Time | Event |
---|---|
9:00 - 9:10 | Welcome and Introduction Antonino Tumeo, John Feo, Vito Giovanni Castellana |
9:10 - 10:00 | Keynote 1 - Chair: Vito Giovanni Castellana (PNNL) Khronos SYCL: Heterogeneous Programming with Modern C++ and Machine Learning Michael Wong (Codeplay) |
10:00 - 10:30 | Coffee Break |
10:30 - 11:30 | Session 1: Distributed Systems and Irregular Applications - Chair: Ryan Friese (PNNL) Conveyors for Streaming Many-To-Many Communication Maley, DeVinney Extending a Work-Stealing Framework with Priorities and Weights Nakashima, Yoritaka, Yasugi, Hiraishi, Umatani RDMA vs. RPC for Implementing Distributed Data Structures (short) Brock, Chen, Yan, Owens, Buluç, Yelick |
11:30 - 12:30 | Session 2: Mixed Precision and new memory hierarchies for Irregular Applications - Chair: Sayan Ghosh (PNNL) A Mixed Precision Multicolor Point-Implicit Solver for Unstructured Grids on GPUs Walden, Nielsen, Diskin, Zubair Mixed-Precision Tomographic Reconstructor Computations on Hardware Accelerators Doucet, Ltaief, Gratadour, Keyes Metall: A Persistent Memory Allocator Enabling Graph Processing (short) Iwabuchi, Lebanoff, Gokhale, Pearce |
12:30 - 14:00 | Lunch break (on your own) |
14:00 - 14:50 | Keynote 2 - Chair: Marco Minutoli (PNNL) Sparse Linear Algebra in Facebook's Deep Learning Models Jongsoo Park (Facebook) |
14:50 - 15:00 | Session 3: Dealing with Irregular Algorithms - Chair: Vinay Amatya (PNNL) iPregel: Strategies to Deal with an Extreme Form of Irregularity in Vertex-Centric Graph Processing (short) Capelli, Brown, Bull |
15:00 - 15:30 | Coffee Break |
15:30 - 15:55 | Session 3 (continued): Dealing with Irregular Algorithms - Chair: Vinay Amatya (PNNL) Stretching Jacobi: A Two-Stage Pivoting Approach for Block-Based Factorization Thuerck |
15:55 - 16:40 | Session 4: Hardware evaluation and mechanisms for Irregular Applications - Chair: Tania Lorido Botran (PNNL) A Hardware Prefetching Mechanism for Vector Gather Instructions Takayashiki, Sato, Komatsu, Kobayashi Performance Impact of Memory Channels on Sparse and Irregular Algorithms (short) Green, Fox, Young, Shirako, Bader Cascaded DMA Controller for Speedup of Indirect Memory Access in Irregular Applications (short) Kashimata, Kitamura, Kimura, Kasahara |
16:40 - 17:30 | Debate - Moderator: Flavio Vella (Free University of Bozen) Proposition: "Programming abstractions can greatly simplify the expression of complex workflows of data analytic applications, but make it difficult to customize source code for heterogeneous computer systems. Unfortunately, it is an either or situation --- easy expression or fast performance" Panelists: Nesreen Ahmed (Intel), Jonathan Beard (ARM), Tyler Sorensen (Princeton University), Ana Lucia Verbanescu (University of Amsterdam) |