BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160210Z
LOCATION:Track 7
DTSTART;TZID=America/New_York:20201119T140000
DTEND;TZID=America/New_York:20201119T143000
UID:submissions.supercomputing.org_SC20_sess293_drs109@linklings.com
SUMMARY:Algorithm Design for High-Performance CFD Solvers on Structured Gr
 ids
DESCRIPTION:Doctoral Showcase\n\nAlgorithm Design for High-Performance CFD
  Solvers on Structured Grids\n\nWang, Chandramowlishwaran\n\nComputational
  Fluid Dynamics (CFD) with structured grids has been widely used in many e
 ngineering disciplines including Aerospace Engineering and Vehicle Design.
  Current grid partitioners in CFD are biased towards minimizing the commun
 ication volume, and rely on graph partition methods for complex multi-bloc
 k structured grids. I propose novel algorithms specifically for multi-bloc
 k structured grids and a new cost function which not only accounts for the
  communication volume and edge cuts but also factors in the network's late
 ncy and bandwidth. Coupled with an MPI+OpenMP Jacobi solver, the new parti
 tioner outperforms the state-of-the-art by 1.5x to 3x on the Mira supercom
 puter. Current temporal tiling algorithms for stencil computations in CFD 
 are mainly designed for a single block grid on shared-memory machines, whe
 reas multi-block grids distributed across many nodes are the norm. I propo
 se a pipelined distributed stencil algorithm called Pencil. Pencil identif
 ies the optimal combination of MPI, OpenMP and temporal tiling, demonstrat
 ing up to 1.9x speedup over Pluto, the state-of-the-art polyhedral optimiz
 er, and extends the tiling to multi-block structured girds. Furthermore, P
 encil pipelines the communication and computation for overlap and obtains 
 1.33x to 3.41x speedup over MPI+OpenMP with space tiling on 32 Broadwell n
 odes, and excellent weak and strong scaling on 16 to 128 nodes. The state-
 of-the-art neural network-based CFD solvers are typically unable to predic
 t the solution for geometries unseen during the training. I am developing 
 a network that understands the local patterns of fluid dynamics. Coupled w
 ith an iterative scheme to combine local patterns, it can predict flow in/
 over arbitrary geometries.\n\nRegistration Category: Tech Program Reg Pass
END:VEVENT
END:VCALENDAR