G10: Enabling An Efficient Unified GPU Memory and Storage Architecture with Smart Tensor Migrations.
Haoyang ZhangYirui Eric ZhouYuqi XueYiqi LiuJian HuangPublished in: CoRR (2023)
Keyphrases
- memory management
- real time
- storage devices
- memory capacity
- memory access
- data transfer
- heterogeneous computing
- higher order
- data storage
- limited memory
- management system
- secondary storage
- memory requirements
- high order
- data flow
- multithreading
- compute intensive
- memory storage
- random access
- diffusion tensor
- disk storage
- associative memory
- graphics processors
- single instruction multiple data
- storage media
- graphic processing unit
- memory bandwidth
- level parallelism
- gpu implementation
- processing elements
- data management systems
- computational power
- hardware implementation