2019 International Conference On Computer Aided Design

The Premier Conference Devoted to Technical Innovations in Electronic Design Automation

November 4-7, 2019The Westin Westminster Westminster, CO

v class="event-details"> MP Associates, Inc.

WEDNESDAY November 07, 1:15pm - 3:15pm | Monte Carlo
EVENT TYPE: REGULAR SESSION

SESSION 9B
High-Throughput Computing and Efficient Data Movement
Moderator:
Jing Xie - Qualcomm Research
The papers in this session exploit code generation and transformation techniques to achieve high throughput and efficient data movement. The first paper presents a framework for efficiently mapping stencil computations to FPGA platforms while the second presents a compilation framework based on the polyhedral model that translates high-level arithmetic-intensive kernels to systolic arrays on FPGAs. The third paper extracts data access patterns from an application to partition memory and generate data reuse logic. The fourth paper proposes compiler-automated transformations to optimize data layout for streaming accesses in addition to parameter tuning methods for high performance and low energy.

9B.1*SODA: Stencil with Optimized Dataflow Architecture
 Speaker: Yuze Chi - Univ. of California, Los Angeles
 Authors: Yuze Chi - Univ. of California, Los Angeles
Jason Cong - Univ. of California, Los Angeles
Peng Wei - Univ. of California, Los Angeles
Peipei Zhou - Univ. of California, Los Angeles
9B.2PolySA: Polyhedral-Based Systolic Array Auto-Compilation
 Speaker: Jie Wang - Univ. of California, Los Angeles
 Authors: Jason Cong - Univ. of California, Los Angeles
Jie Wang - Univ. of California, Los Angeles
9B.3An Efficient Data Reuse Strategy for Multi-Pattern Data Access
 Speaker: Wensong Li - Fudan Univ.
 Authors: Wensong Li - Fudan Univ.
Fan Yang - Fudan Univ.
Hengliang Zhu - Fudan Univ.
Xuan Zeng - Fudan Univ.
Dian Zhou - Fudan Univ.
9B.4Optimizing Data Layout and System Configuration on FPGA-Based Heterogeneous Platforms
 Speaker: Hou-jen Ko - Purdue Univ.
 Authors: Hou-jen Ko - Purdue Univ.
Zhiyuan Li - Purdue Univ.
Samuel Midkiff - Purdue Univ.


* Indicates Best Paper Candidate