---

Welcome to PMBS 2018
12th November 2018
Location: D165

---

Session 1: HPC Interconnects
Chair: Stephen Jarvis

---

9:00 - 9:30
Improving MPI Reduction Performance for Manycore Architectures with OpenMP and Compression  [abstract] [presentation]

Hongzhang Shan, Samuel Williams
Lawrence Berkeley National Laboratory, Berkeley, CA

Calvin W. Johnson
San Diego State University, San Diego, CA

---

9:30 - 10:00
Exploring and Quantifying How Communication Behaviors in Proxies Relate to Real Applications  [abstract] 

Omar Aazi, Jeanine Cook, Courtenay Vaughan
Sandia National Laboratories, Albuquerque, NM

Jonathan Cook
New Mexico State University, Las Cruces, NM

---

10:00 - 10:30 Coffee Break

---

Session 2: Machine/Deep Learning
Chair: Sebastiano Fabio Schifano

---

10:30 - 11:00
Deep Learning at Scale on NVIDIA V100 Accelerators  [abstract] [presentation]

Rengan Xu, Frank Han, Quy Ta
Dell EMC, Austin, TX

---

11:00 - 11:30
Benchmarking Machine Learning Methods for Performance Modeling of Scientific Applications  [abstract] [presentation]

Preeti Malakar, Prasanna Balaprakash, Venkatram Vishwanath, Vitali Morozov, Kalyan Kumaran
Argonne National Laboratory, Lemont, IL

---

Session 3: Late-breaking Research and Preliminary Techniques
Chair: Prasanna Balaprakash

---

11:30 - 11:50
Algorithm Selection of MPI Collectives using Machine Learning Techniques  [abstract] 

Sascha Hunold
TU Wien, Vienna, Austria

Alexandra Carpen-Amarie
Fraunhofer ITWM, Kaiserslautern, Germany

---

11:50 - 12:10
miniVite: A Graph Analytics Benchmarking Tool for Massively Parallel Systems  [abstract] [presentation]

Sayan Ghosh, Ananth Kalyanaraman, Assefaw H. Gebremedhin
Washington State University, Pullman, WA

Manhantesh Halappanavar, Antonino Tumeo
Pacific Northwest National Laboratory, Richland, WA

---

12:10 - 12:30
Unified Cross-Platform Profiling of Parallel C++ Applications  [abstract] [presentation]

Vladyslav Kucher, Florian Fey, Sergei Gorlatch
University of Muenster, Muenster, Germany

---

12:30 - 14:00 Lunch

---

Session 4: System Evaluation
Chair: Jesmin Jahan Tithi

---

14:00 - 14:30
A Metric for Evaluating Supercomputer Performance in the Era of Extreme Heterogeneity  [abstract] [presentation]

Brian Austin, Chris Daley, Douglas Doerfler, Jack Deslippe, Brandon Cook, Brian Friesen, Thorsten Kurth, Charlene Yang, Nicholas J. Wright
Lawrence Berkeley National Laboratory, Berkeley, CA

---

14:30 - 15:00
Evaluating SLURM Simulator with Real-Machine SLURM and Vice Versa  [abstract] [presentation]

Ana Jokanovic, Marco D'Amico, Julita Corbalan
Barcelona Supercomputing Center, Barcelona, Spain

---

15:00 - 15:30 Coffee Break

---

Session 5: Performance
Chair: Steven Wright

---

15:30 - 16:00
Is Data Placement Optimization Still Relevant On Newer GPUs?  [abstract] [presentation]

Md Abdullah Shahneous Bari
Stony Brook University, Stony Brook, NY

Larisa Stoltzfus
University of Edinburgh, Edinburgh, UK

Pei-Hung Lin, Chunhua Liao, Murali Emani
Lawrence Livermore National Laboratory, Livermore, CA

Barbara Chapman
Stony Brook University, Stony Brook, NY
Brookhaven National Laboratory, Upton, NY

---

16:00 - 16:30
Approximating a Multi-Grid Solver  [abstract] [presentation]

Valentin Le Fèvre
École Normale Supérieure de Lyon, Lyon, France

Leonardo Bautista-Gomez, Osman Unsal, Marc Casas
Barcelona Supercomputing Center, Barcelona, Spain

---

Session 6: Simulation and Prediction
Chair: John Pennycook

---

16:30 - 17:00
Evaluating the Impact of Spiking Neural Network Traffic on Extreme-Scale Hybrid Systems  [abstract] [presentation]

Noah Wolfe, Mark Plagge, Christopher D. Carothers
Rensselaer Polytechnic Institute, Troy, NY

Misbah Mubarak, Robert B. Ross
Argonne National Laboratory, Lemont, IL

---

17:00 - 17:30
Automated Instruction Stream Throughput Prediction for Intel and AMD Microarchitectures  [abstract] [presentation]

Jan Laukemann, Johannes Hofmann
Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany

Julian Hammer, Georg Hager, Gerhard Wellein
Erlangen Regional Computing Center, Erlangen, Germany

---