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:20190719T085745Z
LOCATION:HG E 3
DTSTART;TZID=Europe/Stockholm:20190614T133000
DTEND;TZID=Europe/Stockholm:20190614T140000
UID:submissions.pasc-conference.org_PASC19_sess164_msa147@linklings.com
SUMMARY:Simulating Diverse HEP Workflows on Heterogeneous Architectures
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Physi
 cs\n\nSimulating Diverse HEP Workflows on Heterogeneous Architectures\n\nL
 eggett, Shapoval\n\n<br />The next generation of HPC facilities show an in
 creasing use of GPGPUs and other accelerators in order to achieve their hi
 gh FLOP counts. In general, High Energy Physics computing workflows have m
 ade little use of GPUs due to the relatively small fraction of kernels tha
 t run efficiently on GPUs, and the expense of rewriting code for rapidly e
 volving GPU hardware. However, the computing requirements for HL-LHC are e
 normous, and it will become essential to be able to make use of supercompu
 ting facilities that rely heavily on accelerator technologies. ATLAS has b
 egun investigating strategies to efficiently schedule the offloading of co
 mputational components to accelerators. The same applies to LHCb, which, w
 hile sharing the same underlying framework as ATLAS, has considerably diff
 erent workflow. CMS's framework also has the ability to efficiently offloa
 d tasks to accelerators. But before investing heavily in writing many kern
 els for specific offloading architectures, we need to better understand th
 e performance metrics and throughput bounds of the workflows with various 
 accelerator configurations. This can be done by simulating a diverse set o
 f workflows, using real metrics for task interdependencies and timing, as 
 we vary fractions of offloaded tasks, latencies, and accelerator offloadin
 g parameters such as CPU/GPU ratios and speeds.
END:VEVENT
END:VCALENDAR