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:20210916T132456Z
LOCATION:
DTSTART;TZID=Europe/Stockholm:20210706T173000
DTEND;TZID=Europe/Stockholm:20210706T190000
UID:submissions.pasc-conference.org_PASC21_sess182_post152@linklings.com
SUMMARY:P26 - COSMA: Accelerating Electronic Structure Calculations withou
 t Changing the Code
DESCRIPTION:Poster\n\nP26 - COSMA: Accelerating Electronic Structure Calcu
 lations without Changing the Code\n\nKabic, VandeVondele\n\nA building blo
 ck of many Electronic Structure Calculations, that is often dominating the
  total runtime, is a distributed multiplication of dense matrices, or tens
 ors. As an example, when simulating 128 water molecules with CP2K code, us
 ing the Random Phase Approximation Method (RPA), dense matrix multiplicati
 on takes more than 90% of the total runtime in some configurations. Theref
 ore, using a more efficient matrix multiplication routine can have a signi
 ficant impact on the performance of many simulations. Currently, the faste
 st known algorithm for dense matrix multiplication is the recently develop
 ed COSMA algorithm, that is provably communication-optimal for all matrix 
 shapes and available memory constraints. With a newly developed ScaLAPACK 
 compatible-API, COSMA can be used just by linking to it, making its adopti
 on almost trivial. Here, we show the first results of using the COSMA algo
 rithm with CP2K, demonstrating a performance improvement without making an
 y change in the CP2K simulation code.
END:VEVENT
END:VCALENDAR