BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T143139Z
LOCATION:B306
DTSTART;TZID=America/New_York:20241121T114500
DTEND;TZID=America/New_York:20241121T120000
UID:submissions.supercomputing.org_SC24_sess810_drs104@linklings.com
SUMMARY:High-Performance Computing Resilience Analysis Using Large Languag
 e Models
DESCRIPTION:Hailong Jiang (Kent State University)\n\nThis doctoral showcas
 e highlights three pivotal works conducted during my PhD that collectively
  advance the field of high-performance computing (HPC) resilience analysis
  using large language models (LLMs).\n\nThe first work introduces HAPPA, a
  modular platform for HPC Application Resilience Analysis. HAPPA integrate
 s LLMs to understand long code sequences, employing innovative code repres
 entation techniques to predict resilience accurately. Through the DARE dat
 aset, HAPPA demonstrates superior predictive accuracy over existing models
 , achieving a mean squared error (MSE) of 0.078 in Silent Data Corruption 
 (SDC) prediction, significantly outperforming the PARIS model.\n\nBuilding
  on this foundation, the second work investigates the resilience of loops 
 in HPC programs through a semantic approach. By analyzing the computationa
 l patterns known as the 13 dwarfs of parallelism, this study quantifies th
 e SDC rates for each pattern. Utilizing LLMs with prompt engineering, the 
 research identifies loop semantics, providing insights into which loops ar
 e more error-prone and enhancing the development of resilient HPC applicat
 ions.\n\nExpanding the scope further, the third work evaluates the capabil
 ities of LLMs in comprehending the syntax and semantics of Intermediate Re
 presentation (IR) code. The study conducts a comprehensive analysis using 
 models like GPT-4o, GPT-3.5, and CodeLlama. By performing tasks such as de
 compiling IR code, generating CFGs, and simulating IR code execution, the 
 research provides insights into the effectiveness of LLMs in handling low-
 level code analysis and their potential applications in program analysis.\
 n\nThese studies collectively demonstrate the potential of LLMs in enhanci
 ng the resilience of HPC applications through innovative analysis techniqu
 es and predictive modeling.\n\nRegistration Category: Tech Program Reg Pas
 s\n\nSession Chair: Alan Sussman (University of Maryland)\n\n
END:VEVENT
END:VCALENDAR