A 22nm 832Kb Hybrid-Domain Floating-Point SRAM In-Memory-Compute Macro with 16.2-70.2TFLOPS/W for High-Accuracy AI-Edge Devices.
Ping-Chun WuJian-Wei SuLi-Yang HongJin-Sheng RenChih-Han ChienHo-Yu ChenChao-En KeHsu-Ming HsiaoSih-Han LiShyh-Shyuan SheuWei-Chung LoShih-Chieh ChangChung-Chuan LoRen-Shuo LiuChih-Cheng HsiehKea-Tiong TangMeng-Fan ChangPublished in: ISSCC (2023)
Keyphrases
- floating point
- high accuracy
- fixed point
- square root
- artificial intelligence
- knowledge base
- instruction set
- sparse matrices
- memory requirements
- low power
- random access memory
- memory bandwidth
- power consumption
- interval arithmetic
- main memory
- general purpose
- random access
- memory space
- image processing
- fast fourier transform