BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20241120T082409Z
LOCATION:HG E 3
DTSTART;TZID=Europe/Stockholm:20240605T090000
DTEND;TZID=Europe/Stockholm:20240605T093000
UID:submissions.pasc-conference.org_PASC24_sess131_msa195@linklings.com
SUMMARY:Enzyme.jl: High-Performance, Cross-Language, and Parallel Automati
 c Differentiation in Julia
DESCRIPTION:Minisymposium\n\nWilliam Moses (University of Illinois Urbana-
 Champaign)\n\nAutomatic differentiation (AD) is key to training neural net
 works, Bayesian inference, and scientific computing. Applying these techni
 ques requires rewriting code in a specific machine learning framework or m
 anually providing derivatives. This talk presents Enzyme, a high-performan
 ce automatic differentiation compiler plugin for the low-level virtual mac
 hine (LLVM) compiler capable of synthesizing gradients of programs express
 ed in the LLVM intermediate representation (IR). Enzyme differentiates pro
 grams in any language whose compiler targets LLVM, including C/C++, Fortra
 n, Julia, Rust, JaX, Swift, etc., thereby providing native AD capabilities
  in these languages with state-of-the-art performance. Unlike traditional 
 tools, Enzyme performs AD on optimized IR. We show that AD on optimized IR
  achieves a geometric mean speedup of 4.2x over AD on IR before optimizati
 on, and orders of magnitude speedups on GPU accelerator codes.\n\nThis tal
 k will discuss AD from the lens of Enzyme.jl, Julia bindings for Enzyme. W
 hile Enzyme is applicable to any LLVM-based programming language, working 
 within Julia presents several opportunities and challenges. Julia makes it
  easy to write generic code that can be automatically retargeted for any b
 ackend, without the programmer needing to also become an expert in these m
 odels. This flexibility, however, comes at a cost of just-in-time compilat
 ion, and garbage collection.\n\nDomain: Climate, Weather, and Earth Scienc
 es, Physics, Computational Methods and Applied Mathematics\n\nSession Chai
 rs: Samuel Omlin (ETH Zurich / CSCS); Ludovic Raess (University of Lausann
 e, ETH Zurich); and Michael Schlottke-Lakemper (University of Augsburg)
END:VEVENT
END:VCALENDAR