cost DRAM architecture that enables dynamic configuration for high capacity or low latency at the granularity of a row. By allowing a single DRAM row to switch between max-capacity and high-performance modes, it reduces key timing parameters, improves system performance, an

0 views • 42 slides

Different memory technologies present unique strengths, and a hybrid memory system combining DRAM and PCM aims to leverage the best of both worlds. This research explores the challenge of data placement between these diverse memory devices, highlighting the use of row buffer locality as a key criter

0 views • 34 slides

Fast Bayesian Optimization optimizes hyperparameters for machine learning on large datasets efficiently. It involves black-box optimization using Gaussian Processes and acquisition functions. Regular Bayesian Optimization faces challenges with large datasets, but FABOLAS introduces an innovative app

0 views • 12 slides

The BEER project explores how on-die ECC complicates DRAM reliability studies by concealing error characteristics. It aims to uncover the unique ECC function of DRAM chips and infer error locations in error-prone cells. The study highlights the challenges in identifying and correcting bit flips obfu

0 views • 17 slides