An Energy-efficient Floating-Point DNN Processor using Heterogeneous Computing Architecture with Exponent-Computing-in-Memory.
Juhyoung LeeJihoon KimWooyoung JoSangyeob KimSangjin KimDonghyeon HanJinsu LeeHoi-Jun YooPublished in: HCS (2021)
Keyphrases
- floating point
- heterogeneous computing
- instruction set
- compute intensive
- floating point arithmetic
- computer architecture
- level parallelism
- memory bandwidth
- floating point unit
- fixed point
- memory access
- parallel processors
- data mining
- memory management
- grid computing
- wireless sensor networks
- computer systems
- sensor networks
- high performance computing
- application specific
- parallel computing
- differential equations
- multithreading
- sensor nodes
- fine grained
- database management systems
- sufficient conditions
- peer to peer
- low cost