Searcharxiv⌕ Search

arXiv subjects

Onur Mutlu

Publications and source records attributed to Onur Mutlu.

At least 73 records · Page 4Linked to original sources

DRAM Bender: An Extensible and Versatile FPGA-based Infrastructure to Easily Test State-of-the-art DRAM Chips

To understand and improve DRAM performance, reliability, security and energy efficiency, prior works study characteristics of commodity DRAM chips. Unfortunately, state-of-the-art open source infrastructures capable of conducting such studies are obsolete, poorly supported, or difficult to use, or their inflexibility limit the types of studies they can conduct. We propose DRAM Bender, a new FPGA-based infrastructure that enables experimental studies on state-of-the-art DRAM chips. DRAM Bender offers three key features at the same time. First, DRAM Bender enables directly interfacing with a DRAM chip through its low-level interface. This allows users to issue DRAM commands in arbitrary order and with finer-grained time intervals compared to other open source infrastructures. Second, DRAM Bender exposes easy-to-use C++ and Python programming interfaces, allowing users to quickly and easily develop different types of DRAM experiments. Third, DRAM Bender is easily extensible. The modular design of DRAM Bender allows extending it to (i) support existing and emerging DRAM interfaces, and (ii) run on new commercial or custom FPGA boards with little effort. To demonstrate that DRAM Bender is a versatile infrastructure, we conduct three case studies, two of which lead to new observations about the DRAM RowHammer vulnerability. In particular, we show that data patterns supported by DRAM Bender uncovers a larger set of bit-flips on a victim row compared to the data patterns commonly used by prior work. We demonstrate the extensibility of DRAM Bender by implementing it on five different FPGAs with DDR4 and DDR3 support. DRAM Bender is freely and openly available at https://github.com/CMU-SAFARI/DRAM-Bender.

cs.AR↗

ColumnDisturb: Understanding Column-based Read Disturbance in Real DRAM Chips and Implications for Future Systems

We experimentally demonstrate a new widespread read disturbance phenomenon, ColumnDisturb, in real commodity DRAM chips. By repeatedly opening or keeping a DRAM row (aggressor row) open, we show that it is possible to disturb DRAM cells through a DRAM column (i.e., bitline) and induce bitflips in DRAM cells sharing the same columns as the aggressor row (across multiple DRAM subarrays). With ColumnDisturb, the activation of a single row concurrently disturbs cells across as many as three subarrays (e.g., 3072 rows) as opposed to RowHammer/RowPress, which affect only a few neighboring rows of the aggressor row in a single subarray. We rigorously characterize ColumnDisturb and its characteristics under various operational conditions using 216 DDR4 and 4 HBM2 chips from three major manufacturers. Among our 27 key experimental observations, we highlight two major results and their implications. First, ColumnDisturb affects chips from all three major manufacturers and worsens as DRAM technology scales down to smaller node sizes (e.g., the minimum time to induce the first ColumnDisturb bitflip reduces by up to 5.06x). We observe that, in existing DRAM chips, ColumnDisturb induces bitflips within a standard DDR4 refresh window (e.g., in 63.6 ms) in multiple cells. We predict that, as DRAM technology node size reduces, ColumnDisturb would worsen in future DRAM chips, likely causing many more bitflips in the standard refresh window. Second, ColumnDisturb induces bitflips in many (up to 198x) more rows than retention failures. Therefore, ColumnDisturb has strong implications for retention-aware refresh mechanisms that leverage the heterogeneity in cell retention times: our detailed analyses show that ColumnDisturb greatly reduces the benefits of such mechanisms.

cs.AR↗

Understanding and Mitigating Covert Channel and Side Channel Vulnerabilities Introduced by RowHammer Defenses

DRAM chips are vulnerable to read disturbance phenomena (e.g., RowHammer and RowPress), where repeatedly accessing or keeping open a DRAM row causes bitflips in nearby rows. Attackers leverage RowHammer bitflips in real systems to take over systems and leak data. Consequently, many prior works propose defenses, including recent DDR specifications introducing new defenses (e.g., PRAC and RFM). For robust operation, it is critical to analyze other security implications of RowHammer defenses. Unfortunately, no prior work analyzes the timing covert and side channel vulnerabilities introduced by RowHammer defenses. This paper presents the first analysis and evaluation of timing covert and side channel vulnerabilities introduced by state-of-the-art RowHammer defenses. We demonstrate that RowHammer defenses' preventive actions have two fundamental features that enable timing channels. First, preventive actions reduce DRAM bandwidth availability, resulting in longer memory latencies. Second, preventive actions can be triggered on demand depending on memory access patterns. We introduce LeakyHammer, a new class of attacks that leverage the RowHammer defense-induced memory latency differences to establish communication channels and leak secrets. First, we build two covert channel attacks exploiting two state-of-the-art RowHammer defenses, achieving 39.0 Kbps and 48.7 Kbps channel capacity. Second, we demonstrate a website fingerprinting attack that identifies visited websites based on the RowHammer-preventive actions they cause. We propose and evaluate three countermeasures against LeakyHammer. Our results show that fundamentally mitigating LeakyHammer induces large performance overheads in highly RowHammer-vulnerable systems. We believe and hope our work can enable and aid future work on designing better solutions and more robust systems in the presence of such new vulnerabilities.

cs.CR↗

RawBench: A Comprehensive Benchmarking Framework for Raw Nanopore Signal Analysis Techniques

Nanopore sequencing technologies continue to advance rapidly, offering critical benefits such as real-time analysis, the ability to sequence extremely long DNA fragments (up to millions of bases in a single read), and the option to selectively stop sequencing a molecule before completion. Traditionally, the raw electrical signals generated during sequencing are converted into DNA sequences through a process called basecalling, which typically relies on large neural network models. Raw signal analysis has emerged as a promising alternative to these resource-intensive approaches. While attempts have been made to benchmark conventional basecalling methods, existing evaluation frameworks 1) overlook raw signal analysis techniques, 2) lack the flexibility to accommodate new raw signal analysis tools easily, and 3) fail to include the latest improvements in nanopore datasets. Our goal is to provide an extensible benchmarking framework that enables designing and comparing new methods for raw signal analysis. To this end, we introduce RawBench, the first flexible framework for evaluating raw nanopore signal analysis techniques. RawBench provides modular evaluation of three core pipeline components: 1) reference genome encoding (using different pore models), 2) signal encoding (through various segmentation methods), and 3) representation matching (via different data structures). We extensively evaluate raw signal analysis techniques in terms of 1) quality and performance for read mapping, 2) quality and performance for read classification, and 3) quality of raw signal analysis-assisted basecalling. Our evaluations show that raw signal analysis can achieve competitive quality while significantly reducing resource requirements, particularly in settings where real-time processing or edge deployment is necessary.

q-bio.GN↗

A decentralized future for the open-science databases

Continuous and reliable access to curated biological data repositories is indispensable for accelerating rigorous scientific inquiry and fostering reproducible research. Centralized repositories, though widely used, are vulnerable to single points of failure arising from cyberattacks, technical faults, natural disasters, or funding and political uncertainties. This can lead to widespread data unavailability, data loss, integrity compromises, and substantial delays in critical research, ultimately impeding scientific progress. Centralizing essential scientific resources in a single geopolitical or institutional hub is inherently dangerous, as any disruption can paralyze diverse ongoing research. The rapid acceleration of data generation, combined with an increasingly volatile global landscape, necessitates a critical re-evaluation of the sustainability of centralized models. Implementing federated and decentralized architectures presents a compelling and future-oriented pathway to substantially strengthen the resilience of scientific data infrastructures, thereby mitigating vulnerabilities and ensuring the long-term integrity of data. Here, we examine the structural limitations of centralized repositories, evaluate federated and decentralized models, and propose a hybrid framework for resilient, FAIR, and sustainable scientific data stewardship. Such an approach offers a significant reduction in exposure to governance instability, infrastructural fragility, and funding volatility, and also fosters fairness and global accessibility. The future of open science depends on integrating these complementary approaches to establish a globally distributed, economically sustainable, and institutionally robust infrastructure that safeguards scientific data as a public good, further ensuring continued accessibility, interoperability, and preservation for generations to come.

cs.DB↗

Chronus: Understanding and Securing the Cutting-Edge Industry Solutions to DRAM Read Disturbance

We 1) present the first rigorous security, performance, energy, and cost analyses of the state-of-the-art on-DRAM-die read disturbance mitigation method, Per Row Activation Counting (PRAC) and 2) propose Chronus, a new mechanism that addresses PRAC's two major weaknesses. Our analysis shows that PRAC's system performance overhead on benign applications is non-negligible for modern DRAM chips and prohibitively large for future DRAM chips that are more vulnerable to read disturbance. We identify two weaknesses of PRAC that cause these overheads. First, PRAC increases critical DRAM access latency parameters due to the additional time required to increment activation counters. Second, PRAC performs a constant number of preventive refreshes at a time, making it vulnerable to an adversarial access pattern, known as the wave attack, and consequently requiring it to be configured for significantly smaller activation thresholds. To address PRAC's two weaknesses, we propose a new on-DRAM-die RowHammer mitigation mechanism, Chronus. Chronus 1) updates row activation counters concurrently while serving accesses by separating counters from the data and 2) prevents the wave attack by dynamically controlling the number of preventive refreshes performed. Our performance analysis shows that Chronus's system performance overhead is near-zero for modern DRAM chips and very low for future DRAM chips. Chronus outperforms three variants of PRAC and three other state-of-the-art read disturbance solutions. We discuss Chronus's and PRAC's implications for future systems and foreshadow future research directions. To aid future research, we open-source our Chronus implementation at https://github.com/CMU-SAFARI/Chronus.

cs.CR↗

Memory-Centric Computing: Solving Computing's Memory Problem

Computing has a huge memory problem. The memory system, consisting of multiple technologies at different levels, is responsible for most of the energy consumption, performance bottlenecks, robustness problems, monetary cost, and hardware real estate of a modern computing system. All this becomes worse as modern and emerging applications become more data-intensive (as we readily witness in e.g., machine learning, genome analysis, graph processing, and data analytics), making the memory system an even larger bottleneck. In this paper, we discuss two major challenges that greatly affect computing system performance and efficiency: 1) memory technology & capacity scaling (at the lower device and circuit levels) and 2) system and application performance & energy scaling (at the higher levels of the computing stack). We demonstrate that both types of scaling have become extremely difficult, wasteful, and costly due to the dominant processor-centric design & execution paradigm of computers, which treats memory as a dumb and inactive component that cannot perform any computation. We show that moving to a memory-centric design & execution paradigm can solve the major challenges, while enabling multiple other potential benefits. In particular, we demonstrate that: 1) memory technology scaling problems (e.g., RowHammer, RowPress, Variable Read Disturbance, data retention, and other issues awaiting to be discovered) can be much more easily and efficiently handled by enabling memory to autonomously manage itself; 2) system and application performance & energy efficiency can, at the same time, be improved by orders of magnitude by enabling computation capability in memory chips and structures (i.e., processing in memory). We discuss adoption challenges against enabling memory-centric computing, and describe how we can get there step-by-step via an evolutionary path.

cs.AR↗

Self-Managing DRAM: A Low-Cost Framework for Enabling Autonomous and Efficient in-DRAM Operations

The memory controller is in charge of managing DRAM maintenance operations (e.g., refresh, RowHammer protection, memory scrubbing) to reliably operate modern DRAM chips. Implementing new maintenance operations often necessitates modifications in the DRAM interface, memory controller, and potentially other system components. Such modifications are only possible with a new DRAM standard, which takes a long time to develop, likely leading to slow progress in the adoption of new architectural techniques in DRAM chips. We propose a new low-cost DRAM architecture, Self-Managing DRAM (SMD), that enables autonomous in-DRAM maintenance operations by transferring the responsibility for controlling maintenance operations from the memory controller to the SMD chip. To enable autonomous maintenance operations, we make a single modification to the DRAM interface, such that an SMD chip rejects memory controller accesses to DRAM regions under maintenance, while allowing memory accesses to others. Thus, SMD enables 1) implementing new in-DRAM maintenance mechanisms (or modifying existing ones) with no further changes in the DRAM interface or other system components, and 2) overlapping the latency of a maintenance operation in one DRAM region with the latency of accessing data in another. We evaluate SMD and show that it 1) can be implemented without adding new pins to the DDRx interface with low latency and area overhead, 2) achieves 4.1% average speedup across 20 four-core memory-intensive workloads over a DDR4-based system/DRAM co-design technique that intelligently parallelizes maintenance operations with memory accesses, and 3) guarantees forward progress for rejected memory accesses. We believe and hope SMD can enable innovations in DRAM architecture to rapidly come to fruition. We open source all SMD source code and data at https://github.com/CMU-SAFARI/SelfManagingDRAM.

cs.AR↗

MARS: Processing-In-Memory Acceleration of Raw Signal Genome Analysis Inside the Storage Subsystem

Raw signal genome analysis (RSGA) has emerged as a promising approach to enable real-time genome analysis by directly analyzing raw electrical signals. However, rapid advancements in sequencing technologies make it increasingly difficult for software-based RSGA to match the throughput of raw signal generation. This paper demonstrates that while hardware acceleration techniques can significantly accelerate RSGA, the high volume of genomic data shifts the performance and energy bottleneck from computation to I/O data movement. As sequencing throughput increases, I/O overhead becomes the main contributor to both runtime and energy consumption. Therefore, there is a need to design a high-performance, energy-efficient system for RSGA that can both alleviate the data movement bottleneck and provide large acceleration capabilities. We propose MARS, a storage-centric system that leverages the heterogeneous resources within modern storage systems (e.g., storage-internal DRAM, storage controller, flash chips) alongside their large storage capacity to tackle both data movement and computational overheads of RSGA in an area-efficient and low-cost manner. MARS accelerates RSGA through a novel hardware/software co-design approach. First, MARS modifies the RSGA pipeline via two filtering mechanisms and a quantization scheme, reducing hardware demands and optimizing for in-storage execution. Second, MARS accelerates the RSGA steps directly within the storage by leveraging both Processing-Near-Memory and Processing-Using-Memory paradigms. Third, MARS orchestrates the execution of all steps to fully exploit in-storage parallelism and minimize data movement. Our evaluation shows that MARS outperforms basecalling-based software and hardware-accelerated state-of-the-art read mapping pipelines by 93x and 40x, on average across different datasets, while reducing their energy consumption by 427x and 72x.

cs.AR↗

EasyDRAM: An FPGA-based Infrastructure for Fast and Accurate End-to-End Evaluation of Emerging DRAM Techniques

DRAM is a critical component of modern computing systems. Recent works propose numerous techniques (that we call DRAM techniques) to enhance DRAM-based computing systems' throughput, reliability, and computing capabilities (e.g., in-DRAM bulk data copy). Evaluating the system-wide benefits of DRAM techniques is challenging as they often require modifications across multiple layers of the computing stack. Prior works propose FPGA-based platforms for rapid end-to-end evaluation of DRAM techniques on real DRAM chips. Unfortunately, existing platforms fall short in two major aspects: (1) they require deep expertise in hardware description languages, limiting accessibility; and (2) they are not designed to accurately model modern computing systems. We introduce EasyDRAM, an FPGA-based framework for rapid and accurate end-to-end evaluation of DRAM techniques on real DRAM chips. EasyDRAM overcomes the main drawbacks of prior FPGA-based platforms with two key ideas. First, EasyDRAM removes the need for hardware description language expertise by enabling developers to implement DRAM techniques using a high-level language (C++). At runtime, EasyDRAM executes the software-defined memory system design in a programmable memory controller. Second, EasyDRAM tackles a fundamental challenge in accurately modeling modern systems: real processors typically operate at higher clock frequencies than DRAM, a disparity that is difficult to replicate on FPGA platforms. EasyDRAM addresses this challenge by decoupling the processor-DRAM interface and advancing the system state using a novel technique we call time scaling, which faithfully captures the timing behavior of the modeled system. We believe and hope that EasyDRAM will enable innovative ideas in memory system design to rapidly come to fruition. To aid future research EasyDRAM implementation is open sourced at https://github.com/CMU-SAFARI/EasyDRAM.

cs.AR↗

PuDHammer: Experimental Analysis of Read Disturbance Effects of Processing-using-DRAM in Real DRAM Chips

Processing-using-DRAM (PuD) is a promising paradigm for alleviating the data movement bottleneck using DRAM's massive internal parallelism and bandwidth to execute very wide operations. Performing a PuD operation involves activating multiple DRAM rows in quick succession or simultaneously, i.e., multiple-row activation. Multiple-row activation is fundamentally different from conventional memory access patterns that activate one DRAM row at a time. However, repeatedly activating even one DRAM row (e.g., RowHammer) can induce bitflips in unaccessed DRAM rows because modern DRAM is subject to read disturbance. Unfortunately, no prior work investigates the effects of multiple-row activation on DRAM read disturbance. In this paper, we present the first characterization study of read disturbance effects of multiple-row activation-based PuD (which we call PuDHammer) using 316 real DDR4 DRAM chips from four major DRAM manufacturers. Our detailed characterization show that 1) PuDHammer significantly exacerbates the read disturbance vulnerability, causing up to 158.58x reduction in the minimum hammer count required to induce the first bitflip ($HC_{first}$), compared to RowHammer, 2) PuDHammer is affected by various operational conditions and parameters, 3) combining RowHammer with PuDHammer is more effective than using RowHammer alone to induce read disturbance error, e.g., doing so reduces $HC_{first}$ by 1.66x on average, and 4) PuDHammer bypasses an in-DRAM RowHammer mitigation mechanism (Target Row Refresh) and induces more bitflips than RowHammer. To develop future robust PuD-enabled systems in the presence of PuDHammer, we 1) develop three countermeasures and 2) adapt and evaluate the state-of-the-art RowHammer mitigation standardized by industry, called Per Row Activation Counting (PRAC). We show that the adapted PRAC incurs large performance overheads (48.26%, on average).

cs.AR↗

Revisiting Main Memory-Based Covert and Side Channel Attacks in the Context of Processing-in-Memory

We introduce IMPACT, a set of high-throughput main memory-based timing attacks that leverage characteristics of processing-in-memory (PiM) architectures to establish covert and side channels. IMPACT enables high-throughput communication and private information leakage by exploiting the shared DRAM row buffer. To achieve high throughput, IMPACT (i) eliminates expensive cache bypassing steps required by processor-centric memory-based timing attacks and (ii) leverages the intrinsic parallelism of PiM operations. We showcase two applications of IMPACT. First, we build two covert channels that leverage different PiM approaches (i.e., processing-near-memory and processing-using-memory) to establish high-throughput covert communication channels. Our covert channels achieve 8.2 Mb/s and 14.8 Mb/s communication throughput, respectively, which is 3.6x and 6.5x higher than the state-of-the-art main memory-based covert channel. Second, we showcase a side-channel attack that leaks private information of concurrently-running victim applications with a low error rate. Our source-code is openly and freely available at https://github.com/CMU-SAFARI/IMPACT.

cs.CR↗

Proteus: Enabling High-Performance Processing-Using-DRAM with Dynamic Bit-Precision, Adaptive Data Representation, and Flexible Arithmetic

Processing-using-DRAM (PUD) is a paradigm where the analog operational properties of DRAM are used to perform bulk logic operations. While PUD promises high throughput at low energy and area cost, we uncover three limitations of existing PUD approaches that lead to significant inefficiencies: (i) static data representation, i.e., two's complement with fixed bit-precision, leading to unnecessary computation over useless (i.e., inconsequential) data; (ii) support for only throughput-oriented execution, where the high latency of individual PUD operations can only be hidden in the presence of bulk data-level parallelism; and (iii) high latency for high-precision (e.g., 32-bit) operations. To address these issues, we propose Proteus, the first hardware framework that addresses the high execution latency of bulk bitwise PUD operations by implementing a data-aware runtime engine for PUD. Proteus reduces the latency of PUD operations in three different ways: (i) Proteus dynamically reduces the bit-precision (and thus the latency and energy consumption) of PUD operations by exploiting narrow values (i.e., values with many leading zeros or ones); (ii) Proteus concurrently executes independent in-DRAM primitives belonging to a single PUD operation across multiple DRAM arrays; (iii) Proteus chooses and uses the most appropriate data representation and arithmetic algorithm implementation for a given PUD instruction transparently to the programmer.

cs.AR↗

PIM Is All You Need: A CXL-Enabled GPU-Free System for Large Language Model Inference

Large Language Model (LLM) inference uses an autoregressive manner to generate one token at a time, which exhibits notably lower operational intensity compared to earlier Machine Learning (ML) models such as encoder-only transformers and Convolutional Neural Networks. At the same time, LLMs possess large parameter sizes and use key-value caches to store context information. Modern LLMs support context windows with up to 1 million tokens to generate versatile text, audio, and video content. A large key-value cache unique to each prompt requires a large memory capacity, limiting the inference batch size. Both low operational intensity and limited batch size necessitate a high memory bandwidth. However, contemporary hardware systems for ML model deployment, such as GPUs and TPUs, are primarily optimized for compute throughput. This mismatch challenges the efficient deployment of advanced LLMs and makes users pay for expensive compute resources that are poorly utilized for the memory-bound LLM inference tasks. We propose CENT, a CXL-ENabled GPU-Free sysTem for LLM inference, which harnesses CXL memory expansion capabilities to accommodate substantial LLM sizes, and utilizes near-bank processing units to deliver high memory bandwidth, eliminating the need for expensive GPUs. CENT exploits a scalable CXL network to support peer-to-peer and collective communication primitives across CXL devices. We implement various parallelism strategies to distribute LLMs across these devices. Compared to GPU baselines with maximum supported batch sizes and similar average power, CENT achieves 2.3$\times$ higher throughput and consumes 2.9$\times$ less energy. CENT enhances the Total Cost of Ownership (TCO), generating 5.2$\times$ more tokens per dollar than GPUs.

cs.AR↗

Revisiting DRAM Read Disturbance: Identifying Inconsistencies Between Experimental Characterization and Device-Level Studies

Modern DRAM is vulnerable to read disturbance (e.g., RowHammer and RowPress) that significantly undermines the robust operation of the system. Repeatedly opening and closing a DRAM row (RowHammer) or keeping a DRAM row open for a long period of time (RowPress) induces bitflips in nearby unaccessed DRAM rows. Prior works on DRAM read disturbance either 1) perform experimental characterization using commercial-off-the-shelf (COTS) DRAM chips to demonstrate the high-level characteristics of the read disturbance bitflips, or 2) perform device-level simulations to understand the low-level error mechanisms of the read disturbance bitflips. In this paper, we attempt to align and cross-validate the real-chip experimental characterization results and state-of-the-art device-level studies of DRAM read disturbance. To do so, we first identify and extract the key bitflip characteristics of RowHammer and RowPress from the device-level error mechanisms studied in prior works. Then, we perform experimental characterization on 96 COTS DDR4 DRAM chips that directly match the data and access patterns studied in the device-level works. Through our experiments, we identify fundamental inconsistencies in the RowHammer and RowPress bitflip directions and access pattern dependence between experimental characterization results and the device-level error mechanisms. Based on our results, we hypothesize that either 1) the retention failure based DRAM architecture reverse-engineering methodologies do not fully work on modern DDR4 DRAM chips, or 2) existing device-level works do not fully uncover all the major read disturbance error mechanisms. We hope our findings inspire and enable future works to build a more fundamental and comprehensive understanding of DRAM read disturbance.

cs.AR↗

CiMBA: Accelerating Genome Sequencing through On-Device Basecalling via Compute-in-Memory

As genome sequencing is finding utility in a wide variety of domains beyond the confines of traditional medical settings, its computational pipeline faces two significant challenges. First, the creation of up to 0.5 GB of data per minute imposes substantial communication and storage overheads. Second, the sequencing pipeline is bottlenecked at the basecalling step, consuming >40% of genome analysis time. A range of proposals have attempted to address these challenges, with limited success. We propose to address these challenges with a Compute-in-Memory Basecalling Accelerator (CiMBA), the first embedded ($\sim25$mm$^2$) accelerator capable of real-time, on-device basecalling, coupled with AnaLog (AL)-Dorado, a new family of analog focused basecalling DNNs. Our resulting hardware/software co-design greatly reduces data communication overhead, is capable of a throughput of 4.77 million bases per second, 24x that required for real-time operation, and achieves 17x/27x power/area efficiency over the best prior basecalling embedded accelerator while maintaining a high accuracy comparable to state-of-the-art software basecallers.

cs.AR↗

PyGim: An Efficient Graph Neural Network Library for Real Processing-In-Memory Architectures

Graph Neural Networks (GNNs) are emerging ML models to analyze graph-structure data. Graph Neural Network (GNN) execution involves both compute-intensive and memory-intensive kernels, the latter dominates the total time, being significantly bottlenecked by data movement between memory and processors. Processing-In-Memory (PIM) systems can alleviate this data movement bottleneck by placing simple processors near or inside to memory arrays. In this work, we introduce PyGim, an efficient ML library that accelerates GNNs on real PIM systems. We propose intelligent parallelization techniques for memory-intensive kernels of GNNs tailored for real PIM systems, and develop handy Python API for them. We provide hybrid GNN execution, in which the compute-intensive and memory-intensive kernels are executed in processor-centric and memory-centric computing systems, respectively. We extensively evaluate PyGim on a real-world PIM system with 1992 PIM cores using emerging GNN models, and demonstrate that it outperforms its state-of-the-art CPU counterpart on Intel Xeon by on average 3.04x, and achieves higher resource utilization than CPU and GPU systems. Our work provides useful recommendations for software, system and hardware designers. PyGim is publicly available at https://github.com/CMU-SAFARI/PyGim.

cs.AR↗

PIMDAL: Mitigating the Memory Bottleneck in Data Analytics using a Real Processing-in-Memory System

Database Management Systems (DBMSs) are crucial for efficient data management and analytics, and are used in several different application domains. Due to the increasing volume of data a DBMS deals with, current processor-centric architectures (e.g., CPUs, GPUs) suffer from data movement bottlenecks when executing key DBMS operations (e.g., selection, aggregation, ordering, and join). This happens mostly due to the limited memory bandwidth between compute and memory resources. Data-centric architectures like Processing-in-Memory (PIM) are a promising alternative for applications bottlenecked by data, placing compute resources close to where data resides. Previous works have evaluated using PIM for data analytics. However, they either do not use real-world architectures or they consider only a subset of the operators used in analytical queries. This work aims to fully evaluate a data-centric approach to data analytics, by using the real-world UPMEM PIM system. To this end we first present the PIM Data Analytics Library (PIMDAL), which implements four major DB operators: selection, aggregation, ordering and join. Second, we use hardware performance metrics to understand which properties of a PIM system are important for a high-performance implementation. Third, we compare PIMDAL to reference implementations on high-end CPU and GPU systems. Fourth, we use PIMDAL to implement five TPC-H queries to gain insights into analytical queries. We analyze and show how to overcome the three main limitations of the UPMEM system when implementing DB operators: (I) low arithmetic performance, (II) explicit memory management and (III) limited communication between compute units. Our evaluation shows PIMDAL achieves 3.9x the performance of a high-end CPU, on average across the five TPC-H queries.

cs.AR↗