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:20260730T152637Z
LOCATION:
DTSTART;TZID=America/Los_Angeles:20260728T124400
DTEND;TZID=America/Los_Angeles:20260728T124500
UID:dac_DAC 2026_sess315_RESEARCH1322@linklings.com
SUMMARY:Keeping PIM Busy: Eliminating Execution Overheads for Full Through
 put
DESCRIPTION:Mun Seong Park, Jisung Pack, Junil Kim, Jun Sik Kim, Seok Youn
 g Kim, and Seon Kim (Korea University)\n\nLarge language models (LLMs), su
 ch as GPT-3 and Llama-2, impose extreme memory-bandwidth demands, creating
  severe data movement bottlenecks. Processing-in-Memory (PIM) mitigates th
 is by performing computations near memory; however, current PCIe-based PIM
  systems deliver only a small fraction of their theoretical throughput. On
  our multi-channel PIM emulation platform, PIM cores are active for only 6
 .55% of execution time, primarily due to (1) serialized host–device commun
 ication that limits channel-level parallelism, (2) insufficient request-ge
 neration capability in conventional DMA engines, and (3) per-bank microarc
 hitectures that serialize batch execution.\n\nWe address these bottlenecks
  with a co-designed memory system and microarchitectural solution: Channel
 -Level Burst (CL) for autonomous per-channel operand generation, PDMA for 
 high-throughput PIM-oriented request scheduling, and an integrated PIM cor
 e for enabling true multi-batch parallelism.\nAcross GEMM microbenchmarks 
 and four LLMs, CL provides 9.10x speedup, PDMA adds 29.6x, and multi-batch
  execution contributes 102.4x. Together, they deliver a 145.8x improvement
  over a baseline multi-channel PIM system, enabling PIM to surpass CPU thr
 oughput in memory-bound LLM decode kernels and approach the efficiency of 
 GPUs.\n\nTopics: Quantum\n\nTrack: DES6. Quantum Computing, Systems\n\nSes
 sion Chairs: Hokeun Kim (Arizona State University); BaekGyu Kim (Daegu Gye
 ongbuk Institute of Science and Technology); Xiang Chen (Peking University
 ); Sangyeob Kim (Yonsei University); Robert M. Radway (University of Penns
 ylvania); Prashanth Mohan (University of Pittsburgh); Xiaoxuan Yang (Unive
 rsity of Virginia, Stanford University); Ramtin Zand (University of South 
 Carolina); Hadi Kamali (University of Central Florida); Jiafeng Xie (Villa
 nova University); Prabuddha Chakraborty (University of Maine); Junichiro K
 adomoto (University of Tokyo); and Yosuke Ueno (RIKEN)\n\n
END:VEVENT
END:VCALENDAR