A massively parallel library for matrix and tensor algorithms

Event Sponsor: 
Mathmatics and Computer Science Division Seminar - LANS
Start Date: 
Oct 25 2017 - 10:30am
Building/Room: 
Building 240/Room 4301
Location: 
Argonne National Laboratory
Speaker(s): 
Edgar Solomonik
Speaker(s) Title: 
University of Illinois at Urbana-Champaign
Host: 
Anouar Benali

We present a framework of communication-avoiding parallel algorithms and a distributed-memory library of primitive routines for sparse and dense tensors. Computationally, a tensor represents a multidimensional array of data, while numerically tensors represent multilinear maps that operate and compose by means of contraction. We study the communication complexity of parallel algorithms for contractions of tensors with sparsity and symmetry. The proposed algorithms are implemented as part of Cyclops Tensor Framework. Cyclops supports contractions of tensors with user-defined element types and elementwise functions. We show performance results of application-codes achieving near 1 Petaflop/s performance using Cyclops as well as case-studies using symmetry, sparsity, and custom element-types.

Miscellaneous Information: 

Coffee and goodies will be served.

Please click below to add this event to your calendar.

[schedule.ics]

 

Upcoming Seminars:
 

  • November 1, 2017, "Applications of statistical approaches to the study of climate using large datasets" Jiali Wang, Asst. Atmospheric Scientist, Climate & Atmospheric Science Department, EVS Division, ANL, [more info]
  • November 8, 2017, "TBD" Peter E. Larsen, Assistant Computational Biologist, [more info]
  • November 15, 2017, "TBD" Wendy Di, Assistant Computational Scientist, MCS/ANL, [more info]
  • November 16, 2017, "Recent Progress in Optimal Power Flow: A Survey of Convex Relaxations and an Algorithm for Robust Feasibility" Daniel Molzahn, Computational Engineer, ES Division, Argonne National Laboratory,