BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201804Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251118T080000
DTEND;TZID=America/Chicago:20251118T170000
UID:submissions.supercomputing.org_SC25_sess527_post154@linklings.com
SUMMARY:Accelerating Scientific Workflows with LLM-Driven Compiler Optimiz
 ations for Generated High-Performance Hardware
DESCRIPTION:Robert Ramstad, Nicolas Bohm Agostini, and Antonino Tumeo (Pac
 ific Northwest National Laboratory (PNNL))\n\nThe optimization of computat
 ion kernels is central to high performance computing, directly impacting a
 pplications from scientific computing to artificial intelligence (AI). In 
 experimental workflows with high-throughput or streaming data, software-on
 ly execution often becomes a bottleneck, motivating custom hardware accele
 rators. Field-programmable gate arrays and application-specific integrated
  circuits excel at these workloads by exploiting parallelism, pipelining, 
 and low latency. Yet, mapping optimized kernels to hardware with high-leve
 l synthesis (HLS) requires significant manual effort. To address this, we 
 propose a large language model (LLM)-driven optimization approach. Our met
 hod leverages the MLIR compiler infrastructure and modern LLMs’ capability
  to synthesize code to create tailored optimization strategies for hardwar
 e targets through HLS. This approach achieved 2.7x speedup for an electron
  energy loss spectroscopy autoencoder model targeting the Virtex UltraScal
 e+. These results show that LLM-driven optimization offers a low-effort, h
 igh-performance alternative to manual workflows, paving the way for agenti
 c AI in compilers and high performance computing.\n\nTag: Research & ACM S
 RC Posters\n\nRegistration Category: Technical Program Reg Pass\n\nSession
  Chairs: Kento Sato (RIKEN Center for Computational Science (R-CCS)); Chri
 s Schlipalius (Pawsey Supercomputing Research Centre; Commonwealth Scienti
 fic and Industrial Research Organisation (CSIRO), Australia); and Anja Ger
 bes (Georg-August-Universität Göttingen)\n\n
END:VEVENT
END:VCALENDAR