High Bandwidth Memory with Processing-in-Memory (HBM-PIM) offers an opportunity to reduce data movement by executing computation directly inside memory, but current commercial platforms expose limited instruction sets and require specialized software stacks. In this work, we investigate whether HBM-PIM can serve as a backend for ISA-level matrix acceleration, using the RISC-V Attached Matrix Extension (AME) as a semantic reference. We propose a PEP-based execution model that maps AME element-wise and matrix instructions to HBM-PIM micro-kernels and data instructions in memory operations. Differently from SoA HBM-PIM, we introduce a reduction-free outer-product dataflow that enables accumulation entirely within memory despite the lack of native reduction support. Our approach supports end-to-end execution of element-wise operations, GEMV, and GEMM in PIM mode, minimizing host involvement and off-chip transfers. An experimental evaluation on Samsung Aquabolt-XL shows that AME matrix tile multiplication achieves up to 14.9 GFLOP/s (59.4 FLOP/cycle) on a single HBM pseudo-channel.

CXL has emerged as a promising memory interface that can cost-effectively expand the capacity and bandwidth of a memory system, complementing the traditional DDR interface. However, CXL DRAM presents 2-3x longer access latency than DDR DRAM, forming a tiered-memory system that demands an effective and efficient page-migration solution. Although many page-migration solutions have been proposed for past tiered-memory systems, they have achieved limited success. To tackle the challenge of managing tiered-memory systems, this work first presents a CXL-driven profiling solution to precisely and transparently count the number of accesses to every 4KB page and 64B word in CXL DRAM. Second, using the profiling solution, this work uncovers that (1) widely used CPU-driven page-migration solutions often identify warm pages as hot pages, and (2) certain applications have sparse hot pages, where only a small percentage of words in each of these pages are frequently accessed. Besides, this work demonstrates that the performance overhead of identifying hot pages is sometimes high enough to degrade application performance. Lastly, this work presents M5, a platform designed to facilitate the development of effective CXL-driven page-migration solutions, providing hardware-based hot-page and hot-word trackers in the CXL controller. On average, M5 can identify 47% hotter pages and offer 14% higher performance than the best CPU-driven page-migration solution, even with a simple policy.

The Transformer-based generative model (TbGM), comprising summarization (Sum) and generation (Gen) stages, has demonstrated unprecedented generative performance across a wide range of applications. However, it also demands immense amounts of compute and memory resources. Especially, the Gen stages, consisting of the attention and fully-connected (FC) layers, dominate the overall execution time. Meanwhile, we reveal that the conventional system with GPUs used for TbGM inference cannot efficiently execute the attention layer, even with batching, due to various constraints. To address this inefficiency, we first propose AttAcc, a processing-in-memory (PIM) architecture for efficient execution of the attention layer. Subsequently, for the end-to-end acceleration of TbGM inference, we propose a novel heterogeneous system architecture and optimizations that strategically use xPU and PIM together. It leverages the high memory bandwidth of AttAcc for the attention layer and the powerful compute capability of the conventional system for the FC layer. Lastly, we demonstrate that our GPU-PIM system outperforms the conventional system with the same memory capacity, improving performance and energy efficiency of running a 175B TbGM by up to 2.81× and 2.67×, respectively.

Row Hammer (RH) has been demonstrated as a security vulnerability in modern systems. Although commodity CPUs can handle RH-induced single-bit errors in DRAM through ECC, RH can still give rise to multi-bit uncorrectable errors (UEs) and crash the systems. Meanwhile, recent work has indicated that the DRAM cells vulnerable to RH are determined by manufacturing imperfections and resulting defects. Taking one step further from the recent work, we first conduct RH experiments on contemporary DRAM modules for 3 weeks. This demonstrates that RH-induced UEs occur only at specific DRAM addresses (RH-UE-vulnerable addresses) and the percentage of such addresses is small in these DRAM modules. Second, to protect the systems from RH-induced UEs, we propose two RH defense solutions: H- and S-TAROT (TArgeted ROw-Hammer Therapy). H-TAROT is a software-based solution running on the host CPU. It obtains RH-UE-vulnerable addresses during the system boot and then periodically accesses such addresses before UEs may occur. Since it accesses only a small percentage of addresses, it does not incur a notable performance penalty for throughput applications (e.g., a 1.5% increase in execution time of the SPECrate 2017 benchmark suite running on a system even with 128GB of DRAM). Yet, it imposes a significant performance penalty on latency-sensitive applications (e.g., a 28.2% increase in tail latency of Redis). To minimize the performance penalty, for a system with a SmartNIC (SNIC), S-TAROT offloads H-TAROT from the host CPU to the SNIC CPU. Our experiment shows that S-TAROT increases the execution time and tail latency of the SPECrate 2017 benchmark suite and Redis by only 0.1% and 1.0%, respectively.

Transformer-based generative models, such as GPT, summarize an input sequence by generating key/value (KV) matrices through attention and generate the corresponding output sequence by utilizing these matrices once per token of the sequence. Both input and output sequences tend to get longer, which improves the understanding of contexts and conversation quality. These models are also typically batched for inference to improve the serving throughput. All these trends enable the models' weights to be reused effectively, increasing the relative importance of sequence generation, especially in processing KV matrices through attention. We identify that the conventional computing platforms (e.g., GPUs) are not efficient at handling this attention part for inference because each request generates different KV matrices, it has a low operation per byte ratio regardless of the batch size, and the aggregate size of the KV matrices can even surpass that of the entire model weights. This motivates us to propose AttAcc, which exploits the fact that the KV matrices are written once during summarization but used many times (proportional to the output sequence length), each multiplied by the embedding vector corresponding to an output token. The volume of data entering/leaving AttAcc could be more than orders of magnitude smaller than what should be read internally for attention. We design AttAcc with multiple processing-in-memory devices, each multiplying the embedding vector with the portion of the KV matrices within the devices, saving external (inter-device) bandwidth and energy consumption.

An embedding layer is one of the most critical building blocks of deep neural networks, especially for recommender systems and graph neural networks. The embedding layer dominates a large portion of the total execution time due to its large memory requirements and little data reuse in operations. To accelerate the embedding layers, dual in-line memory module (DIMM) based near-data processing architectures have been proposed. They amplify bandwidth by adding a processing unit to the DIMM’s buffer. However, prior architectures have less capacity scalability due to the limited number of memory channels. Crucially, they are limited in performance improvement due to the load imbalance problem and the limitations of DIMM-based memory systems with a multi-drop bus structure between the processing units and the host.

The demand for precise information on DRAM microarchitectures and error characteristics has surged, driven by the need to explore processing in memory, enhance reliability, and mitigate security vulnerability. Nonetheless, DRAM manufacturers have disclosed only a limited amount of information, making it difficult to find specific information on their DRAM microarchitectures. This paper addresses this gap by presenting more rigorous findings on the microarchitectures of commodity DRAM chips and their impacts on the characteristics of activate-induced bitflips (AIBs), such as RowHammer and RowPress. The previous studies have also attempted to understand the DRAM microarchitectures and associated behaviors, but we have found some of their results to be misled by inaccurate address mapping and internal data swizzling, or lack of a deeper understanding of the modern DRAM cell structure. For accurate and efficient reverse-engineering, we use three tools: AIBs, retention time test, and RowCopy, which can be cross-validated. With these three tools, we first take a macroscopic view of modern DRAM chips to uncover the size, structure, and operation of their subarrays, memory array tiles (MATs), and rows. Then, we analyze AIB characteristics based on the microscopic view of the DRAM microarchitecture, such as 6F2cell layout, through which we rectify misunderstandings regarding AIBs and discover a new data pattern that accelerates AIBs. Lastly, based on our findings at both macroscopic and microscopic levels, we identify previously unknown AIB vulnerabilities and propose a simple yet effective protection solution.

Graph Neural Network (GNN) models have attracted attention, given their high accuracy in interpreting graph data. One of the primary building blocks of a GNN model is aggregation, which gathers and averages the feature vectors corresponding to the nodes adjacent to each node. Aggregation works by multiplying the adjacency and feature matrices. The size of both matrices exceeds the on-chip cache capacity for many realistic datasets, and the adjacency matrix is highly sparse. These characteristics lead to little data reuse, causing intensive main-memory accesses during the aggregation process. Thus, aggregation exhibits memory-intensive characteristics and dominates most of the total execution time. In this paper, we propose GraNDe, an NDP architecture that accelerates memory-intensive aggregation operations by locating NDP modules near DRAM datapath to exploit rank-level parallelism. GraNDe maximizes bandwidth utilization by separating the memory channel path with the buffer chip in between so that pre-/post-processing in the host processor and reduction in NDP modules operate simultaneously. By exploring the preferred data mappings of the operand matrices to DRAM ranks, we architect GraNDe to support adaptive matrix mapping that applies the optimal mapping for each layer depending on the dimension of the layer and the configuration of a memory system. We also propose adj-bundle broadcasting and re-tiling optimizations to reduce the transfer time for adjacency matrix data and to improve feature vector data reusability by exploiting tiling with consideration of adjacency between nodes. GraNDe achieves 3.01× and 1.69× on average, and up to$4.00\times$and$1.98\times$speedups of GCN aggregation over the baseline system and the state-of-the-art NDP architecture for GCN, respectively.

As Row Hammer (RH) attacks have been a critical threat to computer systems, numerous hardware-based (HWbased) RH mitigation strategies have been proposed. However, the advent of non-adjacent RH attacks and lower RH thresholds significantly increase the area and performance overhead of these prior solutions due to their conservative design characteristics.We propose a new in-DRAM RH protection solution named Shuffling Aggressor DRAM Rows (SHADOW). SHADOW dynamically randomizes DRAM row mapping information, preventing an attacker from targeting a specific victim row that may hold critical data. SHADOW is robust against non-adjacent RH attacks because it utilizes the in-DRAM row-shuffle technique. To realize the in-DRAM row-shuffle operation with low performance and energy overhead, we use a novel DRAM microarchitecture optimization technique. We also utilize the recently introduced JEDEC RFM interface to enable in-DRAM RH mitigation without any DRAM interface changes. By exploiting an additional DRAM row per subarray, SHADOW does not require costly SRAM- or CAM-based tracking structures other than intrinsic counters for the RFM interface. We demonstrate the strong probabilistic protection of SHADOW against RH attacks through adversarial pattern analysis and highlight the compelling performance, area, and energy overheads compared to those of state-of-the-art HW-based RH prevention solutions.

Error-correcting code (ECC) has been widely used in DRAM-based memory systems to address the exacerbating random errors following the fabrication process scaling. However, ECCs including the strong form of Chipkill have not been so effective against Row Hammer (RH), which incurs bursts of errors discretely corrupting the whole row beyond the ECC correction capability.

Since its public introduction in the mid-2010s, the Row Hammer (RH) phenomenon has drawn significant attention from the research community due to its security implications. Although many RH-protection schemes have been proposed by processor vendors, DRAM manufacturers, and academia, they still have shortcomings. Solutions implemented in the memory controller (MC) incur increasingly higher costs due to their conservative design for the worst case in terms of the number of DRAM banks and RH threshold to support. Meanwhile, DRAM-side implementation either has a limited time margin for RH-protection measures or requires extensive modifications to the standard DRAM interface. Recently, a new command for RH-protection has been introduced in the DDR5/LPDDR5 standards, referred to as refresh management (RFM). RFM enables the separation of the tasks for RH-protection to both MC and DRAM by having the former generate an RFM command at a specific activation frequency and the latter take proper RH-protection measures within a given time window. Although promising, no existing study presents and analyzes RFM-based solutions for RH-protection. In this paper, we propose Mithril, the first RFM interface-compatible, DRAM-MC cooperative RH-protection scheme providing deterministic protection guarantees. Mithril has minimal energy overheads for common use cases without adversarial memory access patterns. We also introduce Mithril+, an optional extension to provide minimal performance overheads at the expense of a tiny modification to the MC, while utilizing existing DRAM commands.

Mobile and edge devices become common platforms for inferring convolutional neural networks (CNNs) due to superior privacy and service quality. To reduce the computational costs of convolution (CONV) , recent CNN models adopt depth-wise CONV (DW-CONV) and Squeeze-and-Excitation (SE) . However, existing area-efficient CNN accelerators are sub-optimal for these latest CNN models because they were mainly optimized for compute-intensive standard CONV layers with abundant data reuse that can be pipelined with activation and normalization operations. In contrast, DW-CONV and SE are memory-intensive with limited data reuse. The latter also strongly depends on the nearby CONV layers, making an effective pipelining a daunting task. Therefore, DW-CONV and SE only occupy 10% of entire operations but become memory bandwidth bound, spending more than 60% of the processing time in systolic-array-based accelerators. We propose a CNN acceleration architecture called MVP, which efficiently processes both compute- and memory-intensive operations with a small area overhead on top of the baseline systolic-array-based architecture. We suggest a specialized vector unit tailored for processing DW-CONV, including multipliers, adder trees, and multi-banked buffers to meet the high memory bandwidth requirement. We augment the unified buffer with tiny processing elements to smoothly pipeline SE with the subsequent CONV, enabling concurrent processing of DW-CONV with standard CONV, thereby achieving the maximum utilization of arithmetic units. Our evaluation shows that MVP improves performance by 2.6 \( \times \) and reduces energy by 47% on average for EfficientNet-B0/B4/B7, MnasNet, and MobileNet-V1/V2 with only a 9% area overhead compared to the baseline.

Personalized recommendation systems are gaining significant traction due to their industrial importance. An important building block of recommendation systems consists of the embedding layers, which exhibit a highly memory-intensive characteristic. A fundamental primitive of embedding layers is the embedding vector gathers followed by vector reductions, exhibiting low arithmetic intensity and becoming bottlenecked by the memory throughput. To tackle such a challenge, recent proposals employ a near-data processing (NDP) solution at the DRAM rank-level, achieving impressive performance speedups. We observe that prior rank-level-parallelism-based NDP solutions leave significant performance potential on the table as they do not fully reap the abundant transfer throughput inherent in DRAM datapaths.

Recurrent Neural Networks (RNNs) spend most of their execution time performing matrix-vector multiplication (MV-mul). Because the matrices in RNNs have poor reusability and the ever-increasing size of the matrices becomes too large to fit in the on-chip storage of mobile/IoT devices, the performance and energy efficiency of MV-mul is determined by those of main-memory DRAM. Therefore, computing MV-mul within DRAM draws much attention. However, previous studies lacked consideration for the matrix sparsity, the power constraints of DRAM devices, and concurrency in accessing DRAM from processors while performing MV-mul. We propose a main-memory architecture called MViD, which performs MV-mul by placing MAC units inside DRAM banks. For higher computational efficiency, we use a sparse matrix format and exploit quantization. Because of the limited power budget for DRAM devices, we implement the MAC units only on a portion of the DRAM banks. We architect MViD to slow down or pause MV-mul for concurrently processing memory requests from processors while satisfying the limited power budget. Our results show that MViD provides 7.2× higher throughput compared to the baseline system with four DRAM ranks (performing MV-mul in a chip-multiprocessor) while running inference of Deep Speech 2 with a memory-intensive workload.