2017 International Conference On Computer Aided Design

The Premier Conference Devoted to Technical Innovations in Electronic Design Automation

November 13-16, 2017Irvine Marriott Irvine, CA

MP Associates, Inc.
MONDAY November 13, 10:30am - 12:30pm | Salon E
EVENT TYPE: REGULAR SESSION
SESSION 1A
Advanced Caching and In-Memory Processing
Moderator:
Miroslav N. Velev - Aries Design Automation, LLC
This session presents four papers to advance the design of cache memories and take advantage of in-memory processing to accelerate applications. The first paper builds an efficient hybrid cache for multicore processors by exploiting value locality. The second paper explores cache bypassing and partitioning on GPUs. The third paper proposes a virtual persistent cache concept to remedy the long latency in Host-Aware Shingled Magnetic Recording (HA-SMR) drives. The final paper present a memristor-based in-memory processing for accelerating object recognition tasks.

1A.1Leveraging Value Locality for Efficient Design of a Hybrid Cache in Multicore Processors
 Speaker: Amin Jadidi - Pennsylvania State Univ.
 Authors: Mohammad Arjomand - Georgia Institute of Technology
Amin Jadidi - Pennsylvania State Univ.
Mahmut Kandemir - Pennsylvania State Univ.
Chita Das - Pennsylvania State Univ.
1A.2Exploring Cache Bypassing and Partitioning for MultiTasking on GPUs
 Speaker: Yun (Eric) Liang - Peking Univ.
 Authors: Yun (Eric) Liang - Peking Univ.
Xiuhong Li - Peking Univ.
Xiaolong Xie - Peking Univ.
1A.3Virtual Persistent Cache: Remedy the Long Latency Behavior of Host-Aware Shingled Magnetic Recording Drives
 Speaker: Ming-Chang Yang - Chinese Univ. of Hong Kong & Academia Sinica and National Taiwan Univ.
 Authors: Ming-Chang Yang - Chinese Univ. of Hong Kong & Academia Sinica and National Taiwan Univ.
Yuan-Hao Chang - Academia Sinica
Fenggang Wu - Univ. of Minnesota, Twin Cities
Tei-Wei Kuo - National Taiwan Univ. & Academia Sinica
David H.C. Du - Univ. of Minnesota, Twin Cities
1A.4ORCHARD: Visual Object Recognition Accelerator Based on Approximate In-Memory Processing
 Speaker: Mohsen Imani - Univ. of California, San Diego
 Authors: Yeseong Kim - Univ. of California, San Diego
Mohsen Imani - Univ. of California, San Diego
Tajana Rosing - Univ. of California, San Diego