Plasticine - A Universal Data Analytics Accelerator
Abstract
We have developed hardware and software for a universal data analytics accelerator called Plasticine. Plasticine hardware is based on the novel concept of a reconfigurable dataflow architecture (RDA) which has both reconfigurable memories and reconfigurable compute. RDAs provide high energy efficiency without sacrificing programmability. We have fabricated a 7nm chip implementation of Plasticine I that provides significant performance and energy improvements compared to GPUs and FPGAs. Architecture studies for Plasticine II include support for dynamic on-chip networks, sparse-matrix computations and graph analytics. Plasticine software includes high-level and low-level compilers for converting TensorFlow machine learning applications into optimized configurations for Plasticine.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2020
- Accession Number
- AD1092828
Entities
People
- Kunle Olukotun
Organizations
- Stanford University