Previous-PADALs‎ > ‎PADAL17‎ > ‎


Wed August 2

09:00 Welcome

Session 1: Applications Session Abstracts

09:10 Data-related Optimizations for Various Applications on the Sunway supercomputing System Lin Gan
09:30 Benchmarking Stencil on GPUs Naoya Maruyama
09:50 Design and Optimization of a Multi-stencil CFD Solver Aparna Chandramowlishwaran
10:10 Need for data locality in deep learning Prasanna Balaprakash
10:30 Blocking Optimization for Tensor Decomposition Jee Choi
10:50 Break
11:15 Panel for Session 1 Panel Questions

12:00 Lunch

Session 2 Memory Technologies Session Abstracts

13:00 Exposing the locality of new memory hierarchies to HPC applications Brice Goglin
13:20 System abstractions to facilitate data movement in supercomputers with deep memory and interconnect hierarchy Francois Tessier
13:40 Bandwidth and Object Placement Studies on HBM Didem Unat
14:00 Exploiting Data Locality for Unified CPU/GPU Memory Space using OpenMP Lingda Li
14:20 Break
15:00 Deep Memory Abstractions: Beyond Caches and NUMA Nodes Swann Perarnau
15:20 Chapel's New Adventures in Intra-node Locality Brad Chamberlain
15:40 Extreme Heterogeneity and Flexibility Challenges for Data Locality Andrew A. Chien
16:00 Break
16:30 Panel for Session 2 Panel Questions
17:30 Adjourn
18:30 Workshop Dinner at La Petite Folie

Thu August 3

Session 3: Runtime Systems Session Abstracts
09:00 Kokkos Task-DAG: Memory Management and Locality Challenges Conquered H Carter Edwards
09:20 Asynchronous Task-Based Polar Decomposition on Single Node Manycore Architectures Hatem Ltaief
09:40 Addressing Data Locality via Adaptive Runtimes and Higher Level Abstractions Laxmikant (Sanjay) Kale
10:00 Data Locality - Old, New and Future in Legacy Fortran and C++ codes Martin Berzins
10:20 Break
11:00 HiHAT: Retargetable Infrastructure for Tomorrow's Systems CJ Newburn
11:20 Data Locality Challenges in Irregular Applications for Exascale Programing Min Si
11:40 Panel for Session 3 Panel Questions
12:30 Lunch

Session 4: Performance Modeling Session Abstracts
13:30 Performance model generation: Challenges and Approaches Boyana Norris
13:50 Coarse-Grained Data Locality Optimization Algorithm Mohamed Wahib
14:10 Unveiling the Interplay Between Global Link Arrangements and Network Management Algorithms on Dragonfly Networks Vitus J. Leung
14:30 Towards provably locality-enhancing work stealing algorithms Kenjiro Taura
14:50 Break

15:20 OpenMP, Unified Memory, and Prefetching Hal Finkel
15:40 Matrix-Free Techniques for Fast PDE Solvers Karl Rupp
16:00 Panel for Session 4 Panel Questions
17:00 Adjourn

Fri August 4

Session 5: Compilers and DSLs Session Abstracts
09:00 The Loop Chain Abstraction: A Demonstration Catherine Olschanowsky
09:20 How Smart Should Compilers Be When Handling Abstractions for Locality?
Michelle Strout 
09:40 Compiler-Based Code Generation for Stencil Computations with Bricked Data Layout Protonu Basu
10:00 Two-way Python-Fortran transpilation for performance and maintainability Mateusz Bysiek
10:20 DAPP - Data-Centric Parallel Programming for Shared and Distributed Memory Systems Tal Ben-Nun
10:40 Break
11:15 Panel for Session 5 Panel Questions
12:00 Adjourn