BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:America/Los_Angeles X-LIC-LOCATION:America/Los_Angeles BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20240626T180002Z LOCATION:3001\, 3rd Floor DTSTART;TZID=America/Los_Angeles:20240626T153000 DTEND;TZID=America/Los_Angeles:20240626T173000 UID:dac_DAC 2024_sess161@linklings.com SUMMARY:Modeling, Software, and Architecture Just Got Smarter DESCRIPTION:Research Manuscript\n\nThis section presents new developments in embedded modeling, software and architecture. The first paper develops a deep-to-shallow transformable neural architecture search; the second one focuses on efficient code generation for data-intensive Simulink models; the third paper presents a framework of translating regular expressions. T he fourth paper investigates a dynamic mechanism for reducing control flow divergence in GPUs. The next paper investigates methods to boost the effi ciency of data deduplication; the sixth one develops code optimization sch emes to accelerate stencil code kernels; the seventh paper investigates a novel approach to decompose tasks on intermittent computing systems; and t he last one proposes a light-weighted idle time stealing strategy.\n\nALVE ARE: a Domain-Specific Framework for Regular Expressions\n\nRegular Expres sion (RE) matching enables the identification of patterns in datastream of heterogeneous fields ranging from proteomics to computer security. These scenarios require massive data analysis that, combined with the high data dependency of the REs, leads to long computational times and hig...\n\n\nF ilippo Carloni, Davide Conficconi, and Marco D. Santambrogio (Politecnico di Milano)\n---------------------\nHow to Steal CPU Idle Time When Synchro nous I/O Mode Becomes Promising\n\nThe advent of ultra-low-latency storage devices has narrowed the performance gap between storage and CPU in compu ting platforms, facilitating synchronous I/O adoption. Yet, this approach introduces substantial busy waiting time and underutilizes computing units . To address this, we propose a light-w...\n\n\nChun-Feng Wu (National Yan g Ming Chiao Tung University), Yuan-Hao Chang (Academia Sinica), Ming-Chan g Yang (The Chinese University of Hong Kong), and Tei-Wei Kuo (National Ta iwan University)\n---------------------\nDouble-Win NAS: Towards Deep-to-S hallow Transformable Neural Architecture Search for Intelligent Embedded S ystems\n\nThanks to the evolving network depth, convolutional neural netwo rks (CNNs) have achieved impressive performance across various intelligent embedded scenarios towards embedded intelligence. Nonetheless, this trend also leads to degraded hardware efficiency as the network evolves deeper and deeper. In...\n\n\nXiangzhong Luo (Nanyang Technological University); Di Liu (Norwegian University of Science and Technology); and Hao Kong, Shu o Huai, and Weichen Liu (Nanyang Technological University)\n-------------- -------\nControl Flow Divergence Optimization by Exploiting Tensor Cores\n \nKernels are scheduled on Graphics Processing Units (GPUs) in the granula rity of warp, a bunch of concurrently executing threads. When executing ke rnels with conditional branches, threads within a warp may execute differe nt branches sequentially, resulting in a considerable utilization loss and unpre...\n\n\nWeiguang Pang (Qilu University of Technology), Xu Jiang (Un iversity of Electronic Science and Technology of China), Songran Liu (Nort heastern University), Lei Qiao (Beijing Institute of Control Engineering), kexue fu and longxiang Gao (Qilu University of Technology), and Wang Yi ( Uppsala University)\n---------------------\nFinerDedup: Sifting Fingerprin ts for Efficient Data Deduplication on Mobile Devices\n\nData deduplicatio n is promised to extend the lifetime and capacity of storage on mobile dev ices. However, existing data deduplication works show high memory consumpt ion and indexing costs for\nmaintaining a fingerprint for each data block, especially when the duplicate ratio of data blocks on mobile ...\n\n\nXia nzhang Chen, Xingjie Zhou, Wei Li, Xi Yu, Duo Liu, Yujuan Tan, and Ao Ren (Chongqing University)\n---------------------\nSARIS: Accelerating Stencil Computations on Energy-Efficient RISC-V Compute Clusters with Indirect St ream Registers\n\nStencil codes are performance-critical in many compute-i ntensive applications, but suffer from significant address calculation and irregular memory access overheads. This work presents SARIS, a general an d highly flexible methodology for stencil acceleration using register-mapp ed indirect streams. W...\n\n\nPaul Scheffler and Luca Colagrande (ETH Zür ich) and Luca Benini (Università di Bologna)\n---------------------\nEffic ient Code Generation for Data-Intensive Simulink Models via Redundancy Eli mination\n\nSimulink has emerged as the fundamental infrastructure that su pports modeling, simulation, verification, and code generation for embedde d software development. To improve the performance of the code generated f rom Simulink models, state-of-the-art code generators employ various optim ization techniqu...\n\n\nZehong Yu, Zhuo Su, and Yu Jiang (Tsinghua Univer sity); Aiguo Cui (Huawei); and Rui Wang (Capital Normal University)\n----- ----------------\nCache-aware Task Decomposition for Efficient Intermitten t Computing Systems\n\nEnergy harvesting offers a scalable and cost-effect ive power solution for IoT devices, but it introduces the challenge of fre quent and unpredictable power failures due to the unstable environment. \n To address this, intermittent computing has been proposed, which periodica lly backs up the system stat...\n\n\nShuo Xu, Wei Zhang, Mengying Zhao, Zi meng Zhou, and Lei Ju (Shandong University)\n\nTopic: Embedded Systems\n\n Keyword: Embedded Software END:VEVENT END:VCALENDAR