Close

Session

Research Manuscript: Some Things are Best Left In/Near Memories
DescriptionIn/near-memory innovations for emerging workloads push the boundaries of performance and energy efficiency. The first two papers present accelerators for Open Modification Search and point cloud range search. The next three papers discuss graph mining, including an asynchronous callback interface for triangle counting, a data management system for regular path queries, and learned indexes for variable-length keys. The sixth paper presents hardware-managed DRAM caching using a Gaussian mixture model for CXL-enabled memory expansion. Finally, the last two papers introduce adaptive resource allocation for SmartNICs and a framework for optimizing PIM execution sequences using parallel data preparation.
Event TypeResearch Manuscript
TimeThursday, June 271:30pm - 3:30pm PDT
Location3003, 3rd Floor
Topics
Design
Keywords
In-memory and Near-memory Computing Circuits
Presentations
1:30pm - 1:45pm PDTEfficient Open Modification Spectral Library Searching in High-Dimensional Space with Multi-Level-Cell Memory
1:45pm - 2:00pm PDTCAMPER: Exploring the Potential of Content Addressable Memory for 3D Point Cloud Efficient Range Search
2:00pm - 2:15pm PDTSting: Near-storage accelerator framework for scalable triangle counting and beyond
2:15pm - 2:30pm PDTAccelerating Regular Path Queries over Graph Database with Processing-in-Memory
2:30pm - 2:45pm PDTLIVAK: A High-Performance In-Memory Learned Index for Variable-Length Keys
2:45pm - 3:00pm PDTICGMM: CXL-enabled Memory Expansion with Intelligent Caching Using Gaussian Mixture Model
3:00pm - 3:15pm PDTTIGA: Towards Efficient Near Data Processing in SmartNICs-based Disaggregated Memory Systems
3:15pm - 3:30pm PDTExecution Sequence Optimization for Processing In-Memory using Parallel Data Preparation