Research / selected work
Hardware–software co-design for efficient computing.
My research asks how algorithms, software, and hardware can be designed together. The work spans sparse tensor algebra, accelerator dataflows, storage systems, and search over large hardware design spaces.
Open source
Current project
Research output
Selected work
- Accelerating deep learning in the post-Moore’s era of computingPhD dissertation · deep-learning acceleration→
- Accelerating Sparse Data Orchestration via Dynamic Reflexive TilingASPLOS · sparse tensor algebra→
- Mind MappingsASPLOS · accelerator mapping-space search→
- ExTensorMICRO · sparse tensor acceleration→
- BuffetsASPLOS · composable accelerator storage→
- MorphMICRO · flexible 3D-CNN acceleration→
- UCNNISCA · computation reuse in neural networks→