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:20210527T134946Z
LOCATION:Analog 1
DTSTART;TZID=Europe/Stockholm:20200623T145000
DTEND;TZID=Europe/Stockholm:20200623T145500
UID:isc_hpc_ISC High Performance 2020_sess325_post106@linklings.com
SUMMARY:Accelerating Quicksilver – a Monte Carlo Proxy App on Multicores
DESCRIPTION:Research Poster\n\nAccelerating Quicksilver – a Monte Carlo Pr
 oxy App on Multicores\n\nTithi, Liu, Petrini\n\nQuicksilver is a proxy app
  that represents some elements of the Mercury workload used in the Lawrenc
 e Livermore National Laboratory. It solves a simplified dynamic Monte Carl
 o particle transportation problem using multigroup cross-sections. Quicksi
 lver attempts to replicate the memory access patterns, communication patte
 rns, and the branching or divergence of Mercury. By the virtue of captures
  key features of Mercury, lessons learned in optimizing Quicksilver have d
 irect implications on Mercury and as a result, quicksilver is being regula
 rly used in novel architecture co-design eﬀorts, influences hardwar
 e procurements at DOE.\n\nGetting high-performance in Quicksilver is quite
  challenging since its performance is dominated by latency bound table loo
 p-ups and branch divergence. The code is C++ STL heavy: implemented using 
 complex classes containing vectors and arrays; requiring multiple levels o
 f indirections for data lookups and frequent non-unit stride and random me
 mory accesses. Additionally, the control flow is dominated by branch diver
 gence causing low vectorization opportunities. Therefore, it is not surpri
 sing that the state-of-the-art GPUs have been reported to do only marginal
 ly well compared to general-purpose CPUs. In this research, we propose alg
 orithmic and engineering optimizations to Quicksilver that provide 1.8X sp
 eedup on state-of-the-art on Multicore CPUs compared to its original versi
 on. The major speedup comes from asymptotic cost improvement, automatic ve
 ctorizations, better data-reuse, and improved atomics. We are working on o
 ptimizing Quicksilver on various special-purpose accelerators.\n\nTag: Pre
 -Recorded
URL:https://2020.isc-program.com/presentation/?id=post106&sess=sess325
END:VEVENT
END:VCALENDAR

