Transformers have proven effective in language modeling but are limited by high computational and memory demands that grow quadratically with input sequence length. State space models (SSMs) offer a promising alternative by reducing attention complexity from O(L2) to O(L) while also lowering overall memory consumption. Vision Mamba adapts the SSM approach for computer vision tasks, achieving lower latency and memory consumption than traditional transformer models. However, deploying Vision Mamba on edge devices is challenging due to its sequential scan operations, which hinder GPU efficiency. We propose Mamba-X, an end-to-end Vision Mamba accelerator that includes a systolic scan array to maximize parallelism and minimize memory traffic, along with a hybrid, hardware-friendly quantization technique to reduce memory usage and improve hardware efficiency without sacrificing accuracy.

This paper presents a comprehensive evaluation of Intel Gaudi NPUs as an alternative to NVIDIA GPUs, which is currently the de facto standard in AI system design.First, we create microbenchmarks to compare Intel Gaudi-2 with NVIDIA A100, showing that Gaudi-2 achieves competitive performance not only in primitive AI compute, memory, and communication operations but also in executing several important AI workloads end-to-end.We then assess Gaudi NPU's programmability by discussing several softwarelevel optimization strategies to employ for implementing critical FBGEMM operators and vLLM, evaluating their efficiency against GPU-optimized counterparts.Results indicate that Gaudi-2 achieves energy efficiency comparable to A100, though there are notable areas for improvement in terms of software maturity.Overall, we conclude that, with effective integration into high-level AI frameworks, Gaudi NPUs could challenge NVIDIA GPU's dominance in * Both authors contributed equally to this research.

Training recommendation systems (RecSys) faces several challenges as it requires the “data preprocessing” stage to preprocess an ample amount of raw data and feed them to the GPU for training in a seamless manner. To sustain high training throughput, state-of-the-art solutions reserve a large fleet of CPU servers for preprocessing which incurs substantial deployment cost and power consumption. Our characterization reveals that prior CPU-centric preprocessing is bottlenecked on feature generation and feature normalization operations as it fails to reap out the abundant inter-/intra-feature parallelism in RecSys preprocessing. PreSto is a storage-centric preprocessing system leveraging In-Storage Processing (ISP), which offloads the bottlenecked preprocessing operations to our ISP units. We show that PreSto outperforms the baseline CPU-centric system with a 9.6× speedup in end-to-end preprocessing time, 4.3× enhancement in cost-efficiency, and 11.3× improvement in energy-efficiency on average for production-scale RecSys preprocessing.

Graph neural networks (GNNs) can extract features by learning both the representation of each objects (i.e., graph nodes) and the relationship across different objects (i.e., the edges that connect nodes), achieving state-of-the-art performance in various graph-based tasks. Despite its strengths, utilizing these algorithms in a production environment faces several challenges as the number of graph nodes and edges amount to several billions to hundreds of billions scale, requiring substantial storage space for training. Unfortunately, state-of-the-art ML frameworks employ an in-memory processing model which significantly hampers the productivity of ML practitioners as it mandates the overall working set to fit within DRAM capacity. In this work, we first conduct a detailed characterization on a state-of-the-art, large-scale GNN training algorithm, GraphSAGE. Based on the characterization, we then explore the feasibility of utilizing capacity-optimized NVMe SSDs for storing memory-hungry GNN data, which enables large-scale GNN training beyond the limits of main memory size. Given the large performance gap between DRAM and SSD, however, blindly utilizing SSDs as a direct substitute for DRAM leads to significant performance loss. We therefore develop SmartSAGE, our software/hardware co-design based on an in-storage processing (ISP) architecture. Our work demonstrates that an ISP based large-scale GNN training system can achieve both high capacity storage and high performance, opening up opportunities for ML practitioners to train large GNN datasets without being hampered by the physical limitations of main memory size.

A graph engine should possess adaptability to ensure efficient processing despite a variety of graph data and algorithms. In terms of out-of-core graph engines, which exploit a hierarchical memory structure, an adaptive caching scheme is necessary to sustain effectiveness of memory usage. A caching policy selectively stores data likely to be used in the upper-layer memory based on its own expectation about the future workload. However, the graph workload contains a complexity of memory access according to graph data, algorithm, and configurations. This makes it difficult for a static caching policy to respond to the changes in workload. In this paper, we propose a graph-adaptive caching scheme which ensures consistent effectiveness under the changing workloads. Our caching scheme employs an adaptive policy that responds to changes in real-time workloads. To detect the changes, we adopt the competition procedures between two contrasting properties-locality and regularity-that appear in graph workloads. In addition, we combine two window adjustment techniques to alleviate the overhead from competition procedures. The proposed caching scheme is applicable to different types of graph engines, achieving better efficiency in memory usage. Our experimental results prove that our scheme improves the performance of graph processing by up to 65% compared to existing schemes.

Personalized recommendations are one of the most widely deployed machine learning (ML) workload serviced from cloud datacenters. As such, architectural solutions for high-performance recommendation inference have recently been the target of several prior literatures. Unfortunately, little have been explored and understood regarding the training side of this emerging ML workload. In this paper, we first perform a detailed workload characterization study on training recommendations, root-causing sparse embedding layer training as one of the most significant performance bottlenecks. We then propose our algorithm-architecture co-design called Tensor Casting, which enables the development of a generic accelerator architecture for tensor gather-scatter that encompasses all the key primitives of training embedding layers. When prototyped on a real CPUGPU system, Tensor Casting provides 1.9-21× improvements in training throughput compared to state-of-the-art approaches.

Recent studies from several hyperscalars pinpoint to embedding layers as the most memory-intensive deep learning (DL) algorithm being deployed in today's datacenters. This paper addresses the memory capacity and bandwidth challenges of embedding layers and the associated tensor operations. We present our vertically integrated hardware/software co-design, which includes a custom DIMM module enhanced with near-memory processing cores tailored for DL tensor operations. These custom DIMMs are populated inside a GPU-centric system interconnect as a remote memory pool, allowing GPUs to utilize for scalable memory bandwidth and capacity expansion. A prototype implementation of our proposal on real DL systems shows an average 6.2-17.6× performance improvement on state-of-the-art DNN-based recommender systems.