BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Icosahedral Nonhydrostatic (ICON) model: Scalability on Massiv
 ely Parallel Computer Architectures - Florian Prill\,   (Deutscher Wetterd
 ienst (DWD))
DTSTART:20120928T110000Z
DTEND:20120928T112500Z
UID:TALK40247@talks.cam.ac.uk
CONTACT:Mustapha Amrani
DESCRIPTION:Simulation in numerical weather prediction and climate forecas
 ting has a fast-growing demand for memory capacity and processing speed. F
 or the last decade\, however\, computer technology has shifted towards mul
 ti-core chip designs while at the same time on-chip clock rates have incre
 ased only moderately. The parallel implementation of the ICON model's nonh
 ydrostatic dynamical core therefore follows a hybrid distributed/shared me
 mory approach\, based on the Message Passing Interface (MPI) and the OpenM
 P API. \n\nThe ICON code couples the different encapsulated components of 
 the earth system model\, e.g. dynamics\, soil\, radiation\, and ocean\, wi
 th high-level language constructs. Its communication characteristics and p
 rogramming patterns are designed to meet the main challenges in high perfo
 rmance computing\, i.e. load balancing\, cache efficiency\, and low-latenc
 y networking\, and take the unstructured triangular C-grid into account\, 
 which implies indirect addressing. Besides basic optimization strategies s
 uch as loop tiling and grid point reordering\, the implementation employs 
 special domain decomposition heuristics\, parallel range-searching algorit
 hms with logarithmic complexity\, and makes use of asynchronous I/O server
 s to deal with the potentially prohibitive amount of data generated by ear
 th system models. This facilitates the ICON code to extract an adequate le
 vel of performance on a wide range of HPC platforms\, targeting large scal
 ar cluster systems with thousands of cores as well as vector computers.\n\
 n
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR